IEEE 2017 NSS/MIC/RTSD ControlCenter

Online Program Overview Session: M-08

To search for a specific ID please enter the hash sign followed by the ID number (e.g. #123).

MIC Poster Session II

Shortcut: M-08
Date: Thursday, October 26, 2017, 13:40
Room: Grand Hall West
Session type: MIC Session


Click on a content entry to get a more detailed information.

Poster panel
(face) ID: 2

Poster Number:

Development of a 64ch SiPM-based TOF-PET Detector with High Spatial and Timing Resolutions Using Multiplexing Architecture (#1139)

M. Nakazawa1, T. Tsuda1, T. Furumiya1, S. Adachi1, J. Ohi1, K. Kitamura1

1 Shimadzu Corporation, Technology Research Laboratory, Hikaridai 3-9-4, Seika-cho, Soraku-gun, Kyoto, Japan


SiPM has excellent potential as TOF-PET detector because of its fast response. One-to-one connecting of a SiPM and a fast RF amplifier is preferable for maximizing timing performance. ASIC is one solution to realize it in a compact front-end module of a detector, but a huge development cost and a long development time are required.We developed a 64ch SiPM-based TOF-PET detector using the SiPM multiplexing architecture in which SiPMs output signals are multiplexed into one timing signal for reducing timing channels. It causes the degradation of timing resolution due to dark noise and parasitic capacitance of SiPMs, but it can be improved by optimizing the parameter of AC coupling. Timing signal is generated by the leading edge method. Position and energy signals are generated by Anger logic and summing circuits with slow (video rate) amplifiers. Each SiPM is connected the capacitor for timing signals and resistor matrix for position signals. The proposed circuit makes it possible to read the timing signal and position (and energy) signals simultaneously without interference because of the difference in bandwidth of these signals.The developed detector applies four SiPM arrays (S13361-3050-NE, 4 x 4 ch of 3 x 3 mm2, Hamamatsu) with a scintillator block consists of 2.1 mm x 2.1 mm x 15 mm LFS crystal in a 12 x 12 array. The front-end module is composed of two 32ch SiPM signals multiplexing circuits with fast amplifiers and discriminators for timing signal.Coincidence timing resolution of the developed detectors was estimated to be 286 ps (FWHM), and energy resolution was measured as 13.6 %. Moreover, all the crystals (12 x 12) were clearly separated on the 2-D position histogram. These results suggest that our new detector is practical for TOF-PET systems, having excellent spatial and timing resolutions.

Keywords: TOF-PET, SiPM, Multiplexing, Timing Resolution, Spatial Resolution
Poster panel
(face) ID: 5

Poster Number:

Comparison of Various Time Pick-off Techniques for ToF PET (#1291)

C. J. Thompson1, E. Cashman2, J. Moreau1, A. L. Goertzen3

1 Montreal Neurological Institute, Brain Imaging Centre, Montreal, Québec, Canada
2 SensL Tehcnoligies Ltd., Cork, Ireland
3 University of Manitoba, PET/CT Program, Winnipeg, Manitoba, Canada


SiPMs have been shown to be useful in PET due to their low noise and compact size. When applied in a ToF-enabled PET scanner which digitizes the anode signals the precision of the time stamp is limited by the sampling frequency, F, to √2/F unless some form of fitting to several samples is applied.

We have evaluated several time-stamping methods to a data set of 10k coincident events collected from 5x5x5 mm LYSO crystals coupled to SensL 6x6 mm SiPMs. The third, “fast” output of these devices was amplified so as not to saturate, and sampled at 5 GHz with a DRS4 waveform digitizer. The energy resolution was  ~11% FWHM. An energy window covering the photo-peak was applied. In another experiment a source was placed at three different locations 25 mm apart and 10k events were acquired.

When the onset was fitted to spline and the optimal threshold was chosen, the timing resolution was 240 psec. However, when this fast signal was “integrated” by summing the sample’s values and fitting samples close to the onset, the FWHM improved to 194 psec. Another method, where a linear fit of several points at the maximum slope of the leading edge was extrapolated to the baseline gave the best result of 270 psec when samples from 10 to 25 % of the peak amplitude were used. The timing response curves from the three source positions demonstrate that when the sources are 50 mm apart, the data are resolved to 15% of the peak height, but when the sources are 25 mm apart the curves are resolved to the 70% level.

Keywords: ToF-PET, Human-brain-imaging, Time pickoff stragegies
Poster panel
(face) ID: 8

Poster Number:

Development of a DOI-based Compton camera for nuclear medicine application (#1378)

S. Takyu1, S. Liprandi2, F. Nishikido1, A. Mohammadi1, E. Yoshida1, S. Aldawood2, 3, T. Binder2, M. Mayerhofer2, R. Lutter2, I. I. Valencia Lozano2, G. Dedes2, K. Kamada4, K. Parodi2, P. G. Thirolf2, T. Yamaya1

1 National Institutes for Quantum and Radiological Science and Technology (QST), National Institute of Radiological Sciences (NIRS), Chiba, Japan
2 Ludwig-Maximilians-Universität München (LMU), Department of medical physics, Munich, Germany
3 King Saud University, Department of physics and astronomy, Riyadh, Saudi Arabia
4 C&A corporation, Sendai, Japan


We have developed depth-of-interaction (DOI) detectors for PET which can identify the three-dimensional interaction position of gamma-rays in the crystal block. In this paper, we upgrade our DOI absorber detector to a Compton camera, which is expected to enable simultaneous imaging of multiple nuclides and detection of prompt gamma rays in particle therapy. On the other hand, limited energy resolution would limit the performance of Compton imaging. Therefore, we develop a single-layer position sensitive scatterer detector with better energy resolution. At first, we compared the performance of three segmented position sensitive detectors constructed from Gd3Al2Ga3O12(Ce) (GAGG, C&A Corporation) crystal arrays with a 12 x 12 array of 2 x 2 x 6 mm3 crystals, a 15 x 15 array of 1.45 x 1.45 x 6 mm3 crystals, a 22 x 22 array of 0.9 x 0.9 x 6 mm3 crystals, and an MPPC SiPM module (MPPC: 3 x 3 mm2, 8 x 8 array, S13361-3050AE-08, module: C12680-4875, Hamamatsu Photonics K.K.). As the result, we chose the detector with the 22 x 22 GAGG array which has the best position resolution. As an absorber, we used a 4 layer DOI detector using segmented LYSO crystals with 32 x 32 x 4 layers of 1.46 x 1.46 x 4.5 mm3 and a 256 channel flat panel photomultiplier tube (H9500, Hamamatsu Photonics K. K.). In the Compton setup, the center of the GAGG detector was placed at a distance of 50 mm from the 137Cs point source. The center of the 4 layer DOI detector was placed at distances of 50 mm and 200 mm from the center of the GAGG detector. Compton event data in the geometry with different distances of scatterer and absorber were obtained by analyzing the coincidence data in both detectors. The two Compton images were obtained by the back projection of Compton cones. The Compton image obtained for the placement at the 200 mm distance showed better spatial resolution than for the placement at the 50 mm distance. Detailed results with respect to the Compton imaging performance will be presented.

Keywords: depth of interaction, compton imaging, position sensitive scintillation detector, DOI
Poster panel
(face) ID: 11

Poster Number:

Positron annihilation detection towards the limits with 6-sided digital SiPM readout of monolithic scintillators (#1718)

S. E. Brunner1, F. Dachs2, M. Gruber2, A. Hirtl1, K. Kamman3, D. R. Schaart1

1 Delft University of Technology, Radiation Science & Technology, Delft, Netherlands
2 Vienna University of Technology, Institute for Atomic and Subatomic Physics, Vienna, Austria
3 Delft University of Technology, DEMO, Delft, Netherlands


Clinical TOF-PET detectors based on L(Y)SO:Ce crystals and silicon photomultipliers (SiPMs) are currently replacing PMT-based devices due to their advantageous characteristics, e.g., excellent timing performance, compactness, high granularity, and insensitivity to magnetic fields. State-of-the-art TOF-PET systems are usually constructed using arrays of segmented crystals. The spatial resolution of such detectors typically equals the segmentation pitch. Recent SiPM-based commercial TOF-PET systems reach coincidence resolving times (CRT) between 325 ps and 400 ps FWHM. Demonstrator PET systems with CRTs between 200 ps FWHM and 250 ps FWHM have also been built.

A different approach is based on large, monolithic scintillators coupled to an array of photodetectors. Besides improved sensitivity, this approach enables intra-crystal reconstruction of the point of interaction of the 511 keV annihilation photons, thus providing depth-of-interaction (DOI) information. Recently, a spatial resolution of 1.1 mm FWHM, a DOI resolution of 2.4 mm FWHM and a CRT of 147 ps FWHM could be demonstrated. This was achieved by coupling two digital SiPM arrays (i.e. the Philips digital photon counter, DPC) to a 32 mm x 32 mm x 22 mm monolithic LYSO:Ce scintillator.

In the PALADIN (positron annihilation beyond the limits) project, we are exploring the detector performance when attaching a DPC array to each of the 6 surfaces of a monolithic scintillator cube with a size of 32 mm x 32 mm x 32 mm, targeting at a CRT of ~100 ps FWHM, a DOI-resolution of about ~1 mm, and a detection efficiency of up to ~90%. The PALADIN detector provides a total of 96 time stamps and 384 photon counts per event.

Here, we present first experimental results obtained with this detector. We furthermore compare the results with Monte Carlo simulation studies and discuss aspects of the technical realization of the detector, including the implementation of non-standard readout electronics for the DPCs.

Keywords: time-of-flight, positron emission tomography, silicon photomultipliers, monolithic scintillators, SiPM, TOF-PET
Poster panel
(face) ID: 17

Poster Number:

Point-Source Measurements using a Dome Shaped CsI Detector for Cardiac SPECT application (#2156)

J. Dey1, N. Bhusal1, K. Matthews1, P. E. Maggi1, M. More2, C. Brecher2, V. Nagarkar2

1 Louisiana State University, Baton Rouge, Louisiana, United States of America
2 Radiation Monitoring Devices, Watertown, Massachusetts, United States of America


We present feasibility studies of an economically built curved CsI dome for a high performance Cardiac SPECT application. The dome CsI was 99% of standard CsI density and mechanically stable. The light yield was 56% of a standard CsI crystal. Subsequently we successfully performed point-source position-sensitive scintillation light measurement experiments on a CsI:Tl domed sample, with a simple fiber-optic readout. The economical technique of building the detector shown here will be investigated further over the long term future.

Keywords: Curved Csl, Cardiac SPECT, Curved Scintillators
Poster panel
(face) ID: 20

Poster Number:

Composite scintillators for X-ray imaging (#2238)

A. Boyarintsev1, A. Bobovnikov1, A. Gektin1, I. Gerasymov1, T. Sibilieva1, S. Kovalchuk1, T. Nepokupnaya1, Y. Onufriyev1, A. Opolonin1

1 Institute for Scintillation Materials, National Academy of Sciences of Ukraine, Kharkov, Ukraine


Search of cheap scintillators for conventional x-ray imaging detectors is an actual task. The development of x-ray detectors based on composite scintillators is a cost-efficient solution that allows producing inexpensive large area detectors. This work was devoted to develop design of soft x-ray detector based on thin composite film applied directly onto a CCD or CMOS device. The methods of increasing x-ray absorption and light output for detectors based on scintillation powders were discussed.

GAGG:Ce and ZnSe were selected for composite film fabrication. GAGG:Ce powders were synthesized by solid-phase method. Concentration of Ce3+ was 0.1-0.9 at.%. Granules of crushed GAGG:Ce and ZnSe single crystal were used in films fabrication too. The average size of granules was 1 μm. Composite films were fabricated using polysiloxane elastomer as a binder. PVC and PE substrate were selected for composites than 500 μm thick.               

Luminescence and scintillation properties of 0.5 mm thick GAGG:Ce films were studied and compared with the same data for 1 mm thick GAGG:Ce single crystal. The optimum Ce3+ concentration in scintillator was determined. Light output of 500 μm thick GAGG:Ce films were the similar with 1 mm thick single crystal  when irradiated of 90 keV X-ray.

Spatial frequencies for 80 μm thick GAGG:Ce films were up to 9 lp/mm when irradiated of 20-80 keV X-ray. The same parameters for 100 μm and 250 μm thick ZnSe films were up to 8 lp/mm and 6 lp/mm, respectively. The results good correlate with data for structured CsI:Tl film, applied for dental imaging, non destructive testing and crystallography [1].

Use of structured scintillators and light guides is the way to increase x-ray absorption and light collection in thick composite films. Development of design of x-ray detector for non-destructive testing and medical imaging (for the source of 60-300 keV) is in progress.

  1. V.V. Nagarkar, T.K. Gupta, S.R. Miller et al. IEEE Trans. Nucl. Sci., 45 (1998) 492-496.
Keywords: composite scintillator, soft x-ray detector, medical imaging
Poster panel
(face) ID: 23

Poster Number:

Development of a detector module for Time-of-Flight PET with improved timing performance (#2340)

T. Niknejad1, R. Borrego2, R. Bugalho3, A. Di Francesco1, L. Ferramacho3, C. Leong3, M. Rolo4, J. C. Silva1, 3, R. Silva1, 3, M. Silveira3, S. Tavernier3, 5, J. Varela1, 3

1 Laboratory of Instrumentation and Experimental Particle Physics, Lisbon, Portugal
2 Instituto Superior Técnico, Lisbon, Portugal
3 PETsys Electronics, Oeiras, Portugal
4 INFN, Torino, Italy
5 Vrije Universiteit Brussel, Brussels, Belgium


We are developing a detector module with depth of interaction and improved timing performance for whole-body and brain Time-of-Flight PET scanners (TOF-PET). This detector module is based on a pixelated LYSO crystal with pixel size of 3 x 3 x 20 mm3   in  one to one coupling with pixelated SiPM and readout with TOFPET2,  a high performance front-end ASIC with TDC capability (time binning of 30ps). The Depth-of-Interaction (DOI) information is estimated using the light sharing method previously invented in our group, by having a piece of light guide on top the crystal and optically depolish the lateral surfaces of the crystal. The time difference between the signal observed in two detectors for a given positron annihilation depends on DOI of the interaction in case of long crystal. In this new study this DOI information is being used to improve the coincidence timing resolution (CTR). The preliminary study with TOFPET1, the first version the ASIC, and Hamamtsu SiPM have shown an improvement on the average CTR for 511 keV photon pairs from 22Na from 382 ps to 318 ps FWHM. The measurements will be redone with the new version of the ASIC (TOFPET2) with which we are aiming at having a CTR of less than 200 ps

Keywords: PET, Time of flight, Depth of interaction, SiPM, LYSO
Poster panel
(face) ID: 26

Poster Number:

Investigation of a scalable phoswich crystal design for a TOF-DOI PET detector (#2478)

C. - M. Chang1, B. J. Lee2, I. Sacco3, C. S. Levin4

1 Stanford University, Applied Physics, Radiology, Stanford, California, United States of America
2 Stanford University, Mechanical Engineering, Radiology, Stanford, California, United States of America
3 Stanford University, Radiology, Stanford, California, United States of America
4 Stanford University, Radiology, Physics, Electrical Engineering, Bioengineering, Stanford, California, United States of America


We are developing a phoswich PET detector capable of measuring both 511 keV photon time-of-fight (TOF) and depth-of-interaction (DOI). In our previous work, we have demonstrated the effectiveness of using time-over-threshold (ToT) for pulse shape discrimination in a two-layer TOF phoswich detector. However, in that work, the LSO:Ce,Ca (calcium co-doped lutetium oxyorthosilicate) crystals used to construct the phoswich detectors and the polished surface treatment of the crystals may not be realistic for scaling to a full clinical PET system. Therefore, in this work, we investigated the DOI and timing performance of two-layer phoswich detectors made of commercially manufacturable, non-polished (as-cut and chemically etched) LSO-only crystals. Eight two-layer phoswich detector configurations comprising a mixture of combinations of 2 LSO crystal types (fast and slow, both 3.2 x 3.2 x 10 mm3), 2 surface treatments (as-cut and etched) and 2 orientations (fast crystal coupled to SiPM and slow crystal coupled to SiPM) were tested. Two different types of pulse shape discrimination, ToT and pulse integral, to identify hits in the two phoswich layers were used. The relative layer separation accuracy for the ToT spectra was 5%-24% better than for the pulse integral spectra. To evaluate the effect of DOI measurement on coincidence timing performance, layer separation was performed with pulse integral except for the detector configurations that could not be separated by pulse integral, for which ToT was used. The coincidence timing resolution (CTR) improved by 2%-20% with DOI measurement, depending on the surface treatment. A CTR of ~ 200 ps FWHM was achieved for all phoswich detector configurations. Phoswich detectors made of fast (as-cut) and slow (etched) crystals with the slow crystal coupled to SiPM showed the largest layer separation (with a figure of merit ranges from 0.8 to 1.1) among the phoswich detector configurations studied in this work.

Keywords: time-of-flight, TOF, depth-of-interaction, DOI, phoswich, PET, detector, time-over-threshold, module, design, pulse shape, TOT, ToT
Poster panel
(face) ID: 29

Poster Number:

Broadband X-ray Fluorescence Emission Tomograph (#2654)

J. George1, M. C. Veale2, M. D. Wilson2, P. Seller2, L. - J. Meng1

1 University of Illinois at Urbana-Champaign, Nuclear, Plasma and Radiological Engineering, Urbana, United States of America
2 Science and Technology Facilities Council, Rutherford Appleton Laboratory, Oxford, Oxfordshire, United Kingdom of Great Britain and Northern Ireland


In this presentation, we will discuss the experimental results from broadband X-ray fluorescence emission tomography (XFET) imaging system that utilize a newly developed ultrahigh spectral resolution CdTe detector for fluorescence X-rays of energy 5 keV~100 keV.

In this study, we will focus on the feasibility of using this system to monitor the delivery of X-ray induced and nanoparticle-mediated photodynamic therapy (X-PDT). X-PDT, a collimated external X-ray beam is used to irradiate the target region filled with nanoparticles to induce local therapeutic effect through photosensitization, thermal ablation or other effects. During the X-PDT delivery process, the interactions between the externally delivered X-rays and metal atoms encapsulated in PDT agents could generate fluorescence X-rays. Detecting these X-ray signals could allow us to build up a precise 3D distribution of the specific metal in the target area and help to confirm the exact delivery of the therapeutic agents. this work, we focus on PDT agents containing high-Z metal elements, such as Au, Hf, La, and potentially Pt and Gd. The choice of high-Z elements would greatly increase the selective absorption of the excitation X-rays, and also lead to higher-energy fluorescence X-rays (68.8 keV for Au K-a), 43 keV for Gd K-a, 33.4 keV for La-K-a, and 55.8 keV for Hf K-a) that could penetrate a substantial thickness in tissue, which makes it possible for using XFET imaging to monitor the therapeutic delivery in deep tissue.

We will experimentally evaluate both spectral and imaging performance of multiple CdTe detector modules installed around the Xradia Bio MicroCT (MicroXCT-400) as a partial-ring broadband XFET system. We will carry out detailed studies to access spectroscopy capability of the detector for detecting and identifying XF photons, as well as the tomographic imaging performance of an experimentally simulated full-ring XFET system based on the HEXITEC detector modules.

Keywords: X-ray Fluorescence Emission Tomography, X-ray Induced and Nano-Particle Mediated Radiation Therapy
Poster panel
(face) ID: 32

Poster Number:

Depth-of-interaction estimation for a monolithic NaI(Tl) gamma detector with PMT-readout (#2892)

B. Wang1, W. van der Wal1, J. G. Ohm1, F. J. Beekman1, 2, M. C. Goorden1

1 Delft University of Technology, Section Radiation, Detection & Medical Imaging, Delft, Netherlands
2 MILabs B.V, Utrecht, Netherlands


PMT-based Anger cameras are still predominant in SPECT imaging as they provide satisfactory lateral spatial resolution for many applications and are very cost-effective. A limitation of Anger camera is that it does not provide any information on the depth-of-interaction (DoI) of gamma photons in the scintillator. For conventional SPECT tracers that emit gamma photons in the 30-200 keV energy range, this effect is very small as these photons mainly interact in the top of the scintillator. However, in the VECTor pinhole PET-SPECT scanner, a much wider energy range (30-600 keV) is considered. Here we test in a simulation study how well DoI can be estimated by a maximum likelihood algorithm in an Anger camera with optimized lightguide thickness.

A gamma detector employing a 19 mm-thick monolithic NaI(Tl) scintillator is attached to lightguides of different thicknesses (16 mm, 10 mm, 7 mm, 4 mm, 1 mm). Two readout set-ups based on 3 inch and 2 inch PMTs are considered. The lateral and DoI spatial resolution for 511 keV gamma photons of these designs are tested with the well-validated Monte Carlo simulation toolkit GATE. Simulation results show that 3 inch PMT and 7 mm lightguide combination gives about 4 mm lateral resolution and about 6 mm DoI resolution, which is sufficient for our application. Better results can be obtained with 2 inch PMTs combined with a 4 mm thick lightguide. In all cases, there is a trade-off between lateral resolution and DoI resolution when varying the lightguide thickness. To conclude, a maximum likelihood algorithm in combination with a relatively thin lightguide allows obtaining DoI information from a conventional Anger camera.

Keywords: Anger camera, depth of interaction, gamma detector, photomultiplier tube, lightguide
Poster panel
(face) ID: 35

Poster Number:

Development of Cherenkov Detectors for TOF PET (#3011)

V. Sharyy1, C. Canot1, M. Alokhina1, 3, O. Kochebina1, 2, P. Abbon1, S. Jan2, G. Tauzin1, D. Yvon1

1 CEA, IRFU, Gif-sur-Yvette, France
2 CEA, IMIV, SHFJ, Orsay, France
3 Taras Shevchenko National University of Kyiv, Nuclear Physics Department, Kyiv, Ukraine


In this presentation we are investigating the feasibility of the Cherenkov detectors to enhance the TOF performance of the PET scanners.

We discuss the results of the experimental tests and simulation expectations for two types of the Cherenkov detectors.

CaLIPSO project investigates the possibility to build the brain and pre-clinical PET device with a high spatial resolution of the order of 1 mm3.

It uses  an innovative liquid, TriMethyl Bismuth (TMBi), as the photon conversions media and the Cherenkov radiator.

The second project, PECHE, develops the full body PET scanner with the enhanced TOF resolution and uses as a Cherenkov radiator the crystalline lead fluoride. In both project we are using the micro-channel-plate PMT to obtain the highest possible resolution in time. We present the new experimental results on 511 keV photon detection using the 22Na radioactive source.  Our results shows the possibility to detect the Cherenkov lights with a good efficiency, of the order of 25% - 40% in both configurations. We also present the measurement of the time resolution and compare it with the dedicated Geant4 simulation. We extrapolate the measured performance of the detection modules to the full scale brain-oriented and full body PET scanners using Gate simulation. The estimated spatial resolution, coincident resolving time and noise equivalent count rate demonstrate a good potential for the TOF Cherenkov PET.

Keywords: Cherenkov process, TOF PET, MCP-PMT, TMBi, PbF2
Poster panel
(face) ID: 38

Poster Number:

Progress report for an accurate PET detector based on SiPMs and the TOFPET ASIC (#3040)

A. Aguilar1, E. Lamprou1, A. González-Montoro1, J. M. Monzo1, G. Cañizares1, L. F. Vidal1, L. Hernandez1, S. Iranzo1, S. Sánchez1, F. Sánchez1, A. J. González1, J. M. Benlloch1

1 Institute for Instrumentation in Molecular Imaging, i3M, Centro Mixto CSIC-Universitat Politecnica de Valencia, Valencia, Spain


In this contribution, we present a progress report of a study carried out with large SiPM arrays (5x5 cm2), read out with a commercially available ASIC and coupled to a monolithic LYSO crystal, to explore its potential use in PET systems enabling TOF capabilities. In previous works we were showing the performance of such an ASIC-based readout (TOFPET, PETsys Electronics) when using 3x3mm2 MPPCs (Hamamatsu Photonics). A sub-optimization in terms of temperature stability and bias dependency, limited our expected performance. The present work shows results using SiPM arrays from SensL with smoother bias and temperature dependencies. This feature is crucial when the scintillation light is shared among several SiPMs from which the photon impact position is determined.

‪The SiPMs and readout electronics have been evaluated first using a small size 1.6 mm (6 mm height) crystal arrays. Here, all pixels were well resolved, showing also a good energy resolution of nearing 20% (with ToT methods) for the 511 keV photons. Several parameters have been scanned to find the optimum performance, obtaining a CRT as good as 330 ps FWHM. The ASIC channel homogeneity was tested by measuring the CRT with one-to-one SiPM-crystal coupling. Average value of 470 ps FWHM (6x6 mm2 crystals and SiPMs) have been found for 9 direct pairs, with a standard deviation of 13 ps.

So far, when using a black-painted monolithic block, we found a spatial resolution of about 3 mm without correcting for the source size or carrying out mechanical and electronic collimations. Energy resolution appears to be slightly above 20% but again without ToT calibration yet. The analysis of CRT results using these scintillator blocks are still undergoing, but we aim at reaching CRT results in the limit of the present ASIC.

Keywords: PET, SiPM, ASIC, TOFPET, monolithic
Poster panel
(face) ID: 41

Poster Number:

Development of a portable gamma imaging system for absorbed radiation dose control in targeted radionuclide therapy (#3223)

C. Trigila1, M. - A. Verdier2, A. Desbrée3, G. Hull4, C. Esnault1, L. Pinot1, Y. Charon2, M. - A. Duval1, L. Ménard2

1 Univ. Paris Sud et Paris Diderot, IMNC CNRS/IN2P3, Orsay, France
2 Univ. Paris-Diderot, Sorbonne Paris Cite, IMNC CNRS/IN2P3, Orsay, France
3 Laboratory of Internal Dose Assessment, IRSN, Fontenay-aux-roses, France
4 Univ. Paris-Sud, Universite Paris-Saclay,, Institut de Physique Nucléaire, CNRS-IN2P3, Orsay, France


Targeted radionuclide therapy is the most used treatment modality against malign and benign diseases of thyroid. The large heterogeneity of absorbed doses in patients and the range of effects observed state that an individualized dosimetry is essential for optimizing this therapy. The goal of the project is to strengthen the control of the dosed delivered to organs during targeted radionuclide therapy of benign thyroid diseases, providing a novel mobile gamma imaging device specifically dedicated to measurements of the bio-distribution and kinetics of the radio-tracer at the patients’s bedside.  

We report here the preliminary experimental and theoretical studies aiming to optimize the detection head of the camera, that consists of a parallel-hole high-energy tungsten collimator, coupled to a continuous inorganic scintillator, readout by an array of Silicon Photomultiplier (SiPM) detectors. Experimental characterizations is focused on the choice of the best scintillator-photodetector assembly, in terms of spatial and energy performances, among different promising fast inorganic crystals (CeBr3, LYSO, LFS and GAGG) which covered distinct and complementary ranges of features in term of full energy peak efficiency, light yield, energy linearity and cost. They are compared with  LaBr3, which is considered to be the goal standard for its high energy and spatial features.

The theoretical study relies both on the design of the collimator with analytical models and on the optimization of the overall camera with Monte Carlo simulations for the dosimetry of thyroid diseases.

Keywords: Targeted radionuclide therapy, mobile gamma imaging device, thyroid
Poster panel
(face) ID: 44

Poster Number:

Improved Spatial Resolution and Resolution Uniformity of Monolithic PET Detector by Optimization of Photon detector Arrangement (#3584)

Z. Lyu1, 2, P. Fan1, 2, T. Xu1, 2, R. Wang1, 2, Y. Liu1, 2, S. Wang1, 2, Z. Wu1, 2, T. Ma1, 2

1 Tsinghua University, Department of Engineering Physics, Beijing, China
2 Ministry of Education (Tsinghua University), Key Laboratory of Particle & Radiation Imaging, Beijing, China


Compared to pixelated detector, monolithic detector has better performance in sensitivity, light collection efficiency and cost. In PET detector, photon detector arrangement is of critical importance, which can influence detector performance significantly. In this study, three monolithic PET detector designs with different photon detector arrangement are proposed to improve spatial resolution and resolution uniformity.

The PET detector is based on a 21×21×21 mm3 LYSO crystal and SiPM (SensL MicroFJ-30035). To improve detector performance with a reasonable number of SiPM, we proposed three different photon detector arrangements: Dual Ended (top and bottom: 5x5 array), Flexible Design A (top and bottom: 4x4 array, lateral: 2x2 array) and Flexible Design B (top and bottom: 3x3 array, lateral: 2x4 array). The Single Ended and Six Ended arrangement were also evaluated as reference. The solid angle criterion (SAC) is designed to evaluate spatial performance with less computation amounts, comparing with Cramer-Rao lower bound (CRLB). The Single Ended, Dual Ended and Flexible Design B detectors were realized to be evaluated in the preliminary experiment.

The simulation results show that mean CRLB of Dual Ended, Flexible Design A and Flexible Design B are respectively 0.362, 0.363 and 0.325 mm in x/y direction, and 0.316, 0.331 and 0.325 mm in z direction. The SAC shows a strong correlation with CRLB. The preliminary experiment shows the spatial resolution of Flexible Design B (x/y: 2.88mm, z: 5.16mm) is best, comparing with Single Ended (x/y: 3.21mm, z: 5.88mm) and Dual Ended (x/y: 3.05mm, z: 6.23mm). Flexible Design B has best resolution uniformity in both simulation and experiment.

In conclusion, the simulation and experiment results proved that spatial performance of monolithic detector can be improved by optimizing the photon detector arrangement. SAC is a good simple evaluation criterion for spatial performance, which need less computation amounts compared with CLRB.

Keywords: PET, Monolithic detector, Maximum Likelihood Estimation, Artificial Neural Network
Poster panel
(face) ID: 47

Poster Number:

Investigation of the Effects of SiPM Pixel Size and Noise on Edge Crystal Identification of Scintillator Arrays in BGO PET Detectors (#3677)

D. Prout1, A. Chatziioannou1

1 University of California, Los Angeles, Crump Institute, Los Angeles, United States of America


Improvements in SiPMs have decreased the noise in these devices and present the possibility of creating large arrays of BGO PET-detectors that, with multiplexing, could reduce the cost of such detectors. Also, it is now feasible with SiPMs to create arrays consisting of pixels of different surface areas. Smaller pixels (3mm2) can be placed on the edges of the photodetector matrix improving the resolution of the edge crystals for which light collection is challenging. Large pixels (6mm2) may then be used in the central portion of the detector, decreasing the number of SiPM pixels required for each detector head. This scheme along with multiplexing can greatly decrease the number of acquisition channels required. The degree of multiplexing is heavily dependent on the SiPM noise; BGO detectors are quite sensitive to noise as the crystals generate low amounts of light spread over a long time interval. Here, we use an array of SensL 3mm2 SiPMs, with individual channel readouts, coupled to a BGO crystal matrix to determine the level of multiplexing that can be attained while still resolving each crystal. We extend the concept of mixing different area pixels to analog SiPMs (this idea has been studied earlier using digital SiPMs). The pixel readouts were rebinned in software to compare three configurations (1) only 3x3mm2 pixels, (2) only 6x6mm2 pixels, (3) 3x3mm2 pixels on the edges and 6x6mm2 center pixels. Images were formed by a center-of-gravity algorithm and multiplexing was achieved by varying the number of pixels included in the algorithm. Additional virtual pixels were added, using a noise model, to expand the results beyond the pixels for which we had data. Preliminary results show an advantage of using small area pixels along the edges of the array with larger center pixels. However, noise present at the detector temperature of 28o C (realistic for PET tomographs), limits the multiplexing of pixels using such sized BGO crystals to at most 24-3x3mm2 SiPM pixels

Keywords: BGO, SiPM, PET, detector
Poster panel
(face) ID: 50

Poster Number:

Simulation Study of Partially Laser-Processed LYSO:Ce with Single-Side Readout (#3717)

L. Bläckberg1, 2, G. El Fakhri1, H. Sabet1

1 Massachusetts General Hospital & Harvard Medical School, Gordon Center for Medical Imaging, Department of Radiology, Charlestown, Massachusetts, United States of America
2 Uppsala University, Department of Physics and Astronomy, Uppsala, Sweden


We are developing detectors with single-side readout and depth of interaction (DOI) information using Laser Induced Optical Barriers (LIOB) by locally engineering the crystal’s refractive index (RI). The so-called optical barriers can redirect scintillation light and allow for control of the light spread in the detector. In this work we systematically study light transport in 20 mm thick LYSO:Ce with 4 simple pixel-like optical barrier patterns (with 1 mm barrier spacing) read out by a 3.2 mm pitch MPPC array. Our simulations highlight laser-processed detectors as a new detector category with a behavior between standard monolithic and mechanically pixelated detector types. The huge parameter space, in terms of barrier properties (like RI and barrier-crystal interface roughness) and patterns, allows to optimize the detector performance. We found that a detector with optical barriers all-the-way through the crystal thickness can yield high transversal resolution, similar to mechanically cut arrays, and depending on the barrier properties some DOI information can be extracted from the width of the light spread (from 2-3mm FWHM close to the MPPC to 4-8.5 mm FWHM in the top part, for rough interfaces). A pixel-like pattern only in the crystal half closest to the photodetector results in strong DOI dependence (1.2-2.0 mm FWHM decrease per mm depth regardless of barrier properties) in the top half of the crystal where most gamma-ray interactions take place, in combination with higher transversal resolving power compared to a monolithic detector. Results further suggest that up to 7% and 25% improvement in light collection efficiency can be expected compared to monolithic and pixelated detectors. This can lead to improved energy and timing resolution as well as positioning accuracy. Experimentally we show a 20 mm thick LYSO:Ce detector with 1 mm pitch pixel-like pattern, yields good pixel separation (2:1 peak-to-valley ratio) in central region using centroid positioning.  

Keywords: Laser Induced Optical Barriers, Light transport, Depth of Interaction, Positron Emission Tomography, LYSO:Ce scintillator
Poster panel
(face) ID: 53

Poster Number:

Design of ultra-thin anti-reflection films directly coated on LYSO scintillators (#4055)

Q. Sun1, Q. peng3, Z. Wu1, S. Xie1, Q. Huang2, J. Xu1

1 Huazhong University of Science and Technology, Wuhan, China
2 Shanghai Jiaotong University, ShangHai, China
3 Lawrence Berkeley National Laboratory, Berkeley, CA, United States of America


Optical dielectric coating is a technology commonly used to build anti-reflection (AR) films. It is constructed from thin layers of materials such as magnesium fluoride, calcium fluoride, and various metal oxides, which are deposited onto the optical substrate.

In this study, we have investigated the feasibility of applying AR coatings directly on a LYSO scintillator to improve its light output across the entire emission spectrum (370 nm to 600 nm). Four materials (MgF2, TiO2, SiO2 and HfO2) were selected to construct AR films since they have been proved to have excellent adhesive strengths on LYSO scintillators. The simulation results indicated that the single-layer MgF2 film, the 4-layer film constructed with MgF2, SiO2, HfO2, and the 9-layer film constructed MgF2, TiO2, HfO2 are all able to significantly reduce the reflection on the interface between LYSO and the air. The 7-layer film (MgF2, TiO2, HfO2) is the best design to reduce the reflection on the interface between LYSO and the glue.

We conclude that it is possible to significantly improve the light output from a LYSO scintillator to optic media by AR coatings. The AR coatings designed in this study can be applied to improve the performances of PET detectors, which are largely determined by the numbers of scintillating photons entering the photosensors. However, we have also noticed that all the AR coating films designed in this study failed to alter the critical angles on the interface between LYSO (n=1.8) and the optic media (air: n=1.0, glue: n=1.5). The textured coating technology might be a more attractive solution since it allows the photons with an incident angle larger than the critical angle to penetrate the interface between the LYSO and media.

Keywords: Coating, PET, Detector
Poster panel
(face) ID: 56

Poster Number:

Design and evaluation of a SiPM-based detector for PEM system (#1034)

Z. Lu1, 2, X. Huang1, 2, W. Zhou1, 2, J. Gao1, 2, Y. Wang1, 2, L. Li1, 2, Z. Zhang1, 2, L. We1, 2, C. Ma1, 2

1 Chinese Academy of Sciences (CAS), Institute of high energy physics, Beijing, China
2 Chinese Academy of Sciences (CAS), Beijing Engineering Research Center of Radiographic Techniques and Equipment, Beijing, China


Positron emission mammography (PEM) is an effective means for early diagnosis of breast cancer. In this work, we designed a flat-panel detector for PEM based on latest silicon photomultiplier tubes (SiPM). The detector was composed of 5×6 modules with a total area of 168.6×202.4mm2. Each module consisted of a 16×16 LYSO array with the pixel size of 1.9×1.9×15mm3. A 8×8 SiPM array constructed with SensL’s C30035 sensors was coupled to the LYSO with a 1.5mm thick glass. By using a split-type design, active part of the front-end electronics was kept away from SiPMs to avoid the influence of temperature. Self-designed data acquisition system was also used. Test results showed that the module had excellent crystal identification with an average peak-to-valley ratio of 11.8:1. The average energy resolution of 11.39% and coincidence time resolution of 1.32ns were also achieved. Preliminary results suggested that this SiPM-based detector can be well applied to our PEM system.

Keywords: Positron Emission Mammography, Silicon Photomultiplier, Flat-panel detector
Poster panel
(face) ID: 59

Poster Number:

Development and evaluation of a PET detector with 1.0 mm resolution and DOI measurement capability for integrated PET/CT/RT systems (#1298)

X. Cheng1, Y. Zhong1, Y. Shao1

1 UT Southwestern Medical Center, Radiation Oncology, Dallas, Texas, United States of America


The overall balanced imaging performance with compact and practical readout electronics and data acquisition are critically important in developing a PET for integrating with existing cone-beam CT image guided small animal irradiators for radiation therapy researches. In the design study, we optically coupled a latest 8x8 array of 2x2 mm2 MPPCs (SiPMs made by Hamamatsu Photonics K.K.) to each end of a 16x16 array of 1x1x20 mm3 LFS scintillator crystals. With this dual end scintillator readout, we aim for a 1.0 mm uniform imaging resolution based on ~4.0 mm or better depth-of-interaction resolution across a ~6cm diameter FOV. Besides the usual technical challenges to have optimized combination of scintillator surface treatment and inter-scintillator reflector to achieve targeted detector performance with minimal inter-crystal gaps, we applied 64-ch ASIC front-end electronics to read out and process each of 128 MPPC anodes individually in order to minimize the noise and improve the flexibility of processing multi-channel signals gathered from sharing of scintillation photons among different MPPCs. Preliminary test of an initial prototype detector show 16% energy resolution measured from a typical scintillator; Excellent ratio of signals for an individual gamma interaction measured from the dual MPPCs at both scintillator ends which indicates a DOI resolution 2.0-3.0 mm. The crystal identification measured from both flood radiations of a Na-22 isotope source and LFS internal background radiations show the potential capability to identify all crystals if appropriate sharing of scintillation light from one scintillator to multiple MPPCs will be applied, which is under more investigations with standard optimization techniques. The measured coincidence timing resolution is 1.1 ns. In summary, the prototype detector shows promising performance; more optimization studies are going; and more detailed results will be presented.

Keywords: PET, image guided small animal irradiator, MPPC, LFS, scintillator
Poster panel
(face) ID: 62

Poster Number:

Analysis of Tri-PET Background Counts (#1706)

F. P. DiFilippo1

1 Cleveland Clinic, Nuclear Medicine, Cleveland, Ohio, United States of America


Tri-PET is a hybrid modality having characteristics of both SPECT and PET, where a multi-pinhole collimator is used with detectors operating in coincidence mode rather than single-photon mode. When using standard clinical PET detectors, Tri-PET is capable of achieving ~1mm spatial resolution for small animal imaging. A unique feature of Tri-PET is its ability to reject most background counts by comparing the coincidence line of response to the multi-pinhole geometry. However, some background counts remain, and it is necessary to understand the source and magnitude of the background counts. Background counts were analyzed from computer simulations of a 148-pinhole Tri-PET collimator designed for high resolution imaging of small animals in a clinical PET scanner. By modifying parameters of the simulations (tungsten vs non-penetrating collimator, realistic vs ideal detector energy resolution, no scatter medium vs mouse-size phantom), the background counts were classified according to their sources. For this configuration, 74% of Tri-PET counts were classified as true counts, and 26% were classified as background counts. The sources of the background counts, namely penetration through the bulk collimator (2 cm thickness), incorrect pinhole assignment during Tri-PET data processing, and scatter (from the collimator or mouse phantom), were found to be 20%, 5%, and 1%, respectively, of total Tri-PET counts. Background counts are at an acceptable level and may be minimized further through additional modifications in collimator design, by increasing the collimator thickness and the distance between pinholes.

Keywords: Tri-PET, collimator, simulation
Poster panel
(face) ID: 65

Poster Number:

Proposal for a PET scanner with 4 π steradian span (#2182)

D. Perez1, R. Herrera1, R. Chil1, G. Konstantinou1, J. M. Udías3, M. Desco1, 2, J. J. Vaquero1, 2

1 Universidad Carlos III de Madrid, Bioingenieria e Ingenieria Aeroespacial, Leganés (Madrid), Spain
2 Instituto de Investigación Sanitaria Gregorio Marañón, Madrid, Spain
3 Universidad Complutense de Madrid, Department de Física atómica, Molecular y Nuclear, Madrid, Spain


The most frequent PET scanner design is a cylindrical arrangement of detectors with the patient lying along the axial direction. This distribution does not maximize the geometrical sensitivity unless the axial extent becomes much longer that the axial FOV, a very costly solution. The optimal geometry for a PET scanner is a sphere with the specimen located at the center. The state of the art of scintillators and detectors for gamma rays do not facilitate a design with a smooth spherical geometry. In this work we propose a PET scanner design of 4π steradian span as a first approximation to the ideal spherical shape. We present here a scaled design for preclinical imaging as a proof-of concept, shaped as an icosahedron with scintillators and photodetectors covering its twenty facets. This polyhedron is preferred over other platonic solids since it combines high area of directly facing, equidistant detectors surfaces, a unique facet shape, and a low angle subtended by facets. In this way, a great number of lines of response impinges the crystal with almost orthogonal-to-the-surface angles. In the simulations the facets are covered by SiPMs arranged in matrices for multiplexing reading, coupled to a monolithic LYSO scintillator crystals laser-engraved with pixels matching the spatial sampling of the SiPM distribution. Preliminary results are based on simulations of sources of known activity distributions; histograms of lines of response distributions in polar and cylindrical projections as well as sinograms and pseudo-michelograms were produced to assess analytical, algebraical and statistical-iterative 3D reconstructions performance. The results confirm the high-sensitivity and high-resolution of the design, and the possibility of using low tracer doses and/or employ thin scintillators to minimize depth of interaction effects. Considerations about electronics requirements like pile-up rejection, deadtime and bandwidth have also been evaluated.

Keywords: PET, Design optimization, Silicon Photomultipliers, PET scanner
Poster panel
(face) ID: 68

Poster Number:

A Novel Laparoscope for Simultaneous NIR/Gamma/Visible Imaging during Laparoscopic Surgery (#2241)

H. G. Kang1, S. H. Song1, Y. B. Han1, H. - Y. Lee2, K. M. Kim3, S. J. Hong1, 4

1 Eulji University, Department of Senior Healthcare, Daejeon, Republic of Korea
2 Seoul National University, Department of Nuclear Medicine/College of Medicine, Seoul, Republic of Korea
3 Korea Institute of Radiological and Medical Science, Division of Medical Radiation Equipment, Seoul, Republic of Korea
4 Eulji University, Department of Radiological Science, Seongnam-Si, Republic of Korea


We present a novel laparoscope for simultaneous NIR/gamma/visible imaging during laparoscopic surgery. The proposed laparoscope system consists of a custom-made laparoscope, a beam splitter module, and an illumination system. The laparoscope head consists of a tungsten pinhole collimator, a focusing lens, a long pass filter (500 nm~), a GSO crystal (9 × 11 × 2 mm3), and an optical fiber bundle. The laparoscope head was specially designed to form an NIR, gamma, and visible image on the same image plane of the entrance surface of the optical fiber bundle. Subsequently, the three images were sent to the custom-made beam splitter and detected on the three CCD cameras simultaneously. The illumination system can deliver both white light and NIR excitation light into the imaging object simultaneously. The performance of visible and NIR images was evaluated using a USAF 1951 resolution target. An eppendorf tube (EP tube) containing a mixture of 99mTc (2.87 MBq) and ICG (40 μL) was used to obtain simultaneous NIR, gamma and visible images. The NIR and visible images were obtained for 50 ms, while the gamma image was obtained for 30 sec. The NIR, gamma, and visible images of the EP tube showed a good spatial correlation. In future, preclinical trials will be conducted to evaluate the clinical usefulness of the proposed multimodal laparoscopic NIR/gamma/visible imaging system.

Keywords: Laparoscopic surgery, Intraoperative imaging, Near-Infrared (NIR), Radioguided surgery, Indocyianine green (ICG), 99mTc, Sentinel Lymph Node (SLN)
Poster panel
(face) ID: 71

Poster Number:

drimPET: Determination of DoI in LYSO crystals using SiPMs and wavelength shifters (#2462)

P. M. M. Correia1, I. F. C. Castro2, N. Romanyshyn1, J. F. C. A. Veloso1

1 University of Aveiro, i3n and Physics deparment, Aveiro, Aveiro, Portugal
2 RI-TE Radiation Imaging Technologies, Lda, Aveiro, Aveiro, Portugal


The determination of the depth-of-interaction (DoI) of 511 keV gamma photons in scintillator crystals is of great importance in small diameter positron emission tomography (PET) systems, to achieve high spatial resolution with good uniformity within the entire field-of-view. drimPET is new concept for DoI determination wherein a single layer ring of 128 LYSO crystals is read out on both ends: on the outer end, each crystal is directly coupled to a SiPM, while on the inner end groups of several (e.g., 16 or 32) crystals are indirectly coupled to only 1 or 2 SiPMs, using intermediate wavelength-shifters (WLS). This results in a simpler and less expensive readout when compared to the typical dual-ended readout method. Also, less components on the inner side of the ring allow more proximity to the FoV and thus increased sensitivity. Proof-of-concept studies were performed on a single detector cell with electronic collimation. The amount of light detected on each side of the crystal and the achievable DOI resolution with drimPET were measured, as well as the energy resolution, for different positions of a 22Na source along the crystal depth. The single cell was composed of a LYSO crystal, initially with both ends read by SiPMs and afterwards applying the drimPET method. Experimental studies and GATE simulations of drimPET are presented to evaluate the feasibility of applying the method in full detector rings for a small animal PET system, considering different number and dimension of the wavelength-shifting fibers.

Keywords: DOI, wavelength shifters, PET, pre-clinical, SiPM
Poster panel
(face) ID: 74

Poster Number:

Construction of a Novel Endoscopic Cerenkov Luminescence Imaging System for Image-Guided Resection of Hepatocellular Carcinoma on Mice Models (#2601)

Z. Zhang1, H. Guo1, Z. Hu1, J. Tian1

1 Chinese Academy of Sciences (CAS), Institute of Automation, Key Laboratory of Molecular Imaging, Beijing, Beijing, China


Nowadays, Cerenkov luminescence imaging (CLI) becomes a novel cancer molecular imaging method. Nevertheless, most of the existing imaging devices can not readily acquire the CL images for the CLI signals own weak intensity and poor penetration ability. Endoscopic CLI (ECLI) emerged recently as a considerable solution for CLI application. In this study, we construct a novel ECLI system to detect tumors and guide the resection, which mainly consisted of an electron-multiplying charge-coupled device (EMCCD) and a clinical laparoscope. Air-cooling and water-cooling were novelly integrated in the ECLI system (Figure 1), to reduce the working temperature to -95℃ for controlling thermal noise and improving imaging sensitivity. An IVIS system was employed to contrast with the developed ECLI system. Subcutaneous (n=4) and orthotropic (n=4) hepatocellular carcinoma (HCC) mice models were used in this study. CLI and ECLI were acquired respectively to detect the tumors, after vein injection of 18F-FDG (~300 μCi per mouse). According to the detection results, we conducted ECLI guided tumor resection on orthotropic HCC mice models (n=4) and validated the resection performance by pathological verification. The experimental results including: 1) robust linear correlation (R2 = 0.931) of signal counts between CLI and ECLI results was observed, which demonstrated the ECLI system can accurately detect tumors; 2) by applying the novel cooling solution, the developed ECLI system showed higher sensitivity in small tumor detection comparing with the IVIS system; 3) the ECLI system is capable to reveal explicit resection sites during tumor removal operation. As a conclusion, the successful detection and guided-resection of tumor on HCC mice models indicated that the developed ECLI system has a great potential for cancer diagnosis and treatment.

Keywords: Cerenkov luminescence imaging (CLI), Imaging System
Poster panel
(face) ID: 77

Poster Number:

Clear sub-millimeter spatial resolution in small animal PET using high density avalanche chamber technology (#2809)

K. Bolwin1, F. Büther1, 2, K. P. Schäfers1

1 University of Münster, European Institute for Molecular Imaging (EIMI), Münster, North Rhine-Westphalia, Germany
2 University Hospital of Münster, Department of Nuclear Medicine, Münster, North Rhine-Westphalia, Germany


Positron Emission Tomography (PET) allows for the in vivo assessment of molecular processes in biomedical research. The application of PET in small animals, such as mice or rats, is challenging due to small regions of interest.

The spatial resolution of small animal PET systems is limited by several physical factors including positron range, object scattering and non-collinearity effects. The influence of the non-collinearity effect can be reduced by keeping the distance between object and detector as small as possible. To address this, we built a high-resolution prototype PET scanner by minimizing the detector distance of an existing high-resolution small animal PET device using multi-wire proportional chambers (MWPCs). We performed point source and phantom measurements on a prototype device to study the influence of the non-collinearity effect on the scanner's spatial resolution. The spatial resolution could be improved from 1.13 mm to 0.92 in directions perpendicular to the wire direction (x,y), and from 1.09 to 0.68 mm along the wires (z). Furthermore, we developed a module inlet prototype to reduce the wire gap from 1.5 mm to 1 mm which should further improve the spatial resolution in x and y direction.

Keywords: small animal PET, multi-wire proportional chamber, high resolution
Poster panel
(face) ID: 80

Poster Number:

Towards the first high rate PET-MR-Insert (#3072)

R. Becker1, A. Buck2, C. Casella1, V. Commichau1, S. K. Dhawan6, D. Di Calafiori1, G. Dissertori1, A. Eleftheriou3, J. Fischer1, A. S. Howard1, M. Ito1, P. Khateri1, W. Lustermann1, J. F. Oliver4, U. Röser1, C. Ritzer1, P. Solevi5, G. Warnock3, B. Weber3, M. Wyss3, A. Zagozdzinska-Bochenek1

1 ETH Zürich, Institute for Particle Physics, Zürich, Switzerland
2 University of Zürich, Clinic of Nuclear Medicine, Zürich, Switzerland
3 University of Zürich, Institute for Pharmacology and Toxicology, Zürich, Switzerland
4 Universitat de València, Instituto de Física Corpuscular, València, Spain
5 Otto-von-Guericke University Magdeburg, Institute of Medical Technology, Magdeburg, Germany
6 Yale university, Department of Physics, New Haven, United States of America


Within the SAFIR collaboration we are building a high rate positron emission tomography (PET) insert for a preclinical magnetic resonance tomograph (MRT). In this work we present the results that we obtained during our component tests.

In our insert, we will use 2.1 × 2.1 × 12 mm3 LYSO crystals, in an 8x8 matrix coupled to a similar array of 2 × 2 mm2 SiPMs (HAMAMATSU S13361-2050AE-08 SPL). To evaluate the performance, we used a test setup with two matrices and one readout ASIC (PETA6SE). When cutting on the photopeak (411 keV − 611 keV) we measured a timing resolution of 270 ps (FWHM) and an energy resolution of 15.1 % (FWHM of photopeak) after linearisation and at the nominal overvoltage (3 V). By increasing the overvoltage to 6 V, those values change to 224 ps and 21.5 %.

Furthermore we have developed a digital readout electronic, which handles the digital data from the front-end ASICs and transfers those to the readout server sing a commercial SFP optical Ethernet module and optical fibres. We have tested the optical data connection and we could operate it, without disturbing the MR imaging.

Finally we have successfully tested an MR-compatible step-down converter, which can deliver up to 4 A with an efficiency of 80.5 % (converting 18 V to 3.3 V at 3 A). By cooling the device to room temperature we could increase the efficiency to 85 %. We did not observe any distortions of the MR imaging during those tests.

Keywords: PET-MR, Small animal imaging
Poster panel
(face) ID: 83

Poster Number:

Imaging Performance of a Submillimetric Spatial Resolution APD-Based Preclinical PET Scanner Dedicated to Mouse Imaging (#3356)

E. Gaudin1, C. Thibaudeau1, L. Arpin2, J. - D. Leroux3, D. Dufourt-Forget1, J. - F. Beaudoin1, J. Cadorette1, M. Paille1, C. M. Pepin1, K. C. Koua2, A. Samson2, J. Bouchard2, C. Paulin2, R. Fontaine2, R. Lecomte1

1 Université de Sherbrooke, Sherbrooke Molecular Imaging Center of CRCHUS and Department of Nuclear Medicine and Radiobiology, Sherbrooke, Québec, Canada
2 Université de Sherbrooke, Institut Interdisciplinaire d’Innovation Technologique (3IT) and Department of Electrical and Computer Engineering, Sherbrooke, Canada
3 Novalgo Inc., Sherbrooke, Canada


The LabPET II is an APD based PET scanner platform designed to achieve sub-mm spatial resolution using fully pixelated detectors and highly integrated parallel front-end processing electronics. This work provides an evaluation of the physical imaging performance of the mouse-version scanner according to the NEMA NU4-2008 protocol and reports initial in vivo imaging studies with mice.

An effective energy window of 400 to 650 keV and an initial coincidence time window of 20 ns were used for all acquisitions. Images were reconstructed using a 3D MLEM algorithm implementing the physical description of the system matrix with 20 iterations. Spatial resolution, absolute sensitivity, recovery coefficients, image uniformity and spill-over ratios were measured following the NEMA methodology. Whole-body images of mice were obtained to assess the image quality of the scanner for bone (21.4 g mouse injected with 8.9 MBq of Na18F) and cardiac (12.6 g mouse injected with 8 MBq of [18F]-FDG) imaging.

Submillimeter tangential FWHM spatial resolution is achieved across the entire field-of-view, while radial FWHM resolution remains below 1 mm up to 15 mm off-centre in reconstructed images. An absolute sensitivity of 3.31% is reached at the center of the FOV. Air and water spill-over ratios of 32% and 20% were measured using the NEMA image quality phantom, and the standard deviation of the uniform region was 8.2%.

The whole-body bone scan of the mouse shows submillimetric details such as sternocostal joints, vertebra and small cranial bone structures can be clearly differentiated from adjacent structures. The left ventricle of the tiny 12.6 g mouse can also be clearly resolved.

Sub-mm spatial resolution was achieved in phantom and mouse images. The initial scanner performance characteristics and image quality provide convincing evidence of the promising capabilities of the LabPET II technology for biomolecular imaging.

Keywords: Perclinical PET, APD-based, submillimetric
Poster panel
(face) ID: 86

Poster Number:

Performance evaluation of HiPET, a High Sensitivity and High Spatial Resolution DOI PET Tomograph (#3593)

Z. Gu1, R. Taschereau1, D. L. Prout1, N. Vu2, A. Chatziioannou1

1 University of California, Los Angeles, Crump Institute for Molecular Imaging, Los Angeles, California, United States of America
2 Sofie Biosciences, Culver City, California, United States of America


A new preclinical PET tomograph named HiPET is currently being developed at the Crump Institute for Molecular Imaging, at UCLA. This scanner aims to achieve high peak absolute sensitivity (up to 19.1%) and submillimeter spatial resolution.

To achieve this, HiPET consists of ten panel detectors, with a crystal ring diameter of 160 mm and an axial extent of 104 mm. Each panel detector consists of two modules, and each module employs a phoswich depth of interaction (DOI) design composed by two layers of scintillator arrays (LYSO and BGO). The top (gamma ray entrance) layer is a 48 x 48 array of 1.01 x 1.01 x 6.1 mm3 LYSO crystals (1.09 mm pitch). The bottom (towards the PMT) layer is a 32 x 32 array of 1.55 x 1.55 x 8.9 mm3 BGO crystals (1.63 mm pitch). The analog detector signals are digitized by sixty 125 MHz free running ADC and the digital samples are processed in a Xilinx Virtex-6 FPGA.

The average crystal energy resolution is 12.3% measured from 46080 LYSO crystals, and 17.3% measured from 20480 BGO crystals of the HiPET system. For LYSO-LYSO crystal pairs, the measured intrinsic detector spatial resolution was 0.89±0.09 mm FWHM. For BGO-BGO crystal pairs, the measured FWHM was 1.35±0.07 mm. Preliminary imaging studies including the NEMA NU-4 image quality phantom and a mouse bone scan were performed, demonstrating that the HiPET scanner is suitable for producing high quality images for molecular imaging based biomedical research.

Keywords: PET, performance evaluation, DOI, phoswich, preclinical, LYSO, BGO
Poster panel
(face) ID: 92

Poster Number:

Using Higher Energy Radioisotopes for Crystal Identification in a SPECT System (#4064)

D. Stentz1, P. Sankar1, W. Chang1, J. S. Karp1, S. D. Metzler1

1 University of Pennsylvania, Department of Radiology, Philadelphia, Pennsylvania, United States of America


A cost-effective solution for high-resolution clinical single-photon emission computed tomography (SPECT) is a set of modular pixelated scintillation crystal detector readout with an array of medium-sized, conventional photo-multiplier tubes (PMTs).  This is the approach of C-SPECT, a dedicated cardiac SPECT and X-ray transmission computed tomography (tCT) system, which uses NaI(Tl) scintillation crystals and 2 inch PMTs arranged in a hexagonal pattern.  While not a unique difficulty, the process of identification of each crystal pixel using a standard Technetium-99m flood (no collimator) is exacerbated by the small pixel size and the performance of a small number of PMTs used to determine positioning for any given event.  We investigate the improved performance of using a higher energy radioisotope (i.e. Sodium-22 or Germainum-68) for our crystal identification flood.  Improved pixel sharpness with the higher energy radiation source (511 keV photons) is measured and compared to a flood at 140.5 keV.  This improvement results in a better discrimination between pixel boundaries that is crucial near the PMT positions and the edge of the detector.  Finally, we confirm the reliability of using this procedure for pixel identification for our nominal use of Technetium-99m and estimate the fraction of events that are misassigned to neighboring pixels.

Keywords: SPECT, Pixel Assignment
Poster panel
(face) ID: 95

Poster Number:

ZTE-based attenuation correction in head and neck PET/MR (#1430)

M. Khalifé1, R. de Laroche2, D. Bequé3, B. Sgard2, F. Pérez-García1, M. Soret2, M. - O. Habert2, 4, F. Wiesinger3, A. Kas2, 4

1 Institut du Cerveau et de la Moelle Epinière (ICM), CNRS UMR 7225 - Inserm U1127 - Université Paris 6 UPMC UMR S1127, Paris, France
2 Service de médecine nucléaire, Groupe Hospitalier Pitié-Salpêtrière C. Foix, Paris, France
3 GE Global Research Center, Munich, Germany
4 Université Paris 6 UPMC, LIB Inserm U1146, Paris, France


PET/MR in head and neck cancer is still lacking an accurate attenuation correction (AC). Dixon-based AC is used in the clinical routine ignoring facial and vertebral bones. ZTE MRI, previously used to segment skull in brain applications, is used here to detect bone in MRI in 7 patients. To account for fat in the neck, a combined ZTE and Dixon-based AC map is considered as fat and water are separated on the Dixon fat and water MRI. CT acquired on the same patients and registered to ZTE image formed an AC gold-standard. PET images reconstructed with Dixon, ZTE and ZTE-Dixon AC maps are compared to CTAC PET image. SUVmean and SUVmax were compared in volumes of interest (VOI) drawn on physiological uptake in the PET images. Results showed that over all the VOIs ZTE-Dixon AC kept the mean error within a 10% margin. SUVmean and SUVmax in VOIs close to bone (maxillary and mandible) were underestimated when using Dixon AC and overestimated when using ZTE AC whereas ZTE-Dixon AC showed the lowest error compared to CTAC. ZTE-Dixon AC performed better than ZTE AC in near-bone regions. However, ZTE-based AC methods still showed an error higher than Dixon AC in nasal cavities due to partial volume effect. Consequently, the hybrid ZTE and Dixon AC map showed promising results for head and neck PET/MR especially in regions surrounding bones. Future work will explore learning-based methods to improve nasal cavities segmentation using ZTE MRI.

Keywords: PET/MR, Attenuation correction, head and neck, ZTE MRI
Poster panel
(face) ID: 98

Poster Number:

Novel measurement of MTF and axial resolution in a TOF PET scanner with 3.2-mm crystals (#1684)

J. Hamill1, V. Y. Panin1, H. Bal1, D. K. Bharkhada1

1 Siemens Medical Solutions, Molecular Imaging, Knoxville, Tennessee, United States of America


Background: Modern PET scanners provide high spatial resolution based on small crystals and appropriate use of time-of-flight information (TOF), for example TOF-based rebinning. A novel phantom is introduced to measure the axial-direction modulation transfer function (MTF), and to demonstrate that comparable resolution is realized at the scanner’s center and off center.

Methods: A 3D printer was used to create an 18F-filled Defrise-type phantom with hot and cold discs, 6.35 cm in diameter separated by 3, 4 and 5 mm. This was placed approximately parallel to the long axis of a Siemens next generation SiPM PET/CT prototype scanner with 3.2 mm LSO crystals. The phantom was scanned near the central axis and 20 cm off center. List-mode PET data were reconstructed iteratively with TOF rebinning (40 image updates) and with normal span-11 reconstruction (38 updates). Attenuation and scatter corrections were based on CT scans. A profile provided an MTF measurement at each of the three spatial frequencies. This was related to spatial resolution by referring to the MTF of a numerical version of the phantom after Gaussian blurring.

Results: In the central position, MTF with TOF rebinning had values 0.708, 0.532 and 0.306 at 0.1, 0.125 and 0.167 lp/mm respectively, consistent with FWHM resolution less than 4 mm. MTF values 20 cm off center had similar values: 0.706, 0.533 and 0.303. MTF values in the span-11 reconstruction were lower: 0.535, 0.426 and 0.265 at the center, and 0.498, 0.296 and 0.112 off center, indicating degraded resolution in that position.

Conclusions: The novel MTF-based method provided redundant measurements in one scan, i.e. values at different spatial frequencies that agreed, approximately, with a simple interpretation based on Gaussian blur. With TOF rebinning, MTF values were consistent in the center and 20 cm off center.

Keywords: PET, MTF, Defrise phantom, TOF
Poster panel
(face) ID: 101

Poster Number:

MR compatibility evaluation of the HPK C13500 series PET module for organ-specific RF/PET-detectors (#1777)

F. P. Schmidt1, C. Parl1, B. J. Pichler1

1 Eberhard Karls University, Werner Siemens Imaging Center, Department of Preclinical Imaging and Radiopharmacy, 72076, Germany


The goal of our research project is a next level of multimodality imaging, focusing on organ-specific positron emission tomography and magnetic resonance imaging (PET/MRI). Specifically, the future aim is to integrate a dedicated breast PET-radio frequency (RF)-coil setup in a human PET/MRI. Therefore, we investigated new application specific integrated circuits (ASICs) as PET frontend electronics and evaluated the overall performance and MR compatibility.

A promising candidate is a modified version of the C13500-40LC-12 PET module (Hamamatsu Photonics, Hamamatsu, Japan) with its backend electronics that was evaluated in terms of simultaneous operation with a clinical 3 T MRI scanner. The B0 field homogeneity and a RF noise spectrum in presence of the PET modules were measured and the influence of the B0 field and RF pulses on the energy and timing resolution of the PET components were determined.

The proper operation of the PET electronics, i.e. measurements of the 511 keV energy resolution of 8.2±0.8 % FWHM and coincidence resolving time of 328 ps FWHM, were maintained even when RF-pulses of a MRI sequence were applied. This has been evaluated at a distance of 1.05 m between the opening of the MRI bore and the PET module, decreasing the distance resulted in a denial of operation of the PET backend electronics. The B0 homogeneity was distorted by the light sensor and the scintillator block by more than 0.5 ppm.

Although the performance of the PET module looks promising, the MR compatibility for a fully integrated system requires the separation of the frontend and backend electronics. However, modifications are in progress, which allow locating the backend electronics outside the MRI cabin and thus, enabling to perform the compatibility tests at several positions inside the MRI bore. Additionally, the PET modules will be shielded to minimize the RF interference and the material of the scintillator block housing will be changed to reduce the B0 inhomogeneity.

Keywords: PET, MRI, ASIC
Poster panel
(face) ID: 104

Poster Number:

IRIS Projector Adaptation for PET Scanners Based on Monolithic Crystals (#2038)

A. Iborra1, J. Bert1, T. Merlin1, D. Benoit1, S. Sánchez2, P. Conde2, A. J. González2, M. J. Rodríguez-Álvarez2, D. Visvikis1

1 Laboratory of Medical Information Processing (LaTIM, INSERM UMR1101), CHRU Brest, Brest, France
2 Institute for Instrumentation in Molecular Imaging (i3M), CSIC, Universitat Politècnica de València, Valencia, Spain


The Intrinsic Detector Response Function (IDRF), which varies within the imaging field of view, describes the uncertainty in the line of response (LOR) vertex location. The Iterative Random IDRF Sampling (IRIS) projector uses the accumulation of multiple rays to estimate the response associated with a given line of response. Each of these rays connects two points randomly generated using a model of the IDRF. A model for this response function is not straightforward to obtain in the case of a PET system based on monolithic crystal detectors, due to several factors such as depth of interaction dependence, or compression effects introduced by point of interaction determination algorithms. In this work we present a method, based on a hash table, to numerically approximate the uncertainty of a line of response vertex without the need of explicitly model the Response Function, thus, allowing to incorporate the IRIS projector to image reconstruction algorithms for PET systems based on monolithic crystal detectors. A Derenzo-like phantom has been measured with the dedicated brain PET insert MindView, that uses monolithic crystal detectors. Considering the uncertainty regions computed with the proposed method, the Iterative Random Sampling projector successfully increased SNR values of the reconstructed image, especially regarding the events detected closer to the detector surface (at a higher depth of interaction). Being independent of the reconstruction process, the proposed method in combination with the Iterative Random Sampling projector can improve the performance of any image reconstruction algorithm for PET systems based on monolithic crystal detectors.

Keywords: PET, Image reconstruction, IDRF, IRIS
Poster panel
(face) ID: 107

Poster Number:

Machine Learning Based Approaches for SWEDD diagnosis in DaTSCAN SPECT imaging (#2193)

R. Mabrouk1, B. Chikhaoui2, L. Bentabet3

1 University of British Columbia, Brain Research Center, Vancouver, British Columbia, Canada
2 Research institute in Montreal, Montreal, Québec, Canada
3 Bishop's University, Computer Sciences, Sherbrooke, Québec, Canada


DaTSCAN (123I iophlipane injection) is a Single-photon emission computed tomography molecular imaging agent of Dopamine Transporter (DaT). DaTSCAN is widely used as a hallmark of nigrostriatal nerve cells degeneration such as in Parkinson’s Disease (PD). DaT density is commonly assessed by calculating the Striatum Binding Ratio (SBR). For some subjects showing abnormal motor signs, SBR falls in the normal range which is known as Scan Without Evidence of Dopaminergic Deficit (SWEED). The fate of these subjects to develop PD is not obvious. In this paper, we developed Machine Learning based models for PD classification and SWEDD condition prediction using data from the Parkinson’s Progression Markers Initiative (PPMI).  We availed of the Multivariate Logistic Regression (MLR) and Principal Component Analysis-Linear Support Vector Machine (PCA-LSVM) to establish competent models. SBR calculated from the least affected- and most affected-side of Putamen and Caudate regions and the Unified Parkinson's Disease Rating Scale (UPDRS) were used as significant contributors to our models.  The MLR provided a correct classification between PD and controls at an area under the curve (AUC) of 0.98. The R² from Cox and Snell test was significantly high (0.65) which indicate a high significance of the model. At an optimal threshold of 98%, 8 subjects from SWEED were detected as participants who would develop PD. The PCA-LSVM model provided a correct classification between PD and controls at an average accuracy of 0.92. The LSVM predicted 5 SWEED subjects as participants who would develop PD. Four subjects were cross-identified from the MLR and PCA-LSVM. Developed models provided high capacities to distinguish PD participants from controls. The MLR at its optimum threshold coequal the PCA-LSVM model in the prognostic of SWEED group.

Keywords: SPECT, Machine learning, PD, SWEED
Poster panel
(face) ID: 110

Poster Number:

PET detector block with DOI capabilities based on a large monolithic BGO crystal (#2821)

A. González-Montoro1, S. Majewski2, S. Zanettini3, F. Sánchez1, A. Aguilar1, J. M. Benlloch1, A. J. González1

1 Institute for Instrumentation in Molecular Imaging, i3M, Valencia, Spain
2 University of Virginia, Charlottesville, Virginia, United States of America
3 NAPA Technologies, Troyes, France


In PET scanners both scintillation crystals and photosensors are key components defining the system’s performance and cost. Original PET systems used BGO or NaI(Tl) scintillators but achieved limited performance due to its slow decay and relatively low light output. In this work we show the possibility to use again BGO crystals, and in particular monolithic blocks, in low-dose large-size scanners, with significantly improved sensitivity and reduced cost. In addition, while not the focus of this work, recent experiments showed CRT as low as 560 ps with BGO crystals.

In this work we determine performance of a monolithic BGO block as large as 50×50×15 mm3 with black-painted lateral walls with depth of interaction (DOI) capabilities. A directional optical barrier, called retro-reflectors, is coupled to the entrance face bouncing back the scintillation light in the direction of the emission source and, therefore, adding signal while preserving the narrow light distribution. Four configurations namely 12×12 or 16×16 SiPM (3×3 mm2) as photosensors, with or without nanopattern crystal treatment have been studied. The nanopattern structure consisted in a thin layer of specific high refractive index material shaped with a periodic nanopattern, increasing light extraction. The readout returned information for each SiPM row and column, characterizing the X-Y light projections. We have studied the detector spatial resolution using collimated 22Na sources at normal incidence. The DOI resolution was evaluated using beams with lateral incidence.

An average FWHM for the DOI resolution of 5.3±0.3 mm has allowed us to separate the crystal thickness into 3 DOI layers. Here, we have studied the detector spatial resolution as a function of the DOI layer obtaining 2.5±0.4 mm at the entrance layer (15-10 mm), 2.2±0.4 mm (10-5 mm) and 1.6±0.4 mm (5-0 mm, near the photosensor). An average energy resolution as good as 20% FWHM was obtained for the whole crystal volume.

Keywords: BGO crystal, SiPM, Total body PET
Poster panel
(face) ID: 113

Poster Number:

Improving PET sensitivity and resolution by photon interaction sequence timing discrimination (#2911)

V. Ilisie1, V. Giménez-Alventosa1, L. Moliner1, A. Aguilar1, E. Lamprou1, A. J. González1, F. Sánchez1, J. M. Benlloch1

1 Universitat Politecnica de Valencia, Instituto de Instrumentación para Imagen Molecular (I3M), Centro Mixto CSIC, Valencia, Spain


Detectors for current Positron Emitting Tomography scanners are based on the detection of the photoelectric energy peak while information from Compton scattering is being rejected. If an event involves multiple Compton scattering and the total energy deposited lays within the photoelectric peak, then a energy-weighted average is given for the interaction point. This technique introduces blurring in the image reconstruction process. We show that one can drastically improve the quality of the final reconstructed image, by precisely knowing the coordinates of the first interaction point. Compton events are rich source of information and one can also use them in order to discard fake events such as randoms, singles or even patient scattered photons, by means of reconstructing the Compton cones and require them to be compatible (Compton cones for photons that come from the same event must intersect in a line of response). Details of this Compton algorithm are also presented.

Keywords: PET, LaBr3, LYSO, TOF, Compton
Poster panel
(face) ID: 116

Poster Number:

A practical method to perform Depth of Interaction calibration in a single side readout PET detector (#3238)

M. Pizzichemi1, G. Stringhini1, 2, A. Ghezzi2, A. Stojkovic3, M. Paganoni2, E. Auffray1

1 CERN, Geneve, Genève, Switzerland
2 University of Milano-Bicocca, Department of Physics, Milano, Italy
3 Wellesley College, Wellesley, Massachusetts, United States of America


Knowledge of the Depth of Interaction (DOI) of gamma rays in PET detectors is mandatory to reach high levels of spatial resolution and image homogeneity, particularly in small animal and organ dedicated scanners. Reconstructing the real impact position of incident gammas requires precise knowledge of the relation between some observable measured by the scanner and the DOI coordinate. These calibration functions need to be measured frequently, with high level of accuracy, in order to ensure a reliable output of the scanner over time.

In our collaboration, we recently developed a novel method for extracting continuous DOI information in a single side readout detector, based on a simple light sharing and recirculation scheme. The measurement of the DOI calibration functions was anyway performed using an external electronic tagging setup, making this procedure unsuitable for a fully assembled PET scanner, given the physical constrains and the long time required to complete a full characterization. In this work, we develop and validate experimentally a practical method to perform an accurate calibration of the DOI functions that requires just a fast exposition to an external gamma source that can be easily placed in the center of field of view.

Keywords: PET, Depth of interaction, Scintillator detectors
Poster panel
(face) ID: 119

Poster Number:

Anthropomorphic Thorax Phantom Wilhelm for Cardiac and Respiratory Motion Simulation in Hybrid Emission Tomography (#3519)

K. Bolwin1, B. Czekalla1, L. J. Frohwein1, F. Büther2, K. P. Schäfers1

1 University of Münster, European Institute for Molecular Imaging, Münster, North Rhine-Westphalia, Germany
2 University Hospital of Münster, Department of Nuclear Medicine, Münster, North Rhine-Westphalia, Germany


Respiratory and cardiac motion as well as body motion during medical imaging using techniques such as computed tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET), or single emission computed tomography (SPECT) is well known to disturb image accuracy and quality. Motion correction methods are under development to overcome this issue. These correction techniques need to be validated and results need to be compared to previous results or ground truth data which is the domain of hardware motion phantoms. In this study, an anthropomorphic thorax phantom is presented which is capable to simulate the entire process of intrinsic patient motion caused by cardiac contraction and respiratory action. The measurement of small spherical wax "plaque-type" lesions attached to the myocardial wall was used to compare the detectability of these lesions for different reconstruction algorithms with one another. Using the optimal gate for the reconstruction yields an improvement of contrast and lesion detectability. Although the intensity is decreased compared to the static reference. We showed one example of testing motion correction on a simplified left ventricle. To mimic more realistic heart dynamics we are currently creating a new heart design including the right ventricle and coronary vessels.

Keywords: Cardiac and respiratory motion, Hybrid Emission Tomography, Thorax Phantom
Poster panel
(face) ID: 122

Poster Number:

Interpolation based penalty for anisotropic voxels in penalized likelihood PET image reconstruction (#4018)

N. Jain1, S. G. Ross2, S. Ahn3

1 Wipro GE Healthcare, Molecular Imaging, Bangalore, Karnataka, India
2 GE Healthcare, Molecular Imaging, Waukesha, Wisconsin, United States of America
3 GE Global Research, Molecular Imaging, Niskayuna, New York, United States of America


Ordered subset expectation maximization (OSEM) is the most commonly used algorithm in clinical PET image reconstruction.  A penalized likelihood (PL) image reconstruction method has recently been introduced for clinical PET imaging utilizing the relative difference penalty and has demonstrated improvements in quantitation and SNR over OSEM.  The PL method uses a penalty function which utilizes information from neighboring voxels to control image noise.  For each image voxel, the penalty function value is determined using information from the neighboring 26 voxels in 3D imaging.   In typical PET scans, the axial voxel size is fixed while the transaxial voxel size is dependent on the field of view (FOV) and image matrix size.  In most whole-body imaging scenarios, the axial and transaxial voxel sizes are similar and an assumption of equal size is reasonable.  For targeted reconstructions at high matrix size, such as those used for brain imaging, this can result in highly anisotropic voxels where the axial voxel size is much larger than the transaxial voxel size.  In these targeted cases, an assumption of equal voxel size in the penalty function can result in an unequal amount of penalty applied to the axial and transaxial direction.  This can in turn result in increased smoothing in certain views (typically the coronal and sagittal for brain imaging).   This work describes an interpolation based penalty function which can be used to account for the anisotropic voxel size in targeted reconstructions.  The new method utilizes interpolation in one dimension during the penalty calculation and does not pose a significant run-time expense.  Results are shown which demonstrate image quality improvements in phantom and clinical images with the new penalty.

Keywords: Penalized Likelihood, image reconstruction
Poster panel
(face) ID: 125

Poster Number:

Optimization of the coupling of the LYSO scintillator to a photo-sensor (#4090)

R. Pestotnik1, A. Majdič3, S. Korpar2, P. Križan3, 1

1 Jožef Stefan Institute, Experimental Physics department, Ljubljana, Slovenia
2 University of Maribor and JSI, Maribor, Slovenia
3 University of Ljubljana, Department of Physics, Ljubljana, Slovenia


Positron emission tomography is an important medical diagnostic modality, based on the detection of a pair of  annihilation gamma rays, usually detected by a segmented gamma scintillation detector. In the  detector the light emitted in the segmented scintillation crystals is read out by the photo sensor. To maximize the coverage, the transverse size of the scintillation crystal does not necessarily match the photosensitive area of the photo sensor leading to a poor reconstruction of the edge and corner crystal slabs. To optimize the light collection, we have studied several different types of the light guides. The light guides were simulated by using the ANTS2 framework. Flood images based on the determination of center of gravity of the signal were analyzed. A resolvability index was calculated for each of the scintillator slabs. Based on the results, several different designs have been tested and the results compared to the simulation. 

Keywords: light guide, positron emission tomography, gamma ray detector
Poster panel
(face) ID: 128

Poster Number:

Optical imaging of alpha emitter:Ra-223 solution for alpha targeting radionuclide therapy (#1968)

S. Yamamoto1, K. Kato1, H. Kameyama2, S. Abe2

1 Nagoya University Graduate School of Medicine, Nagoya, Aichi, Japan
2 Nagoya University Hospital, Nagoya, Aichi, Japan


Ra-223 is recently introduced for alpha targeting radionuclide therapy. According to the decay scheme of Ra-223, beta particles higher energy than the Cerenkov-light threshold are emitted. In addition, there are reports that alpha particles themselves emit luminescence in water and air. However measured optical imaging results from Ra-223 is not yet reported. Thus we conducted the optical imaging of Ra-223 solution and compared with that of F-18 solution. Ra-223 and F-18 solutions contained in transparent glass vials were imaged using a high sensitivity CCD camera. We could obtain the image of Ra-223 solution with the radioactivity of 0.72 MBq less than 10 sec acquisition time. With the same radioactivity, the intensity of optical signal in the solution part of the vial of Ra-223 was 21 times higher than that of F-18. In the air part in the vials, Ra-223 was 35 times higher luminescence than that of F-18. The light spectra for Ra-223 and F-18 solutions were similar but those in the air parts were different. The optical signals in the solution part of both Ra-223 and F-18 are attributed to Cerenkov-light while that in the air pat of Ra-223 is attributed to the scintillation of N2 gas in air. The luminescence of water with the alpha particles was not observed. We conclude that optical imaging of Ra-223 solution was possible and the light intensity was much higher than that of F-18 solution thus easier to detect and promising for optical imaging.

Keywords: Optical imaging, alpha emitter, Ra-223, radionuclide therapy
Poster panel
(face) ID: 131

Poster Number:

X-ray phase-contrast CT of invasive ductal carcinoma features based on grating interferometry, an ex-vivo study. (#2510)

X. Zhu1, 2, S. Wang3, 4, X. Li1, 2, W. Peng3, 4, Z. Chen1, 2, L. Zhang1, 2

1 Tsinghua University, Department of Engineering Physics, Beijing, China
2 Ministry of Education, Key Laboratory of Particle and Radiation Imaging (Tsinghua University), Beijing, China
3 Cancer Center Fudan University, Department of Radiology, Shanghai, China
4 Shanghai Medical College, Fudan University, Department of Oncology, Shanghai, China


X-ray phase-contrast imaging with Talbot-Lau grating interferometry, offering absorption, dark-field and phase contrast imaging simultaneously, has been proved to help increase soft tissue contrast and overcome some of the limitations, comparing with conventional absorption-based X-ray radiography. X-ray PCI was recently considered to have potential in future clinical applications of breast cancer, with the capability of identification of soft tissue microstructures and micro-calcifications. Here we reported an ex-vivo study of the phase contrast CT of formalin-fixed breast invasive ductal carcinoma specimens, revealing the better differentiation of phase-contrast  CT, which may help the identification of the breast tumors. Due to the limitations of the submitted summary, further comparison and detailed analysis will be later presented 

Keywords: phase contrast CT, Invasive Ductal Carcinoma
Poster panel
(face) ID: 134

Poster Number:

Towards diagnostic X-ray cone-beam imaging of the entire spine in weight-bearing position (#4186)

F. Noo1, A. Fieselmann2, M. Oktay1, M. Herbst2, L. Ritschl2, S. Vogt3, T. Mertelmeier2

1 University of Utah, Radiology and Imaging Sciences, Salt Lake City, Utah, United States of America
2 Siemens Healthcare, GmbH, Forchheim, Bavaria, Germany
3 Siemens Medical Solutions USA, Malvern, Pennsylvania, United States of America


X-ray CB imaging may play a tremendous role in the future of diagnostic radiology. Recently, an innovative, versatile X-ray system (Multitom Rax, Siemens Healthcare, GmbH, Forchheim, Germany) was introduced for diagnostic radiology. This system enables taking X-ray radiographs with high patient positioning flexibility, as well as acquiring semi-circular short cone-beam scans in a variety of orientations. Here, we are interested in programming this system to enable a novel data acquisition geometry that would cover the entire spine. Such an imaging capability has never been demonstrated so far. However, we may expect it to play an important clinical role as clinicians agree that spine diseases would be more accurately interpretable in weight-bearing position. We are furthermore interested in using a geometry that provides complete data so that CB artifacts may be avoided. We report on a first implementation of such a geometry that consists of two circular arcs connected by a line segment. We assessed immediate and short-term motion reproducibility, as well as ability to image the entire spine within a Rando phantom. Strongly encouraging results were obtained. Reproducibility with sub-mm accuracy was observed and the entire spine was accurately reconstructed.

Keywords: Cone-beam, diagnostic, weight-bearing, imaging
Poster panel
(face) ID: 137

Poster Number:

Material decomposition for spectral CT: application to calcium and iodine identification (#1790)

T. Su1, V. Kaftandjian1, P. Duvauchelle1, Y. Zhu2

1 Univ Lyon, INSA Lyon, Laboratoire Vibrations Acoustique, Villeurbanne, France
2 Univ Lyon, INSA Lyon, Universit Claude Bernard Lyon 1, CNRS, Inserm, CREATIS UMR 5220, U1206, Villeurbanne, France


Spectral CT is able to discriminate transmitted photons corresponding to different energy bins and produce spectral information with one data acquisition, making it possible to separate various materials. We propose a material decomposition method for spectral CT based on least log-squares criterion for the objective function, to identify calcium and iodine. We also simulated spectral CT imaging of a computational human thorax phantom. Extra calcium and iodine solutions at various concentrations (calcium: 30, 50, 100, 150, 200mg/cc; iodine: 5, 15, 25, 35, 45mg/cc) were placed within the heart region of the phantom. Decomposition results of the proposed method show that calcium and iodine are well separated and quantified from soft tissues, with 29.83%-4.73% average relative errors at different concentrations for calcium and 21.44%-5.90% for iodine.

Keywords: Computed tomography, spectral CT, material decomposition, photon-counting
Poster panel
(face) ID: 140

Poster Number:

Towards synchrotron speed laboratory based tomography - feasibility studies (#2392)

M. P. Pichotka1, 4, M. Weigt2, S. Hasn1, J. Jakubek3

1 CTU Prague, Institute of Experimental and Applied Physics (IEAP), Praha 2, Czech Republic
2 speCTive GmbH, Freiburg im Breisgau, Germany
3 Advacam s.r.o., Praha 7, Czech Republic
4 Academy of Sciences of Czech Republic, ITAM, Praha 9, Czech Republic


The relevance of fast tomographic analysis in a multitude of scientific fields often justifies the employment of expensive synchrotron technology, which benefits from very high photon flux and a tunable X-ray spectrum of high brilliance. The brilliance of synchrotron beams simultaneously permits for rather simple volumetric analysis and practically eliminates beam-hardening artefacts. Furthermore adjustment of the X-ray energy permits to maximize the imaging contrast, particularly for light materials but also to perform energy scans to obtain spectroscopic datasets.

A transfer of similar imaging speed to laboratory based tomographic faces several challenges. X-ray tubes, as standard high-flux laboratory sources, feature a broad energy spectrum balanced towards the lower energy tail. In non-spectroscopic tomographic application beam polychromaticy gives rise to beam hardening artefacts, therefore necessitating artificial confinement of the spectrum by beam filtering under significant loss of photon statistics.

However, previous research has demonstrated that spectroscopic tomography allows for utilization of the entire photon statistics of broad energy spectra. Spectroscopic algorithmics using common priorization to corellate statistics accross energy channels exists and have been tested successfully. Photon counting ASICS equipped with very efficient high-Z sensors and corresponding readout electronics, meeting the requirements for fast tomography, also are available. However we expect significant improvement by combination of features currently distributed to different detector types.

The current contribution evaluates the bottle-necks of high-speed spectroscopic tomography using current off-the-shelf photon counting detectors. Practical issues of such applications, such as dynamic correction of the detector and sensor response (ghosting, polarisation), on the fly methods etc., are implemented to this end and experimental verification is undertaken.

Keywords: high speed tomography, spectroscopic tomography, material decomposition, photon counting detectors, spectroscopic detectors
Poster panel
(face) ID: 143

Poster Number:

Image-domain Material Decomposition Based on Spectra Reconstruction Using Photon-counting CT (#2833)

R. Li1, L. Li1, Z. Chen1

1 Tsinghua University, Engineering Physics, Beijing, China


Spectral CT systems based on photon-counting detectors (PCDs) are showing multiple advantages and potentials in medical imaging applications because of their abilities of high energy discrimination capability. For these applications, obtaining incident spectra without distortion is always an important but difficult requirement since the discrimination capability directly depends on the accuracy of the attenuation coefficients obtained. Subject to the influence of charge sharing and photon escaping effects, spectra distortion is unavoidable for PCDs. In this paper, this problem is solved by spectra reconstruction method that consists of two steps. The first step is modeling the detector energy response based on X-ray fluorescence (XRF), and the second one is solving the reverse problem of detector response, with which the distortion-free spectra in the form of energy bins are reconstructed. Based on these bin data, image-domain material decomposition of four-material model (water, calcium, iodine and gadolinium) is conducted, using equivalent attenuation coefficients of each energy bin of the four materials, which is calculated by simulation with standard attenuation coefficients. The comparison of results between those with and without spectra reconstruction indicates that the failed decomposition due to spectra distortion are effectively corrected by the proposed anti-distortion method.

Keywords: spectral CT, photon-counting detector, material decomposition, spectra reconstruction, X-ray fluorescence
Poster panel
(face) ID: 146

Poster Number:

Joint Reconstruction of PET Attenuation and Activity from Scattered and Unscattered Data (#1004)

Y. Berker1, 2, J. S. Karp2, V. Schulz1

1 RWTH Aachen University, Physics of Molecular Imaging Systems, Aachen, North Rhine-Westphalia, Germany
2 University of Pennsylvania, Department of Radiology, Philadelphia, Pennsylvania, United States of America


In previous work, we have proposed scatter-to-attenuation reconstruction for positron emission tomography (PET). Scatter-to-attenuation reconstruction aims at recovering object attenuation information in the form of spatial electron-density distributions from coincident photons, one of which has been single-scattered. One idea is to interleave scatter-to-attenuation reconstruction, which inputs an activity distribution and outputs an attenuation map, with trues-to-activity reconstruction, which inputs said attenuation map and outputs an improved activity distribution. However, major uncertainties regarding the applicability of this approach revolve around a) the unknown impact of the initial activity estimate; b) evaluation of reconstructed activity distributions, and c) convergence to the correct solution. Methods: Using low-dimensional simulated PET data (mouse-sized, 18x18-voxels phantom), we start with mostly uniform initial activity and attenuation estimates and iteratively apply maximum-likelihood expectation-maximization (MLEM) and a maximum-likelihood gradient-ascent (MLGA) algorithms to update activity (from unscattered data) and attenuation (from scattered data), respectively. We evaluate results in terms of log-likelihoods of the expected scatter histograms, and normalized mean squared errors with respect to reference image-space distributions of activity and attenuation. Results: In our study, both attenuation and activity converged to the reference distributions, despite MLEM and MLGA starting with incorrect attenuation and activity estimates, respectively. Conclusion: The MLGA scatter-to-attenuation reconstruction algorithm, in combination with MLEM trues-to-activity reconstruction, jointly reconstructs attenuation maps and attenuation-corrected activity distributions from scattered and unscattered coincidences without reliance on a-priori information about the activity distribution.

Keywords: PET, image reconstruction, algorithms, attenuation, Compton scattering
Poster panel
(face) ID: 149

Poster Number:

Event-by-event Non-Rigid Data-driven PET Respiratory Motion Correction Methods: Comparison of PCA and COD (#1406)

S. Ren1, Y. Lu2, O. Bertolli3, K. Thielemans3, R. E. Carson1, 2

1 Yale University, Department of Biomedical Engineering, New Haven, Connecticut, United States of America
2 Yale University, Department of Diagnostic Radiology, New Haven, Connecticut, United States of America
3 University College London, Institute of Nuclear Medicine, London, United Kingdom of Great Britain and Northern Ireland


Respiratory motion is a major cause of degradation of PET image quality. Respiratory gating and motion correction can be performed to reduce the effects of respiratory motion; these methods require motion information, typically from external motion tracking systems. Various groups have studied data-driven motion estimation methods, among which, the Principal Component Analysis method (PCA) was shown to have better and more stable results than other methods. More recently, a new data-driven respiratory motion estimation method was established by calculating the centroid of distribution (COD) of listmode events, which was used with event-by-event respiratory motion correction (EBE-MC), and showed results comparable to those with external motion tracking device. The EBE-MC method only corrected for rigid motion, so that non-rigid components still contributes to motion-induced blurring. Thus, it is desirable to further develop data-driven nonrigid respiratory motion correction (NRMC) to achieve the best respiratory motion correction results. We first evaluated 2 data-driven respiratory motion detection methods, COD and PCA, by comparing the extracted respiratory motion to that acquired by the Anzai system in dynamic studies with two tracers. Then, we developed and performed data-driven EBE-NRMC using either COD- or PCA-derived respiratory motion information. Data-driven correction results were compared with Anzai-based results, with no-motion correction reconstructions performed for comparison. For all tested studies, both COD and PCA showed good-to-excellent match with Anzai signals, with PCA showing a higher correlation with Anzai than COD signals. The data-driven EBE-NRMC results showed that both COD and PCA provide comparable image quality improvement comparing with Anzai-based correction. Although COD shows a lower correlation with Anzai than PCA, COD-based NRMC results are comparable to that of PCA, both of which showed great decrease in motion-induced blurring.

Keywords: non-rigid respiratory motion correction, data-driven respiratory motion detection
Poster panel
(face) ID: 152

Poster Number:

The use of low-dose CT intra- and extra-nodular image texture features to improve small lung nodule diagnosis in lung cancer screening (#1613)

R. Yan1, 3, S. Park4, S. Ashrafinia2, 3, J. Lee4, L. C. Chu3, C. T. Lin3, A. Hussie3, J. Steingrimsson5, A. Rahmim3, 2, P. Huang1, 5

1 Johns Hopkins, Department of Oncology, Baltimore, Maryland, United States of America
2 Johns Hopkins University, Department of Electrical and Computer Engineering, Baltimore, Maryland, United States of America
3 Johns Hopkins University, Department of Radiology, Baltimore, Maryland, United States of America
4 Johns Hopkins University, Radiation Oncology and Molecular Radiation Sciences, Baltimore, Maryland, United States of America
5 Johns Hopkins University, Department of Biostatistics, Baltimore, Maryland, United States of America


Although extensive studies have shown that computer-aided diagnosis (CAD) could greatly improve the diagnostic accuracy, limited study has conducted on low-dose CT (LDCT) images where texture features are generally much weaker and having much larger variations. We introduce a new CAD image feature extraction method to quantify nodule heterogeneity patterns from intra-nodular, surrounding lung parenchyma and extra-nodular areas. This approach was applied to a matched biopsy confirmed case-control (cancer-benign) sample with small lung nodules from the National Lung Screening Trial (NLST). Cancers and benign controls were matched by known risk factors including age, gender, smoking status, COPD status, BMI, and image appearances. Using CAD with nonparametric machine learning, we developed a probability of malignancy (Pm) score from the training set and applied it to the test set. In the validation set, Pm scores from cases were significantly higher than that from the benign controls (p<0.0001). The area under the ROC curve from CAD was 0.9154. The [sensitivity, specificity, positive predictive value, negative predictive value] from CAD and the combined reading of three radiologists were [0.95, 0.88, 0.86, 0.96] and [0.70, 0.69, 0.64, 0.75] respectively. The overall prediction accuracy from CAD (91%) was significantly higher than radiologist’s reading (70%, two-sided p value= 0.0180). Our CAD method increases the diagnostic accuracy by providing independent information other than the known risk factors. Extending our approach into clinical practice by incorporating known risk factors will likely further improve the diagnostic accuracy when evaluating a random sample.

Keywords: Computer-Assisted Diagnosis, Cancer Screening, Lung Cancer, Machine Learning, Computed Tomography, Low-Dose CT, Radiomics
Poster panel
(face) ID: 155

Poster Number:

Impact of μ-map Processing and Transmission Scan Count Statistics on Quantification of HRRT PET Pig Brain Scans (#1993)

S. H. Keller1, B. Vigfúsdóttir2, 1, J. Villadsen2, L. M. Jørgensen2, 3, H. D. Hansen2, M. Sibomana4, G. M. Knudsen2, 3, C. Svarer2

1 Rigshospitalet, University of Copenhagen, Dept. of Clinical Physiology, Nuclear Medicine and PET, Copenhagen, Denmark
2 Rigshospitalet, NRU, Copenhagen, Denmark
3 University of Copenhagen, Faculty of Health and Medical Sciences, Copenhagen, Denmark
4 Sibomana Consulting, Emines, Belgium


Aim: In this work we evaluate the two μ-map methods TXTV and MAP-TR at two different transmission scan acquisition speeds (with different count statistics) against gold standard CT-based μ-maps on PET pig brain scans.  

Material and methods: Nine HRRT pig brain scans with different tracers reconstructed in 5 versions with different μ-maps  (speed 50 or speed 10 transmission scan, each reconstructed and processed using either MAP-TR or TXTV, and a gold standard μ-map from a same day CT scan) were registered to a pig brain atlas. From TACs on relevant VOIs relative differences (∆%) from the CT-based PET to the 4 HRRT TX-based PET and area under the curves (AUC) were generated along with μ-map profile plots.

Results: AUCs showed 2-10% (MAP-TR) and 3-15% (TXTV) difference from CT-based μ-maps on 11 VOIs with almost no difference between using speed 10 or 50 transmission scan acquisitions. Similar differences were found in relative difference, but with an unexpected change over time, which we found to be correlated with a similar change in scatter fractions over time caused by PET to μ-map mismatches when applying μ-maps for masking in scatter correction and scatter scaling. This was primarily seen when using CT-based μ-maps as the fixation of the pig’s head and neck was changed when moving it to the CT scanner giving a warp of soft tissue causing a considerable mismatch to HRRT PET.      

Conclusion: The problems with our gold standard CT-based μ-maps makes it impossible to find the best HRRT TX method for pig brain scans, but from our results, we recommend using MAP-TR on speed 50 transmission scans for HRRT pig brain imaging until a new study with matched HRRT-CT head and neck fixation has been conducted.

Keywords: PET, HRRT, Attenuation correction, Pig brain scans, quantification
Poster panel
(face) ID: 158

Poster Number:

Super-resolution PET Image Reconstruction with Sparse Representation (#2228)

Z. Hu1, T. Li1, Y. Yang1, X. Liu1, H. Zheng1, D. Liang1

1 Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Lauterbur Research Center for Biomedical Imaging, shenzhen, Guangdong, China


In the positron emission tomography (PET) field, reconstructed images are often blurry and contain noise. These problems are mainly caused by the low-count problem. As sparse technology becomes more widely used, sparse prediction is increasingly prone to be selected to solve the problem. In this paper, we propose a new sparse prior method to process low-resolution PET reconstructed images. In the proposed strategy, two dictionaries ( D1 for low-resolution PET images and D2 for high-resolution PET images) are trained from numerous real PET image patches. Subsequently, D1 is used to obtain the sparse representation for each patch of the input PET image. Finally, a high-resolution PET image is generated from this sparse representation using D2 . The results of experiments indicate that the proposed method exhibits stable and superior effects that enhance image resolution, detail recovery. Quantitatively, this method achieves a better performance than traditional methods in terms of root mean square error (RMSE). The proposed strategy provides a new and efficient approach to improve the image quality of reconstructed PET images.

Keywords: Positron emission tomography (PET), super-resolution, sparse representation, dictionaries
Poster panel
(face) ID: 161

Poster Number:

Reproducibility of Cold Uptake Radiomics in 99mTc-Sestamibi SPECT imaging of Renal Cell Carcinoma (#2453)

S. Ashrafinia2, 1, K. Jones1, M. A. Gorin3, S. P. Rowe1, M. S. Javadi1, M. G. Pomper1, M. E. Allaf3, A. Rahmim1, 2

1 Johns Hopkins University, Radiology and Radiological Sciences, Baltimore, Maryland, United States of America
2 Johns Hopkins University, Electrical and Computer Engineering, Baltimore, Maryland, United States of America
3 Johns Hopkins University, Department of Urology, Baltimore, Maryland, United States of America


99mTc-Sestamibi SPECT/CT imaging of renal cell carcinoma (RCC) is recently shown to promisingly distinguish between the benign oncocytomas and malignant RCC, where the former appears as a high tumor uptake and the latter as a cold uptake. Aiming towards radiomics analysis of the cold uptake in SPECT images for potentially discriminating between RCC subtypes, the first step is to assess the reproducibility and reliability of radiomic features. 99mTc-Sestamibi SPECT/CT images of 50 patients with renal mass, plus a contrast-enhanced CT or MRI, were used by a trained radiologist to segment the region of interest (ROI) on the SPECT image. The impact of segmentation on reproducibility was studied by generating three ROIs through removing 1 voxel from, or appending 1 and 2 voxels to the manual ROIs (ROI0, ROI-1, ROI+1, ROI+2). Images were then uniformly quantized with 8 grey-levels(GL) (4 to 512). A total of 363 radiomic features, verified within the Image Biomarker Standardization Initiative, were generated for all 50 patients/4 ROIs/8 GLs.

Voxel intensity in SPECT images is the number of counts, and is a patient-dependent non-normalized quantity; thus, (i)fixed number of bin quantization should be chosen over fixed bin-width, and (ii)non-normalized/non-quantized statistical features should be excluded. The intra-class correlation (ICC) type C-1 was calculated for all features between all four ROIs and all except ROI-1, showing 204 features with ICC>0.7 and the same 204 features with ICC>0.85, respectively. However, a subset of 204 features with emphasis on high GLs across 2nd-order feature classes should be further excluded due to the scarce presence of voxels with high intensities that are prone to segmentation. Moreover, reproducibility analysis of quantization using Spearman rank-correlation showed GLs below 32 should be avoided due to high variation and inconsistencies. The resulting features and settings are recommended for further investigation of predictive value.

Keywords: Radiomics, Texture analysis, heterogeneity analysis, SPECT, renal cell carcinoma, quantitative imaging, 99TC-sestamibi
Poster panel
(face) ID: 164

Poster Number:

Continuous MRI-based Motion Estimation for PET Motion Correction in Integrated PET/MRI (#2798)

L. J. Frohwein1, M. Heß1, F. Büther1, 2, K. P. Schäfers1

1 University of Münster, European Institute for Molecular Imaging (EIMI), Münster, North Rhine-Westphalia, Germany
2 University Hospital of Münster, Department of Nuclear Medicine, Münster, North Rhine-Westphalia, Germany


Respiratory and cardiac motion during a PET acquisition potentially causes image blurring degrading the quality of tracer uptake quantification. Hybrid PET/MRI machines, with which PET and MRI can be acquired simultaneously, offer the possibility to utilize MRI data for motion estimation and correction of PET data. Most of the existing motion correction methods are model-based and thus not able to capture irregular motion patterns or bulk motion of the patient. In this work, we present a novel MRI-based method to determine 3D motion throughout the entire PET scan duration. For this purpose, we use a standard single-shot TurboFLASH sequence with a short T_R (=slice repetition time). 20 slices are acquired in interleaved mode forming two sub-volumes (odd and even slices). Using the respiratory signal from principal component analysis and the ECG R-peaks, temporally coherent 3D volumes can be formed with a temporal resolution of 350 ms per volume by complementing each sub-volume with slices of the other. Each newly formed full volume is non-rigidly registered to a reference volume creating motion vector fields for every time-step.

Keywords: PET/MRI, PET-MRI, PET motion correction
Poster panel
(face) ID: 167

Poster Number:

Implementation and Validation of an efficient decomposition based system matrix approach incorporating subject’s physical phenomena (#3005)

B. Auer1, 2, F. Boisson1, 2, V. Bekaert1, 2, D. Brasse1, 2

1 Université de Strasbourg, Institut Pluridisciplinaire Hubert Curien (IPHC), Strasbourg, France
2 CNRS/UMR7178, Strasbourg, France


In small animal Single Photon Emission Computed Tomography (SPECT), attenuation and scatter introduce important artifacts in the reconstructed images, which could lead to misdiagnosis for subject’s follow-up. Gold standard Monte Carlo Simulation (MCS) is one of the well-established tools that has been used in SPECT image reconstruction due to its ability to accurately model photon transport.

However, MCS requires extensive computation time to obtain a low noise system matrix and are therefore inappropriate for the rate of daily exams performed in both clinical and preclinical routine: an improvement in simulation speed is thus mandatory.

In this work, we validated, compared to a state of the art approach and by using a modified NU-4 IQ phantom, our efficient and simplified modeling of the physical phenomena occurring in the subject. Our approach based on a system matrix decomposition, associated to a scatter pre-calculated database method, demonstrated an acceptable time on a standard computer for daily imaging small animal follow-up (∼ 1h), leading to a personalized image reconstruction.

The reconstruction workflow leads to significant image artifacts reduction as well as a 13% (on average) improvement in terms of recovery coefficients. Results presented in this study, conduct to the validation of the developed approach in comparison with a state of the art one which appears to be far too long for a daily exam perspective.

Keywords: small animal SPECT, Monte Carlo Simulation, Attenuation and scatter correction, System matrix, daily exam perspective
Poster panel
(face) ID: 170

Poster Number:

Component-Based Normalization for a 1mm3 Resolution Clinical PET System (#3214)

D. L. Freese1, D. F. - C. Hsu1, D. Innes2, C. S. Levin2, 1

1 Stanford University, Electrical Engineering, Stanford, California, United States of America
2 Stanford University, Radiology, Stanford, California, United States of America


We are constructing a 1 mm3 resolution clinical PET system dedicated to locoregional cancer imaging, focusing primarily on the breast as well as the head and neck regions. The two-panel system comprises 98,304 1x1x0.9 mm LYSO crystals, for a total of 2.4e9 LORs, thus, it is impractical to gather even a low number of statistics per LOR for normalization.  Based upon LOR symmetry, we combine our 2.4e9 LORs into 3.25e6 symmetric LORs (SLORs).  We simulate the two-panel system and a uniform slab of activity, and demonstrate improvements in noise when using the SLOR efficiencies to generate a component-based normalization.  We then show in simulation the ability to jointly estimate crystal efficiencies and geometric (SLOR) efficiencies.  Finally, we image a phantom comprising spheres of 4.0mm to 7.9mm in diameter in a warm background with 5:1 contrast where we demonstrate a 33.2% improvement in contrast-to-noise ratio when using component-based normalization compared to direct normalization.

Keywords: PET, PEM, Normalizatoin, Correction, Clinical, Limited-angle tomography
Poster panel
(face) ID: 173

Poster Number:

Non-rigid Event-by-event Body Motion Correction with Automated Data-driven Motion Detection for Static and Dynamic PET (#3296)

Y. Lu1, S. Ren1, J. - D. Gallezot1, M. Naganawa1, K. Fontaine1, J. Wu1, T. Mulnix1, V. Y. Panin2, M. E. Casey2, R. E. Carson1, C. Liu1

1 Yale University, New Haven, Connecticut, United States of America
2 Siemens Medical Solutions USA, Inc., Molecular Imaging, Knoxville, Tennessee, United States of America


Background: Patient body motion is a well-documented source of error in PET imaging, particularly for dynamic studies with long scan durations. Body motion between CT and PET can cause additional attenuation mismatch artifacts. Conventional dynamic image frame-based detection/correction approaches suffer from intra-dynamic-frame motion, which is not corrected.

Methods: We investigated an automated data-driven body motion detection method based on a centroid-of-distribution (COD) trace, followed by an event-by-event non-rigid correction. Body motion between CT-PET scans was corrected by registering CT to MLAA generated attenuation map. The proposed detection and correction method were evaluated using two human dataset with 18F-FPDTBZ (pancreas tracer) and 18F-FDG (lung cancer), and a dog cardiac data with 18F-ROS (reactive oxygen species). We compared our approach to the conventional frame-based post reconstruction registration method.

Results: COD can sensitively and reliably detect body motion. With intra frame motion correction, the event-by-event correction showed a substantial improvement over conventional frame-based registration approach in static and dynamic studies. Non-rigid motion correction provided a superior performance as compared to rigid motion correction methods.

Conclusions: The proposed methods of data-driven body motion detection and non-rigid correction can effectively eliminate body motion, particularly for dynamic PET with longer scan durations.

Keywords: Body motion detection, data-driven, body motion correction, non-rigid event-by-event correction, PET
Poster panel
(face) ID: 176

Poster Number:

Dual-spillover Correction Does Not Improve SPECT Myocardial Blood Flow Measurement (#3506)

R. G. Wells1, J. Renaud1, R. A. deKemp1, T. D. Ruddy1

1 University of Ottawa Heart Institute (UOHI), Cardiac Imaging, Ottawa, Ontario, Canada


Purpose: Myocardial blood flow (MBF) measurement is possible using stationary cardiac SPECT cameras. Limited spatial resolution and blood vessels in the myocardium produce mixing of the blood and myocardial time-activity curves used for kinetic analysis. Our purpose was to evaluate the effect of correction for both spillover into the myocardium from blood and into the blood from myocardium on the accuracy of SPECT MBF estimation.

Methods: Images from 31 patients who had dynamic 1-day rest-stress myocardial SPECT studies using Tc99m-tetrofosmin were retrospectively processed both with attenuation correction (AC) and without (NC), and using a single spillover correction (SSO = blood to myocardium) and a dual spillover correction (DSO).  K1 values from kinetic analysis with a 1-tissue compartment model were fit to a Renkin-Crone extraction fraction model (EF=1-exp(-a-b/MBF)) using 50 repeats of 2-fold cross validation to determine the method-specific K1-to-MBF conversion function. True MBF was measured using the clinical standard of PET imaging, with either Rb82 or N13-ammonia, in the same patients.

Results: The model fit of SPECT K1 to PET MBF was not improved with dual spillover correction. With SSO, the Renkin-Crone model parameters were a=0.140, b=0.282 (AC, R2=0.71) and a=0.187, b=0.400 (NC,R2=0.75).  With DSO, the parameters were a=0.110, b=0.265 (AC, R2=0.63) and a=0.161, b=0.352 (NC,R2=0.67). The 95% confidence limits for the global relative MBF difference (SPECT – PET MBF/mean) were similar for dual- vs single-spillover correction, both with AC ([-0.74 – 0.77] ml/ming/g vs [-0.62 – 0.64] ml/min/g) and without ([-0.67 – 0.71] ml/ming/g vs [-0.55 – 0.59] ml/min/g).

Conclusion: Dual spillover correction alters the K1-to-MBF conversion function but does not improve the accuracy or precision of SPECT MBF measurements.

Keywords: SPECT, myocardial blood flow, spill-over correction, dynamic imaging, cardiac imaging
Poster panel
(face) ID: 179

Poster Number:

Learning-based attenuation correction for brain PET/MRI using artificial neural networks (#3561)

B. Yang1, J. Tang1

1 Oakland University, Department of Electrical and Computer Engineering, Rochester, Michigan, United States of America


Hybrid PET/MRI systems do not provide direct measurement of attenuation maps for PET image reconstruction. Learning-based techniques, without the need for deformable registration and quick in applying the trained model, have achieved promising results for deriving attenuation maps from MR images. The purpose of this study is to use the artificial neural network (ANN) in building a learning-based attenuation map prediction scheme, to combine patch operation into the learning procedure and to handle subject-specific anatomical structures. We cast the attenuation map prediction as a regression problem that models a nonlinear mapping between the MR image patches and the corresponding attenuation map patches. An ANN with supervised learning was used to solve this regression problem through learning from examples. Using the BrainWeb phantoms, we simulated MR and attenuation map atlas pairs of 20 subjects. For each anatomical model, the BrainWeb provides simulated T1-weighted MR image. The corresponding attenuation map was created by assigning different attenuation coefficients for bone, air, and soft tissues. The ANN was trained with the image patches from the atlas pair of 1 subject using a backpropagation algorithm. We tested the trained ANN to predict the attenuation maps for the other 19 subjects. To quantitatively evaluate the prediction results, the mean absolute error comparing the predicted attenuation map with the atlas correspondence was calculated on regions of soft tissues and bone. The mean relative errors for soft tissues and bone across 19 subjects are 2.2% and 6.6%, respectively. We demonstrate that the predicted attenuation maps achieve strong agreement with the atlas correspondence and the ANN model trained using the atlas pair of one subject applies well to other subjects. Our attenuation map predicting scheme is promising for brain PET/MR imaging and its accuracy in reconstructed PET images is currently under evaluation.

Keywords: MR based attenuation correction, artificial neural networks, PET/MRI
Poster panel
(face) ID: 182

Poster Number:

Super-resolution based on a new approach for thoracic motions correction in cardiac PET (#3857)

M. A. A. Ahmed1, P. Xiao1, 2, Q. Xie1, 2

1 Huazhong University of Science and Technology, Biomedical Engineering, Wuhan, Hubei, China
2 Wuhan National Laboratory for Optoelectronics, Wuhan, Hubei, China


In several studies applying super-resolution to the field of thoracic motions correction in PET imaging, the estimation of the inter-frame motions are based on noisy PET images resulting from the gating process. Consequently, this leads to inaccurate motion estimation. The usage of 4D CT images for the motion estimation is unacceptable either due to radiation exposure increase.

The New Approach for simultaneous respiratory and cardiac Motion Correction (NAMC) is a method recently introduced as gating alternative. Motivated by its high SNR, and contrast, super-resolution algorithm based NAMC approach is proposed (NAMC-SR). The NAMC-SR method is evaluated in comparison to standard Gating-Based Super-resolution using realistic PET imaging using GATE simulation of the thorax with respiratory and cardiac motion simulated the thorax digital phantom (XCAT).

The comparison between the NAMC-SR and Gating-Based Super-resolution approaches showed that the NAMC-SR is far better in terms of resolution, SNR and the motion compensation accuracy. Therefore, this new approach will have its positive impact on solving the problem of respiratory and cardiac motion in thoracic PET imaging and the improvement of its quality.

Keywords: Super-resolution, gating, NAMC, motion correction, cardiac PET
Poster panel
(face) ID: 185

Poster Number:

PET-driven respiratory phase tracking and self-gating of PET data: clinical demonstration of enhanced lesion detectability in cardiovascular PET/MR imaging (#4014)

N. A. Karakatsanis1, P. M. Robson1, M. R. Dweck1, 2, M. G. Trivieri1, C. Calcagno1, V. Mani1, D. D. Faul3, C. Tsoumpas1, 4, Z. A. Fayad1

1 Icahn School of Medicine at Mount Sinai, Translational and Molecular Imaging Institute, New York, New York, United States of America
2 University of Edinburgh, British Heart Foundation, University Centre for Cardiovascular Science, Edinburgh, United Kingdom of Great Britain and Northern Ireland
3 Siemens Healthcare, Malvern, Pennsylvania, United States of America
4 University of Leeds, Division of Biomedical Imaging, School of Medicine, Leeds, United Kingdom of Great Britain and Northern Ireland

This work was supported by NIH/NHLBI R01HL071021 grant


The advent of cardiovascular PET/MR in clinic may offer superior motion compensation capabilities by exploiting the non-ionizing nature of MR to sufficiently track and model respiratory and cardiac motion at high spatial resolution. However, the estimated motion models need to be continuously driven by the respiratory and cardiac motion phase for the accurate motion correction of the PET and MR data, thereby demonstrating the need for constant cardio-respiratory phase tracking throughout the entire scan period. Although MR-based continuous monitoring of cardio-respiratory phase is possible, it could reserve valuable time from other diagnostic MR sequences, thus diminishing PET/MR clinical potential. In this study, we validate a readily applicable in clinic PET list-mode (LM) data-driven respiratory phase extraction method to enable robust respiratory self-gating of PET data. The method relies on the hypothesis that for certain tracers the total LM PET counts can be sensitive to the periodic movement of hot or cold activity regions in and out of the PET field of view during breathing. The respiratory phase is tracked by continuously monitoring the total LM counts temporal profile, after smoothing with moving average filters (MAFs) to automatically correct for deep breath hold and drifting patterns. Subsequently the phase is used to drive the self-gating of the PET data into five respiratory gates and a special breath-hold gate. The period of the respiratory phase extracted with our proposed method matched well with the expected period of normal human breathing. Moreover, the automatically identified irregular LM PET count patterns corresponded in time to breath-hold MR acquisitions. The clinical application of the proposed method on our cardiovascular PET/MR studies demonstrated feasibility, while quantitative assessment of 18F-NaF coronary lesions detectability suggested a 10% improvement in lesion contrast and contrast-to-noise scores for the self-gated PET images.

Keywords: PET/MR, respiratory, phase, monitoring, gating, data-driven, irregular, breathing, coronary, motion, compensation
Poster panel
(face) ID: 188

Poster Number:

Improvement of the Sign Determination Method for Data-Driven respiratory signal in TOF-PET (#4086)

O. Bertolli1, S. Arridge2, C. W. Stearns3, S. D. Wollenweber3, B. F. Hutton1, K. Thielemans1

1 University College London, Institute of Nuclear Medicine, London, United Kingdom of Great Britain and Northern Ireland
2 University College London, Centre for Medical Image Computing, London, United Kingdom of Great Britain and Northern Ireland
3 GE Healthcare, Waukesha, Wisconsin, United States of America


Respiratory gating and motion correction can reduce resolution loss in PET chest imaging, and require a respiratory signal. Data-driven (DD) methods produce a reliable respiratory signal from PET data, avoiding the use of devices. Principal Component Analysis (PCA) is an easy to implement DD algorithm whose signals are closely correlated with respiration, except for the unique determination of the direction of motion.

In previous work, we developed a method to determine the PCA signal direction in non-TOF PET, assuming the biggest change in the data is due to axial motion. Here we extend the method to TOF data, showing that using TOF considerably improves its performance especially when combined with selection of parts of the sinogram corresponding to internal region of the chest.

TOF data are unlisted into 5-D dynamic sinograms. The effect of rigid axial motion is simulated in sinogram space to obtain a trace u(t) representing motion in a known direction. The PCA signal w(t) is then compared to u(t) via Pearson correlation. If negative, w(t) is flipped. To reduce the influence of outer parts of the body, we apply a selection on the input data for PCA and the sign-determination method. The sinograms (TOF or non-TOF) are masked with the projection of a centrally placed cylinder, to remove the outer parts.

Correlation between u(t) and w(t) was evaluated on 16 FDG-PET 360s patient studies (acquired with the respiratory device RPM), and on shorter intervals (50,100,200, 300s), with non-TOF and TOF data, without and with masking. RPM was used to check the PCA sign.

The use of TOF-sinograms with masking provided distinctly higher values for the correlation between u(t) and w(t) than in the non-TOF case, with and without the mask, on all intervals. Higher values suggest that u(t) is in good agreement with w(t) and therefore a reliable method to determine the PCA sign. In summary, the increased spatial information provided by TOF is beneficial for the sign-determination method.

Keywords: PET, respiratory gating, PCA, sign-determination
Poster panel
(face) ID: 191

Poster Number:

Optimization of Q.Clear reconstruction for noisy Y-90 PET images (#1903)

H. Ma1, X. Hou3, F. Benard2, A. Celler3

1 University of British Columbia, Physics and Astronomy, Vancouver, British Columbia, Canada
2 British Columbia Cancer Agency, British Columbia Cancer Research Centre, Vancouver, British Columbia, Canada
3 University of British Columbia, Radiology, Vancouver, British Columbia, Canada


Liver radioembolization therapy with Y-90 requires that prescribed activity be delivered to the targeted mass, while normal liver is spared. Activity distribution in the liver is usually verified by performing post-injection imaging studies. Quantitative Y-90 time-of-flight (TOF) PET has been shown to be feasible, but due to low intensity of positrons, the quality of images is poor. It can be improved by using regularized reconstruction algorithms such as Q.Clear, with smoothness of the image adjusted by the β parameter. A recent patient study suggested a constant β=4000.  Our objective was to check this claim and optimize β as a function of image noise. The IEC phantom with a lung insert and 6 spheres (d=0.5-37mm) was filled with 3GBq of Y-90 (SBR=7.5). Four scans were acquired (day 0, 3, 5 and 7) with sphere activity concentrations decreasing from 2.45MBq/mL to 0.40MBq/mL. The images were reconstructed with Q.Clear and increasing β (from 1000-8000 for day 0 to 4000-22000 for day 7). We optimized β as a function of activity using figures-of-merit adopted from NEMA: cold contrast recovery ratio (Qc) and contrast-to-noise ratio (CNR). We also calculated background variability (BV). For each noise level, the values of Qc and CNR were plotted as a function of β for the four largest spheres. Parameter β was considered to be optimized when CNR and Qc reached their respective maxima. The quality of images reconstructed with optimized b was substantially improved over those obtained with β=4000. Optimization based on Qc and CNR yielded quite similar b (b=2000, 6000, 10000 and 18000 for Qc and b=2500, 6000, 8000 and 12000 for CNR for days 0, 3, 5 and 7, respectively). The BVs ranged from 11% - 17%. These results indicate that the visual quality of TOF-PET Y-90 images can be substantially improved by adjusting parameter β used in the Q.Clear reconstructions to the noise level of the data. This can be achieved by searching for maximum values of CNR and Qc as a function of β.

Keywords: Reconstruction, Regularized, Liver Radioembolization, Y-90 PET imaging
Poster panel
(face) ID: 194

Poster Number:

IQ measurements proving dose reduction potential of a new CT geometry - a simulation study (#2253)

N. Saeid Nezhad1, K. Kumar1, C. Hoeschen1

1 Otto von Guericke universität Magdeburg, Institute of Medical Engineering, Magdeburg, Saxony-Anhalt, Germany


A new promising type of a CT scanner combining advantages of fourth, third and first generation has been proposed to reduce the ionizing radiation exposure to the patient. The proposed system is a combination of a new geometrical design called WATCH (Well-Advanced Technique for Computed Tomography with High resolution) system with a novel image reconstruction algorithm called Orthogonal Polynomial Expansion on Disk (OPED). For the WATCH system a spatial resolving ring detector is employed with the x-ray source fixed in the center of the detector ring. The source rotates around the object while the detector ring performs a coupled movement in such a way that x-ray source always stays at the center of the detector ring and detector ring keeps its initial orientation. The purpose of this study is to quantify the patient radiation dose reduction of this system in comparison to a conventional CT scanner for the first time by utilizing a MC simulation based on the Geant4 toolkit. There are two exposure reduction possibilities for the WATCH system. The first possibility is a dose reduction during the collection of the projections and the second one is a dose reduction resulting from employment of the new geometry and a corresponding reconstruction algorithm. The collection of the data with the WATCH system requires to gather less projections than the short scan of clinical CT systems by twice the fan angle. In addition, higher image quality of the reconstructed image from the novel scanner has been investigated through objective image quality assessment parameters. This paper focuses on the evaluation of image contrast, noise, contrast to noise ratio, modulation transfer function, noise power spectrum and noise equivalent quanta (NEQ) characteristics. The novel system shows about 25 percent better NEQ in comparison to the conventional CT scanner. The patient dose could still be reduced by improving the OPED algorithm which can potentially reduce the noise in CT images.

Keywords: Micro-CT, MTF, NPS, NEQ, Monte carlo simulation
Poster panel
(face) ID: 197

Poster Number:

Quantitative evaluation of breast CT reconstruction by means of figures of merit based on similarity metrics (#3031)

P. Oliva1, B. Golosio2, F. Arfelli3, P. Delogu4, F. Di Lillo5, D. Dreossi6, V. Fanti2, L. Fardin3, C. Fedon7, E. Loi2, G. Mettivier5, L. Rigon3, P. Russo5, A. Sarno5, G. Tromba6, R. Longo3

1 University of Sassari and INFN Cagliari, Dipartimento di Chimica e Farmacia, Sassari, Italy
2 Università di Cagliari and INFN Sezione di Cagliari, Dipartimento di Fisica, Cagliari, Italy
3 Università degli Studi di Trieste and INFN Sezione di Trieste, Dipartimento di Fisica, Trieste, Italy
4 Università di Siena, and INFN Sezione di Pisa, Dipartimento di Scienze Fisiche, della Terra e dell’Ambiente, Siena, Italy
5 Università di Napoli Federico II and INFN Sezione di Napoli, Dipartimento di Fisica “Ettore Pancini”, Napoli, Italy
6 Elettra-Sincrotrone Trieste S.C.p.A, Basovizza, Italy
7 INFN Sezione di Trieste, Trieste, Italy


Iterative CT reconstruction algorithms coupled with edge-preserving filters are attracting a growing interest in the field of biomedical X-ray imaging. In many cases such algorithms demonstrate improved reconstruction quality compared with analytical reconstruction algorithms, for instance in the case of measurements with a reduced number of projections. Their performance is often evaluated on test phantoms using conventional figures of merit, as for example contrast-to-noise ratio or spatial resolution at sharp edges. However, this approach can lead to an optimistic evaluation of the reconstruction quality: compared to test phantoms, biomedical images typically present complex structures and strong inhomogeneities. The reconstruction algorithms can generate artifacts that can hide small details, produce fake structures or alter the shape of actual ones.

Figures of merit based on similarity metrics are a valuable tool for an objective evaluation of the reconstruction quality, in particular in the case of biomedical images.

In this work we present an application of such figures of merit to breast-CT reconstruction with a simultaneous algebraic reconstruction technique (SART) algorithm combined with a bilateral filter, at varying number of projections. The results show that, for a given reconstruction, the figures of merit are fairly stable for a wide range of variations of the bilateral filter parameters, thus showing the robustness of the reconstruction technique.

Keywords: CT reconstruction, Similarity metrics, Image quality
Poster panel
(face) ID: 200

Poster Number:

A new method to calculate Energy Resolution based upon NEC phantom (#3658)

D. K. Bharkhada1, H. Rothfuss1, M. Conti1

1 Siemens Medical Solutions USA, Inc.,, Knoxville, Tennessee, United States of America


Energy resolution is an important parameter which affects the capability of scanner to distinguish scatter counts from trues count. A better energy resolution can reduce the amount of scatter in the acquired data. In this article we propose a new method to measure energy resolution using the NEMA NEC phantom. This method is evaluated by comparing the energy resolution obtained from proposed method with that obtained using a line-source in air at center of scan FOV. The measurement is performed on Siemens next generation PET/CT SiPM PET/CT scanner. Energy resolution changes slightly over the whole count rate range. At low count rate, an energy resolution of 10.07% is observed using line-source in NEC phantom after proper corrections, and 10.04% is obtained for line-source in air

Keywords: NEMA; Energy Resolution; PET
Poster panel
(face) ID: 203

Poster Number:

Attenuation Correction of Cerebellum in PET/MR Data (#4036)

E. Rota Kops1, H. Hautzel2, H. R. Herzog1, C. Lerche1, N. J. Shah1

1 Forschungszentrum Jülich GmbH, Institute of Neuroscience and Medicine (INM-4), Juelich, Germany
2 Heinrich-Heine-University Duesseldorf, Dept. of Nuclear Medicine at the Forschungszentrum Juelich, Duesseldorf, Germany


Most approaches of modelling neuroreceptor PET studies apply the cerebellum as reference area. Hence, the respective attenuation correction (AC) method for reconstructing the PET data should be most appropriate regarding these areas. PET data from PET/MR scanners, requiring alternative AC methods, has to be tested with respect to the performance within the cerebellum. This study aimed to compare various AC methods for PET/MR data focussing on the cerebellum. Data of 16 subjects undergoing 18FDG imaging in the Siemens 3TMR-BrainPET scanner and a whole head CT scan at the same day were used. The latter were transformed to CT-based attenuation maps (AMCT). The MR images were used to obtain AMs using the Boston-MGH method (AMMGH), the London-UCL method (AMUCL), the CT-template-based (AMCT-Juel) and Tx-template-based (AMTx-Juel) Juelich methods. BrainPET emission data were reconstructed with the five AMs. Using the SUIT tool the cerebellum was extracted from the MR images and normalized to a cerebellum VOI atlas. The cerebellum of the PET data was extracted by applying the same parameters. Correlation plots with regression equations, coefficients of determination R2, normalized and absolute normalized errors (NErr) between AMCT and the other four AMs were calculated. The values of cerebellar NErr varied to a high extent between the four AC methods. In one case AMMGH showed the lowest NErr (3.03±1.60%), in three cases the AMUCL performed best (NErr from -0.13±2.33% to 0.91±2.93%), in three case the AMTx-Juel had the best results (NErr from -1.24±3.01% to 0.86±2.07%), while for the remaining seven subjects the AMCT-Juel performed best (NErr from -2.08±0.94% to 7.98±3.43%). Our results demonstrate that the quantitation of radiotracer uptake in the cerebellum is very susceptible to the respective attenuation correction applied to the PET data. This, in turn, has to be considered in neuroreceptor modelling studies which rely on the cerebellum as reference.

Keywords: Attenuation Correction, Hybrid PET/MR
Poster panel
(face) ID: 206

Poster Number:

Tensor Tomography of Dark Field Scatter using X-ray Interferometry with Biprisms (#1265)

G. T. Gullberg1, M. Fuller2, U. Shrestha1, Y. Seo1

1 University of California San Francisco, Radiology and Biomedical Imaging, San Francisco, California, United States of America
2 TF Instruments, Salinas, California, United States of America


X-ray CT is the foundation of the medical imaging industry. However, medical imaging has not taken full advantage of its potential to measure tissue properties that are possible by measuring and modeling the full extent of the physics involved in the interaction of x-rays with soft tissue. To obtain measures of the full tissue properties, one needs imaging systems that can provide highly resolved x-rays with spatially-modulated intensity. It is known from basic optics that amplification of coherent photons can be obtained through interferometry. This is feasible even with a conventional x-ray tube by using a Talbot–Lau interferometer most often coupled with gratings for differential phase-contrast imaging (DPCI). The same could be achieved using biprisms with which there are no optics in the x-ray beam between the sample and the detector reducing loss of detected x-rays, radiation exposure, and optical alignment difficulties over grating-based interferometers. Our goal is to develop imaging systems based on biprism interferometry with improved polychromatic performance and to develop algorithms for tomographic reconstructions from projections of phase contrast data to retrieve images of absorption, differential phase, and dark field small-angle scatter. In this presentation, we focus on the reconstruction of small-angle scatter. At every image voxel, a fixed set of scatter vector directions is assumed for which the coefficients are estimated from the measured projections. A model of the imaging process using biprisms instead of gratings is developed for the reconstruction of this tensor of small angle scattering directions. A likelihood function is formulated assuming the detection of photons follows a Poisson distribution. The estimates of the tensor parameters are obtained by maximizing a likelihood function using the expectation maximization (EM) algorithm. Biprism-based interferometry has the potential to revolutionize phase-contrast and dark-field x-ray CT.

Keywords: interferometry, biprism, tensor tomography, dark field scatter
Poster panel
(face) ID: 209

Poster Number:

Evaluation of SRW-OSEM Using Clinical Data (#1343)

J. - C. (. Cheng1, 2, J. Matthews2, R. Boellaard3, 4, R. Laforest5, V. Sossi6

1 The University of British Columbia, Pacific Parkinson's Research Centre, Vancouver, British Columbia, Canada
2 The University of Manchester, Division of Informatics, Imaging, and Data Sciences, Manchester, United Kingdom of Great Britain and Northern Ireland
3 VU University Medical Centre, Radiology and Nuclear Medicine, Amsterdam, Netherlands
4 University Medical Centre Groningen, Nuclear Medicine and Molecular Imaging, Groningen, Netherlands
5 Washington University School of Medicine, Radiology, St. Louis, Missouri, United States of America
6 The University of British Columbia, Physics and Astronomy, Vancouver, British Columbia, Canada


We describe evaluations of the SRW-OSEM algorithm using clinical patient data. SRW-OSEM is an iterative reconstruction method which incorporates scatter and randoms corrections within the weighting component of the system matrix analogous to the attenuation weighted reconstruction algorithm. Our previous results obtained from small animal phantom data showed that SRW-OSEM can accelerate the reconstruction task and reduce the storage cost as well as improving the image quality. In this work, further evaluations were conducted using clinical patient data whose scatter fraction is much higher than the phantom data previously used. As a result, the trues fraction was ~30% for the patient data as compared to ~80% for the small animal phantom data. Convergence in contrast recovery and image profiles were compared between the SRW-OSEM and OP-OSEM. Higher improvement in convergence with respect to OP-OSEM was observed from SRW-OSEM for the patient data as compared to the improvement observed previously from the small animal phantom data. As expected, the higher the background contamination (e.g. scatter and randoms fractions) the higher the improvement in convergence achieved by SRW-OSEM with respect to OP-OSEM. In particular, 3-4 times faster convergence with respect to OP-OSEM was achieved by SRW-OSEM in this case; i.e. image reconstructed with 3 iterations of SRW-OSEM contains similar contrast as compared to that reconstructed with 12 iterations of OP-OSEM. Furthermore, lower noise was observed from the SRW-OSEM image as compared to the OP-OSEM image likely due to the lower number of iterations used in the reconstruction. 

Keywords: Scatter and Randoms Weighted, PET reconstruction, SRW-OSEM
Poster panel
(face) ID: 212

Poster Number:

Noise Analysis of Iterative Algorithm for Lower Iteration Number and Weighting Effects (#1387)

G. L. Zeng1, 2, R. L. Frazier1

1 Weber State University, Engineering, Ogden, Utah, United States of America
2 University of Utah, Radiology, Salt Lake City, Utah, United States of America


It has been observed by many researchers that low intensity regions of images reconstructed by iterative or analytical algorithms tend to have low noise and high intensity regions tend to have high noise. This phenomenon in fact can be predicted by a closed-form mathematical expression. However, when the number of iterations is small this observation may not hold, especially when some weighting functions are used.

Keywords: Noose Analysis, Tomography, Iterative Reconstruction, Analytic Reconstruction
Poster panel
(face) ID: 215

Poster Number:

Longitudinal Multi-Dataset PET Image Reconstruction (#1787)

S. Ellis1, A. J. Reader1

1 King's College London, Division of Imaging Sciences and Biomedical Engineering, London, United Kingdom of Great Britain and Northern Ireland


In positron emission tomography (PET), a subject may be scanned two or more times to monitor longitudinal functional changes. These images often appear very similar except for localised, relatively small regions of change. This observation led to the proposal of the maximum a posteriori simultaneous longitudinal reconstruction (MAP-SLR) method, which reconstructs longitudinal datasets together with regularisation to encourage sparse difference images between pairs of scans. In this work we extend MAP-SLR to application to a multi-scan treatment response simulation study. To do this, five 2D [18F]fluorodeoxyglucose head scan datasets (designed to emulate a brain tumour longitudinal study) were simulated and then reconstructed with MAP-SLR. The resulting images were compared to: maximum likelihood expectation-maximisation (MLEM) reconstructions; longitudinally smoothed MLEM reconstructions; and MLEM applied to a reference dataset with five times the number of counts. When using MAP-SLR, the noise (in terms of regional coefficient of variation) in a longitudinally unchanging white matter region was reduced and with sufficient regularisation these noise levels approached the high counts reference case. In the tumour, whilst a longitudinal bias is obtained with MAP-SLR, the bias is much smaller than that obtained when performing a noise-matched longitudinal smooth on MLEM reconstructions. With an appropriate level of regularisation the tumour bias is small enough to produce reconstructed images which preserve the longitudinal changes seen in the independent dataset MLEM reconstructions, but with noise reduction of 40% in regions which do not change. The results suggest that MAP-SLR is a simple and effective way of achieving noise reduction in longitudinal PET imaging. Future work will involve application specific testing and investigation into the inclusion of other longitudinally defined penalties into the simultaneous reconstruction.

Keywords: positron emission tomography, image reconstruction, longitudinal image reconstruction, regularised reconstruction, sparsity, treatment response
Poster panel
(face) ID: 218

Poster Number:

Statistical Image Reconstruction for Shortened Dynamic PET Using a Dual Kernel Method (#1936)

B. Spencer1, G. Wang1

1 UC Davis Medical Center, Radiology, Sacramento, California, United States of America


Dynamic F-18 FDG PET imaging along with tracer kinetic modeling can provide parametric images of physiologically important parameters for characterization of tumor and other diseases. This technique often requires a 1-hour long scanning time, which is less practical in clinic; a more practical method is to use a shortened dynamic scan time of thirty or forty minutes. However, a shortened dynamic scan acquires less data and tracer kinetic modeling becomes more sensitive to high noise in dynamic PET. To address the noise challenge, the kernel method has been developed for efficient dynamic PET image reconstruction. Previous kernel approaches use a single kernel type, which exploits either nonlocal or local spatial correlations from image priors but does no explore the full potential of the kernel framework. In this work, we propose a new dual-kernel approach to further enhance kernel-based dynamic PET image reconstruction. The dual kernel combines the existing non-local kernel with a local convolutional kernel that can be easily trained from image priors. We evaluated the new kernel approach for shortened dynamic FDG-PET imaging using a digital brain phantom. Simulation results have demonstrated that the dual-kernel approach can achieve better image quality than standard reconstruction approach and the single kernel approach.

Keywords: Dynamic PET, Kernel Method, Image Reconstruction, Kinetic Modeling
Poster panel
(face) ID: 221

Poster Number:

Enhancement of Time-of-Flight PET Image Reconstruction for Long-lived Positron Emitters Using Information from a Prompt Gamma Ray (#2176)

G. Chinn1, 2, C. S. Levin1, 2

1 Stanford, Dept. of Radiology, Stanford, California, United States of America
2 Stanford, MIPS, Stanford, California, United States of America


Isotopes such as I-124 emit a prompt gamma photon along with a positron leading to a triple coincidence of a 603 keV photon along with a pair of 511 keV annihilation photons. In conventional PET, this third photon degrades image quality and accuracy by creating multiple and random coincidences. Consequently, such prompt gamma emitting isotopes are not desirable.  However, using time-of-flight measurements from all three photons, the event can be more accurately positioned along the line of response between the two 511 keV photons compared to conventional TOF PET that uses the timing information from only the two 511 keV photons. In this work, we present a formulation for positioning positron-gamma TOF-PET events by employing weighted least squares estimation along with numerical simulation results of the positioning accuracy compared to conventional TOF-PET. Our numerical simulation shows that the detection of the third gamma can improve the event localization in positron-gamma (three photon coincidence) TOF-PET from 2% to 15% depending on the position of the detected prompt gamma ray relative to the detected annihilation photons.

Keywords: PET, Time-of-flight, Image Reconstruction
Poster panel
(face) ID: 224

Poster Number:

Gradient-Domain PET Reconstruction (#2256)

M. Magdics1, L. Szirmay-Kalos1, L. Neumann2, 3

1 Budapest University of Technology and Economics, Department of Control Engineering and Information Technology, Budapest, Hungary
2 University of Girona, Computer Vision and Robotics Institute (VICOROB), Girona, Spain
3 ICREA, Barcelona, Spain


In gradient domain reconstruction approaches we seek the partial derivatives of the unknown image instead of directly recovering the image itself, which is then computed from its gradient using a Poisson solver or convex optimization. The motivation behind this method is that the sparsity of the individual partial derivatives often exceeds the sparsity in the image domain or in the field of gradient magnitudes, which may lead to improved quality compared to traditional methods and the total variation regularization. Previous work has demonstrated this beneficial property in magnetic resonance imaging (MRI) and computed tomography (CT) applications. This work applies the gradient domain reconstruction methodology for iterative positron emission tomography (PET). We apply separate iterative PET reconstructions to recover the partial derivatives, expressed as finite differences, of the objective radio-tracer density. This, however, requires the knowledge of the projection data that would be obtained by measuring the derivative of the objective function. We show that we can approximate the projections of the partial derivatives by the partial derivatives of the measurement. As the derivative of the measurement is no longer of Poisson distribution, we use the iterative Landweber algorithm instead of the maximum likelihood expectation maximization (ML-EM) method. The final reconstruction is then obtained by integrating the reconstructed partial derivatives using a Poisson solver. We demonstrate the gradient domain approach on a hypothetical 2D PET scanner and show that results are comparable to that of the ML-EM reconstruction.

Keywords: Image reconstruction, Positron emission tomography, Gradient domain reconstruction, Poisson solver
Poster panel
(face) ID: 227

Poster Number:

Robust reconstruction for fluorescence molecular tomography based on correntropy matching pursuit (#2606)

S. Zhang1, X. Ma2, H. Meng2, S. Wei1, J. Tian2

1 Shandong University, Institute of Biomedical Engineering, School of Control Science and Engineering, Jinan, China
2 Chinese Academy of Sciences (CAS), Institute of Automation, Chinese Academy of Sciences, Beijing, China


Fluorescence molecular tomography (FMT), as a promising imaging modality in preclinical research, can obtain the three-dimensional location information of the specific tumor for small animal imaging. A lot of methods based on orthogonal matching pursuit (OMP) have been used for FMT reconstruction. OMP and most of its variants use the mean square error criterion to estimate the sparse vector which depends on the Gaussian assumption of the error distribution. However, this method has poor performance on reconstruction with non-Gaussian noise. In this study, we propose correntropy matching pursuit (CMP) method to alleviate this problem of OMP. Unlike some other matching pursuit methods, CMP method is independent of the error distribution. This method can assign small weights on severely corrupted entries of data and large weights on clean ones adaptively, thus reducing the effect of large noise. The numerical simulation studies, adding different types of noise including Poisson, Exponential and Gaussian noise on the measured data, were used to evaluate the performance of the proposed method. The position errors of the OMP, regularized OMP (ROMP) and CMP were 2.65mm, 2.54mm, 1.13mm; 2.36mm, 2.47mm, 0.98mm; 0.93mm, 1.08mm, 1.04mm with 20% Poisson, Exponential and Gaussian noise respectively. Compared with OMP and ROMP, CMP method shows better performance on FMT reconstruction with non-Gaussian noise and similar performance with Gaussian noise.

Keywords: matching pursuit (MP), correntropy matching pursuit (CMP), Robust, FMT reconstruction
Poster panel
(face) ID: 230

Poster Number:

Iterative, direct LOR PET Image Reconstruction of Human Brain Data for the Siemens mMR Biograph (#2748)

J. J. Scheins1, J. Baran2, C. Lerche1, N. J. Shah1, 2, Z. Chen2, G. Egan2

1 Forschungszentrum Jülich GmbH, Institute of Neuroscience and Medicine (INM-4), Jülich, North Rhine-Westphalia, Germany
2 Monash University, Monash Biomedical Imaging, Melbourne, Australia


In this paper we present first results of an implementation of an iterative, direct LOR PET image reconstruction for the Siemens mMR Biograph using OP-OSEM. The implementation is based on a speed-optimised version of the PET Reconstruction Software Toolkit (PRESTO) to obtain moderate reconstruction time on a single computer. For optimal Single Instruction Multiple Data (SIMD) support projection data and image data are stored in rows of 56x4=224 consecutively words in memory according to the symmetry properties of the scanner which consists of 56 detector heads in the tomographic (transaxial) plane. In this way, PRESTO provides individual, geometrically matching forward-backward projectors for any physical scanner LOR and axial/transaxial LOR compression can be fully omitted. The total system matrix size is 15 GB. PRESTO reconstructions are performed on an Intel Xeon E5-2680v2@2.8GHz (2x10 cores) with 128 GB RAM using up to 20 simultaneous threads and ordered subsets with 4 subsets. One subset iteration update needs approx. 3 sec, thus typical images (require  approx. 100 subset iteration updates) can be obtained in less than 5 minutes on a single machine without any GPU support. Reconstructed images for Iida phantom data and FDG human brain data are compared to images of the Siemens mMR Biograph PSF reconstruction which is optionally available in the Siemens e7tools using Span 11 axial compression. Images obtained by the LOR-PRESTO reconstruction provide better image quality in terms of resolution, e.g. Iida phantom features are more adequately represented. In contrast, images of the Siemens PSF reconstruction suffer from visible Gibbs artifacts whereas for the direct LOR reconstruction no evident Gibbs artifacts are observed. For this reason, the direct LOR reconstruction can help to improve quantification accuracy due to the reduction of the bias evoked by applying physical PSF kernels.

Keywords: Iterative Image Reconstruction, PET, Biograph mMR, Direct LOR Reconstruction
Poster panel
(face) ID: 233

Poster Number:

A direct ray tracing reconstruction algorithm using an adaptive median filter (#2848)

S. Sánchez1, A. Iborra2, P. Conde1, A. J. González1, J. M. Álvarez-Gómez1, A. Soriano1, D. Grau-Ruiz1, A. González-Montoro1, A. Aguilar1, G. Cañizares1, E. Lamprou1, L. Moliner1, F. Sánchez1, M. J. Rodríguez-Álvarez1, J. M. Benlloch1

1 Universidad Politecnica de Valencia/CSIC, Institute for Instrumentation in Molecular Imaging (I3M), Valencia, Spain
2 Laboratory of Medical Information Processing (LaTIM - INSERM UMR 1101), Brest, France


A direct ray tracing algorithm has been proposed under the development of a dedicated brain PET insert scanner, based on monolithic scintillation crystals. This algorithm avoids the use of histograms or system matrices, as in the case of the Filtered Back Projection (FBP) or the Maximum Likelihood Expectation Maximization (MLEM) algorithms. Moreover, this method follows naturally from the concept of Line Of Response (LOR), and easily can account for the photon depth of interaction or annihilation photons time of flight (TOF) information.

The algorithm requires a filtering process that has been carried out in two steps. First, an adaptive median filter with a window array of 3×3×3 is proposed as a suitable method to cancel the discontinuities produced by a fine pixilation of the image space. Second a Butterworth filter is applied on the reconstructed image to wipe out the low oversampled frequencies. The performance of the proposed algorithm has been evaluated with simulated and real data, and compared with some of the gold-standard algorithms, such as FBP or List Mode Ordered Subset (LMOS). In the case of the acquired data, a Derenzo-like phantom was used showing the 1.6 mm rods, with a measured peak-to-valley of 2, in contrast to 1.25 and 1.23 for FBP and LMOS, respectively. In addition to the already mentioned advantage, this algorithm provides the option of real time reconstruction.

Keywords: Direc Ray Tracing, Filtered Back Projection, Maximum Likelihood Expectation Maximization, Line Of Response, adaptive median filter, Butterworth filter
Poster panel
(face) ID: 236

Poster Number:

Image reconstruction by nonconvex inexact half-quadratic optimization (#3068)

M. Robini1, P. Niu1, F. Yang2, Y. Zhu1

1 CREATIS, INSA Lyon, Villeurbanne, France
2 Beijing Jiaotong University, School of Computer and Information Technology, Beijing, China


We consider the problem of reconstructing an image from noisy projection data. The problem is formulated as the minimization of a generic objective function that can be particularized to the standard objectives used in medical image reconstruction (the most common choice is the sum of a convex data-fidelity term and a gradient-based regularization term). Our main focus is on half-quadratic (HQ) optimization. HQ optimization lies at the intersection of different classes of algorithms (fixed-point, quasi-Newton, alternating minimization, and majorization-minimization); it comes down to solving a sequence of positive definite systems, or inner systems, and has the advantages of simplicity and versatility. HQ algorithms have been meticulously studied, both theoretically and experimentally, but two difficulties remain: first, the inner systems are solved approximately in practice, which may hamper convergence; second, pointwise convergence is not guaranteed when the set of stationary points of the objective contains a continuum. We present new results that do not suffer from these limitations, and we propose an efficient and numerically stable implementation based on conjugate gradient. The resulting inexact HQ algorithm sets a threshold on the accuracy of the computed solutions to the inner systems; it converges globally to a stationary point of the objective under a minimal set of conditions ubiquitous in tomographic reconstruction. We investigate the behavior of the inexact HQ algorithm in the context of limited-data tomography. Our experiments show that it performs well in various nonconvex scenarios (including when the regularization term has flat continua) and converges to solutions accurate to full machine precision.

Keywords: Image reconstruction, nonconvex optimization, half-quadratic optimization, conjugate gradient
Poster panel
(face) ID: 239

Poster Number:

System Modeling and Image Reconstruction for 3D Micron-resolution Autoradiography (#3193)

M. P. Nguyen1, W. Ensing2, 3, B. Vastenhouw2, 3, C. Kamphuis2, 3, R. Ramakers1, 2, M. C. Goorden1, F. J. Beekman1, 2

1 Delft University of Technology, Section of Radiation Detection and Medical Imaging, Delft, Netherlands
2 MILabs B.V., Utrecht, Netherlands
3 University Medical Center Utrecht, Department for Translational Neuroscience, Brain Center Rudolf Magnus, Utrecht, Netherlands


Traditional autoradiography requires tissue cryo-sectioning by a cryo-microtome, followed by scanning of the sections with a digital 2D film or phosphor screen reader. Reconstructing 3D images from a stack of 2D images is extremely difficult and often impossible. Recently, MILabs BV introduced EXIRAD-3D, automated 3D autoradiography that eliminated those time-consuming and error prone steps. EXIRAD-3D also enables 3D multi-isotope  imaging that was very hard to achieve  with  traditional autoradiography methods. It is based on a specially designed collimator that can be inserted into the U-SPECT-II system to allow for SPECT imaging of radiolabeled molecules down to 0.15 mm resolution. This work evaluated two different methods that were employed to obtain a system matrix for image reconstruction in EXIRAD-3D. A first method, referred to as “traditional” method, used model-based interpolation to generalize measured point spread functions (PSFs) over the whole field-of-view. A second method, referred to as “raytracing” method, obtained some geometrical information from measured PSFs and then generated a full system matrix with a raytracing code. Reconstruction results were evaluated on a Jaszczak phantom scan with 99mTc (rod diameters of 0.15, 0.17, 0.19, 0.22, 0.26, and 0.30 mm). The smallest 0.15 mm rods could be distinguished when the traditional method was used. The raytracing method was only able to resolve the rods > 0.17 mm diameter and some rod distortion was visible which was probably caused by non-accurate information on the collimator geometry and detector response.

Keywords: 3D autoradiography, multi-isotope, SPECT, micron-resolution, image reconstruction, system modeling
Poster panel
(face) ID: 242

Poster Number:

Joint Reconstruction of Parametric PET and MR Images (#3323)

Y. Yu1, 2, T. Ma1, 2, S. Wang1, 2, Y. Liu1, 2, S. Yao3

1 Tsinghua University, Department of Engineering Physics, Beijing, China
2 Ministry of Education, Key Laboratory of Particle & Radiation Imaging(Tsinghua University), Beijing, China
3 Chinese PLA General Hospital, Department of Nuclear Medicine, Beijing, China


Both the static and dynamic parametric PET images have strong correlations with anatomical MR images for many commonly used tracers. MAP reconstruction can improve the quality of parametric PET images based on the similarity between the parametric PET and higher-resolution MR images within the direct 4D image reconstruction framework.

With the application of current PET/MR scanner, PET and MR acquisitions can be performed at the exact same time duration and bed position, which results in more reliable joint information of two modalities with neither temporal nor spatial mismatch. Utilizing simultaneous functional and anatomic information acquired from PET/MR scanner, PET/MR joint reconstruction can improve the images quality of both modalities by exploring the structure similarities of them.

We proposed a novel framework to directly reconstruct both the PET parametric images based on Patlak linear model and MR images by minimized one single cost function. The similarity term between the Patlak slope and the MR image was defined with three different priors: joint entropy (JE), joint total variation (JTV) and quadratic parallel level sets (QPLS). The functional was minimized with the same quasi-Newton optimized method. The algorithm was validated by simulation of a 60-minute 18F-FDG PET dynamic scan and a MR scan of one Zubal brain phantom. PET/MR images without any priors (NP) were also generated for comparison.

Parametric images obtained by joint reconstruction have less overall reconstruction noise than that of NP, as MR based anatomic information has been introduced by all three priors. Sharpness and visibility of fine features were also clearly improved. Joint reconstruction also helped to obtain more accurate parameter estimation especially within the tumor. JE prior gave the smallest bias and variances for all regions.

In conclusion, joint reconstruction of parametric PET and MR images substantially improved resolution and accuracy of PET parametric images.

Keywords: Joint Reconstruction; PET/MR; Parametric Imaging;
Poster panel
(face) ID: 245

Poster Number:

GPU-based List-mode Direct Parametric Reconstruction for Dynamic Cardiac SPECT (#3594)

L. Shi1, J. Wu1, Y. Lu1, J. - D. Gallezot1, S. Thorn1, A. J. Sinusas1, R. E. Carson1, C. Liu1

1 Yale University, New Haven, Connecticut, United States of America


Introduction: Recently introduced stationary dedicated cardiac SPECT scanners provides new opportunities to quantify myocardial blood flow (MBF) using dynamic SPECT.  However, comparing to PET, the low sensitivity of SPECT scanners affects MBF quantification due to the high noise level, especially for 201Tl which has a long half-life (73 hours). To reduce image noise and improve MBF quantification accuracy, we developed a GPU-based list-mode direct reconstruction method that produces voxel-by-voxel K1 images directly from list-mode SPECT data for the GE Alcyone SPECT with 19 pinhole collimators and CZT detectors.

Methods: A GPU-based MLEM algorithm was developed for list-mode direct parametric reconstruction with a one-tissue compartment model (1T). The proposed direct method was compared to the indirect method in both simulated data and in-vivo porcine data with 201Tl tracer. Myocardium (MYO) ROI mean values and coefficients of variation were evaluated.

Results: Compared to CPU implementation, our GPU-based reconstruction achieved over 2000-fold acceleration. Our list-mode recon­struction generated nearly identical results to the projection-based MLEM method on a static SPECT dataset. For dynamic imaging, in the MYO-only simulation, direct parametric reconstruction achieved lower image noise and faster convergence as compared to the indirect method. In the MYO+blood pool (BP) simulation and in the in-vivo porcine study, incorporating the blood volume (BV) term in the model substantially improved the fitting of both MYO and BP.

Conclusions: Compared to the indirect method, the direct reconstruction method achieved lower image noise and faster convergence. The BV term is critical for dynamic cardiac studies and will be incorporated into our direct reconstruction in the future.

Keywords: direct parametric reconstruction, cardiac SPECT, GPU acceleration
Poster panel
(face) ID: 248

Poster Number:

Bayesian Reconstruction Using Region of Interest Markov Chain Monte Carlo Simulation (#3816)

X. Lai1, K. Kim1, G. El Fakhri1, Q. Li1

1 Massachusetts General Hospital, Radiology Department, Boston, Massachusetts, United States of America


Recently there are growing interestes in fully assessing likelihood or posterior distribution based on Markov Chain Monte Carlo (MCMC) sampling. Such fully sampling not only allows us to get a point estimator that ML or MAP provide but also enables us to assess the uncertainty of the esimator and carry Bayesian tests. However, the computing complexity and heavy calculation load hinder its application in high-dimensional image reconstruction problems. In this paper, we will present region of interest MCMC sampling for Bayesian reconstruction. It could remarkably reduce computing load compared to full-space MCMC sampling, while allowing us to assess the postior in the ROI.

Keywords: MCMC, Bayesian Reconstruction
Poster panel
(face) ID: 251

Poster Number:

Cerebral perfusion CT reconstruction via infimal convolution of total generalized variation regularization (#3881)

Y. Liao1, 2, S. Li1, 2, J. He1, 2, Y. Wang1, 2, D. Zeng1, 2, Z. Bian1, 2, J. Huang1, 2, J. Ma1, 2

1 Southern Medical University, School of Biomedical Engineering, Guangzhou, China
2 Southern Medical University, Guangzhou Key Laboratory of Medical Radiation Imaging and Detection Technology, Guangzhou, China


Ischemic stroke is the leading cause of serious and long-term disability worldwide. The cerebral perfusion CT (CPCT) can be used to diagnose the ischemic stroke for which can provide cerebral hemodynamic information. However, due to the dynamic sequential scans in CPCT, the associative radiation dose unavoidably increases compared with conventional CT. To address this issue, in this study, we propose a CPCT reconstruction method at low-dose cases wherein a recently proposed infimal convolution of total generalized variation (ICTGV) regularization is introduced into the reconstruction process. For simplicity, the proposed algorithms is terms as “PWLS-ICTGV”. Specifically, The ICTGV allows automatic decomposition of the CPCT data into two components with different degrees of temporal regularity, each of component constrained by a second-order tensor generalized variation regulation (TGV), which could achieve regularization via optimal local balancing between spatial and temporal regularity. Moreover, an alternating minimization algorithm is employed to solve the proposed method. Experimental results on the phantom data show that the proposed method can achieve promising gains over the filtered back-projection (FBP) and tensor total variation (TTV) based algorithm in terms of noise reduction and parameters estimation of hemodynamic information.

Keywords: Cerebral perfusion, CT reconstruction, total generalized variation
Poster panel
(face) ID: 254

Poster Number:

Spatially-variant Strength for Anatomical Priors in PET Reconstruction (#4073)

Y. - J. Tsai1, G. Schramm2, J. Nuyts2, A. Bousse1, S. Ahn4, C. W. Stearns3, S. Rose3, S. Arridge5, K. Thielemans1

1 University College London, Institute of Nuclear Medicine, London, United Kingdom of Great Britain and Northern Ireland
2 University of Leuven, Department of Imaging & Pathology, Leuven, Belgium
3 GE Healthcare, Waukesha, Wisconsin, United States of America
4 GE Global Research, Niskayuna, New York, United States of America
5 University College London, Department of Computer Science, London, United Kingdom of Great Britain and Northern Ireland


In emission tomography penalized image reconstruction methods can be used to control noise. The use of anatomical information allows incorporation of prior information on sharp edges into the regularization. This study explores the use of a spatially-variant penalty strength, proposed initially for quadratic penalties, in penalized image reconstruction using anatomical information. The expected benefit is to obtain local contrast which is independent of changes to the surrounding activity. This should reduce the need for tuning of the global penalty strength.

We have used the recently proposed Parallel Level Sets anatomical prior as it has shown promising results in the literature. It was incorporated into the previously proposed preconditioned algorithm (L-BFGS-B-PC) for achieving both good image quality and fast convergence rate. To demonstrate how the surrounding activity can affect the visual appearance and the quantitative accuracy of a region of interest, simulations were performed on simple phantoms including a 2D disc phantom and XCAT images. Lesions were inserted and surrounded by either high or low background. Anatomical information was provided by the attenuation maps, with different cases simulated with lesion information in the attenuation to be able to evaluate the influence from the anatomical penalty. The reconstructed images and the quantitative results for both phantoms with and without applying the spatially-variant penalty strength were compared. Based on the results for the disc phantom, lesion delineation and quantitative accuracy is more consistent (i.e. independent of background activity) when the spatially-variant penalization is applied.  This is the case both with and without anatomical information. Although the improvement is less apparent for the XCAT phantom in our 2D simulations, further work to explore the potential benefit in fully 3D imaging is warranted.

Keywords: penalized reconstruction, L-BFGS-B, Parallel Level Sets, uniform resolution
Poster panel
(face) ID: 257

Poster Number:

A method for detection of microaneurysms in fundus images (#1095)

R. Rosas-Romero1

1 Universidad de las Américas-Puebla, Electrical & Computer Engineering, San Andrés Cholula, Puebla, Mexico


Diabetes increases the risk of developing any deterioration in the blood vessels that supply the retina, an ailment known as Diabetic Retinopathy (DR). It can only be diagnosed by an ophthalmologist; however, the growth of the number of ophthalmologists is lower than the growth of the population with diabetes. Preliminary, affordable and accessible ophthalmological diagnosis will give the opportunity to perform routine preventive examinations, indicating the need to consult an ophthalmologist during a stage of non proliferation. During this stage, there is a lesion on the retina, known as microaneurysm (MA), which is one of the first clinically observable lesions that indicate the disease. This work proposes a new approach for MA detection based on (1) reduction of non-uniform illumination; (2) normalization of image grayscale content to improve dependence of images from different contexts; (3) application of the bottom-hat transform to leave reddish regions intact while suppressing bright objects; (4) binarization of the image of interest with the result that objects corresponding to MAs, blood vessels, and other reddish objects (Regions of Interest - ROIs) are completely separated from the background; (5) application of the hit-or-miss Transformation on the binary image to remove blood vessels from the ROIs; (6) two features are extracted from a candidate to distinguish real MAs from FPs, where one feature discriminates round shaped candidates (MAs) from elongated shaped ones (vessels) through application of Principal Component Analysis (PCA); the second feature is a count of the number of times that the radon transform of the candidate ROI, evaluated at the set of discrete angle values, is characterized by a valley between two peaks. The proposed approach is tested on the public databases DiaretDB1. The proposed MA detection method achieves sensitivity, specificity and precision of 92.32%, 93.87% and 95.93%.

Keywords: Morphological Processing, Eye Fundus Image, Principal Component Analysis, Radon Transform
Poster panel
(face) ID: 260

Poster Number:

A MR Guided De-noising for PET Using IHYPR-LR (#1341)

J. - C. (. Cheng1, 2, J. Matthews2, R. Boellaard3, 4, V. Sossi5

1 The University of British Columbia, Pacific Parkinson's Research Centre, Vancouver, British Columbia, Canada
2 The University of Manchester, Division of Informatics, Imaging, and Data Sciences, Manchester, United Kingdom of Great Britain and Northern Ireland
3 VU University Medical Centre, Radiology and Nuclear Medicine, Amsterdam, Netherlands
4 University Medical Centre Groningen, Nuclear Medicine and Molecular Imaging, Groningen, Netherlands
5 The University of British Columbia, Physics and Astronomy, Vancouver, British Columbia, Canada


We describe a MR guided post processing method to de-noise PET image based on Iterative HighlY constrained back-PRojection Local Reconstruction (IHYPR-LR). IHYPR-LR is a modified version of HYPR-LR with the composite image updated iteratively, and HYPR-LR is a de-nosing method originally developed for time-resolved MRI. In this work, a co-registered T1-weighted MR image with high resolution and low noise was used as the initialization of the composite image in the IHYPR-LR frame work for PET de-noising. A [11C]DASB Parkinsonian patient study conducted on the High Resolution Research Tomograph (HRRT) was used for the evaluations of the proposed method. The study was divided into high and low count frames with similar tracer distribution. The high count data were used to extract the optimal number of IHYPR iterations which minimizes the bias introduced by the MR composite without excessively degrading the level of noise reduction. The optimal number of iterations was then applied to low count PET data. The de-noised images were generated using the original HYPR-LR and IHYPR-LR and compared with the OSEM images with a standard 2mm FWHM Gaussian post filter for the HRRT. As expected, since MR images do not always share the same contrast and structures with PET images, bias in contrast was observed from the de-noised PET image using the original HYPR-LR. On the other hand, 3 iterations of IHYPR-LR successfully reduced the bias and outperformed the post filtered PET image in terms of noise reduction and structure boundary definitions. In summary, the proposed MR guided de-noising method achieves noise reduction without degrading the PET contrast. Further optimizations and investigations will be conducted.

Keywords: MR guided de-noising, PET, post-processing
Poster panel
(face) ID: 263

Poster Number:

Multi-View Depth-Aware Rigid 2-D/3-D Registration (#1752)

R. Schaffert1, 2, J. Wang2, P. Fischer2, A. Borsdorf2, A. Maier1, 3

1 Friedrich-Alexander University Erlangen-Nürnberg, Pattern Recognition Lab, Erlangen, Bavaria, Germany
2 Siemens Healthcare GmbH, Forchheim, Bavaria, Germany
3 Erlangen Graduate School in Advanced Optical Technologies (SAOT), Erlangen, Bavaria, Germany


Overlays of CT images onto X-ray images are used in minimally invasive procedures to provide assistance to the physicians. 2-D/3-D image registration is usually required to achieve the high accuracy needed for the overlay. Registration using single-view images is often not sufficient regarding accuracy and robustness, especially for a limited field of view. Therefore, multi-view registration is required, where the typical setup is to have 90° between the views. However, this can lead to high obstruction during a workflow and may be not necessary to obtain optimal robustness and accuracy. Recently, a depth-aware rigid 2-D/3-D registration framework was proposed and shown to achieve high accuracy and robustness in the single-view scenario, but it cannot be directly used for multi-view registration. In this paper, we extend this method to multiple views by reformulating the used point-to-plane correspondence model to make it independent of the view. This allows for simultaneous optimization over multiple views. Furthermore, we systematically investigate the effect of the viewing angle and the angle between the views in order to achieve optimal registration results while minimizing the obstruction of the workflow. We perform experiments on a publicly available “gold standard” angiography dataset as well as on clinical spine data. For the native 2-D images of the angiography dataset, our method achieves a capture range of 17 mm and an accuracy of 0.21 mm, outperforming the state-of-the-art method (capture range of 12 mm and accuracy of 0.23 mm). On the spine data, our method shows highly increased robustness (capture range of up to 25 mm) compared to single-view registration (capture range of up to 6 mm). We demonstrate that an angular distance of 15° between views increases the robustness considerably compared to single-view registration and the robustness reaches a plateau at 30° for the used data. Our evaluation also shows a dependency on the actual viewing direction.

Keywords: 2D/3D registration, multi-view, image fusion, interventional radiology
Poster panel
(face) ID: 266

Poster Number:

Individual prediction of brain tumor histological grading using radiomics on structural MRI (#1853)

S. Bonte1, 2, I. Goethals2, R. Van Holen1

1 Ghent University, Medical Image and Signal Processing (MEDISIP) - Department of Electronics and Information Systems - IMEC, Gent, Belgium
2 Ghent University Hospital, Department of Nuclear Medicine, Gent, Belgium


The accurate diagnosis of brain tumors is of primary importance for optimal therapy planning. In clinical practice, this is determined on a biopsy, exposing the patient to the risk of complications. Moreover, sampling bias and performer variability may influence the obtained result. Several studies have shown efforts to determine the histological grading of brain tumors in a non-invasive way by extracting features from medical images. A multicenter study where both tumor grade and cell type are simultaneously predicted is however lacking. In this study we collected structural MRI-scans from 327 patients with glioma acquired in different centers (the local hospital and two online databases). The goal was to predict tumor grade and cell type of individual patients using a radiomics study with Random Forests. In a multiclass design, we obtain a global accuracy of 69.0% to predict tumor grade and 45.6% to predict cell type. Converting the problem to binary classification, we obtain an accuracy of 85.1% to distinguish between low grade (grade 2 and 3) and high grade (grade 4), and 84.5% to distinguish between low-grade glioma and glioblastoma. A high degree of diagnosis variability and overlap between different low-grade classes might cause the reduced prediction accuracy. Our results however show that radiomics on structural MRI is a suitable approach for non-invasively assessing brain tumor diagnosis and might be used for individual treatment planning.

Keywords: Brain tumor, glioma, MRI, radiomics, random forests
Poster panel
(face) ID: 269

Poster Number:

Surface Registration to Estimate Motion in CBCT (#1893)

B. Bier1, M. Unberath1, N. Ravikumar1, 2, J. Maier1, A. Gooya2, Z. A. Taylor2, A. F. Frangi2, G. Gold3, R. Fahrig3, 4, A. Maier1

1 Friedrich-Alexander University Erlangen-Nürnberg, Pattern Recognition Lab, Erlangen, Germany
2 The University of Sheffield, CISTIB Center for Computational Imaging and Simulation Technologies, Sheffield, United Kingdom of Great Britain and Northern Ireland
3 Stanford University, School of Medicine, Stanford, United States of America
4 Siemens Healthcare GmbH, Erlangen, Germany


C-arm cone-beam CT (CBCT) has been used recently to acquire images of a patient's knee under weight-bearing conditions. The resulting reconstructions allow to assess knee joint health under more realistic conditions than in supine configuration. However, motion during the scan severely corrupts image quality of the reconstructions. Recently, a method to compensate for this motion has been proposed that, in contrast to its predecessors, uses a depth camera in addition to the CBCT. The method is based on the popular Iterative Closest Point (ICP) algorithm, that requires the selection of a reference and, therefore, induces bias.

In this work, we investigate a group-wise registration approach to estimate motion, which is more robust to noise and outliers. Further, the group-wise character can estimate the mean shape more accurately. We compare our results to a state-of-the-art marker-based method, as well as to the previously proposed ICP-based method.

Image quality improves compared to the ICP-based method with an improvement of the Structural Similarity (SSIM) from 0.96 to 0.98. Streaks in the images could be reduced slightly.

The preliminary results presented here are promising. Dense surface information together with the stochastic formulation of the proposed method allows for the incorporation of more complex, e.g. compound motion, that better reflects true joint motion. Further investigations in this direction are subject to future work.

Keywords: Cone-beam CT, Motion Estimation, Group-wise Registration, Range Imaging, Reconstruction, Weight-bearing Imaging
Poster panel
(face) ID: 272

Poster Number:

Motion Estimation in Rotational Angiography with Alpha-Expansion Moves (#2506)

L. Felsner1, A. Aichert1, M. Unberath1, A. Maier1

1 Friedrich-Alexander University Erlangen-Nürnberg, Pattern Recognition Lab., Computer Science Department, Erlangen, Germany


Rotational coronary angiography allows for 3D reconstruction in interventional guidance of percutaneous coronary interventions. While gated reconstruction is able to account for cardiac motion, residual breathing motion must be compensated for. Epipolar consistency has been shown to accurately extract the craniocaudal displacement of the heart during contraction and breathing. However, optimization of consistency over detector shifts proves complicated as recurrent motion of similar structures produces multiple local minima in the objective function. This prevents the use of typical local non-linear optimization methods. Related work, therefore, relies on grid-search, which is prohibitively expensive in high-dimensional parameter space. This, each projection is optimized individually, making existing methods slow and prone to unnatural jumps in the estimated motion. Therefore, we propose a new method for motion estimation using a graph-based approach. We reformulate the objective function in terms of a labeling problem and use alpha-expansion moves to solve it. For each label, representing a specific shift, a minimum cut of a graph represents the optimal assignment of that label to each projection versus its current estimate. Our approach is aware of the local neighborhood, which makes it more efficient to optimize, as well as more robust and elegant. It is also preferable to optimize all projections at once, in contrast to the order-dependent optimization with grid-search. The robustness and reliability of the optimization is evaluated on two phantom datasets with different heart and breathing motion. Our method is more stable with respect to the selection of parameter range and sequence of optimization. Using a graph based method we further gain some flexibility in how we choose the neighborhood term. In addition, the full optimization with grid-search takes 140 minutes, while alpha–expansion merely requires 6 minutes.

Keywords: Epipolar Consistency, Graph Cuts, Alpha-Expansion Moves, Motion Estimation, Rotational Angiography
Poster panel
(face) ID: 275

Poster Number:

Clinical Relevance of Partial-Volume Effect: Dependence on Lesion size and Shape (#2852)

T. Nguyen1, H. Zaidi1, 2, P. F. Høilund-Carlsen1, W. Vach3

1 Odense University Hospital, Dept. of Nuclear Medicine, Odense, Denmark
2 Div. of Nuclear Medicine and Molecular Imaging, Geneva University Hospital, Geneva, Switzerland
3 Inst. of Medical Biometry and Statistics, University of Freiburg, Freiburg, Germany


This study sought to systematically assess the influence of partial-volume effect (PVE) and segmentation on PET quantitative measures, hinging on lesion geometry. How this affects, e.g., prevalent maximum standardised uptake values, SUVmax, and potentially impact clinical applications, e.g., response evaluation has yet to be fully discerned. From PET simulations with open-source software of variable-sized ellipsoidal lesions inserted in an anthropomorphic phantom, images of two contrasts and resolutions were generated. SUVmax and volumetric indices extracted with six different segmentations were compared for variability and test-retest repeatability. The study showed similar or larger shape dependent variability and lower repeatability in SUVmax than SUVmean. Alternative volumetric indices might provide more robust measures but require better contouring than common thresholding. Thus findings suggested significant impact of PVE and segmentation in relevant clinical lesion sizes that can bias interpretation of SUV changes.

Keywords: PET quantitation, Partial-volume effects, PET segmentation, Standardized uptake values (SUV), Treatment assessment
Poster panel
(face) ID: 278

Poster Number:

Estimation of Architectural Distortion in Mammograms using Fractal Features (#2955)

M. K. Bhowmik1, A. Roy1, U. R. Gogoi1, N. Nath2

1 Tripura University, Department of Computer Science and Engineering, Suryamaninagar, Tripura, India
2 New York Institute of Technology, Department of Life Sciences, New York, New York, United States of America


Moving towards accurate breast cancer detection, X-ray mammography is the gold standard in medical imaging for its efficiency and reliability. Abnormalities often encountered in mammograms are in the form of benign or malignant masses, calcifications, asymmetry and architectural distortion. Other than masses and calcifications, architectural distortion should not be overlooked since, it is often the major cause of non-palpable cancer. However, due to the appearance variability and subtle differences of the abnormalities from the tissues, the radiologists feel ambiguous to detect and differentiate the malignant one from the benign one. Due to the existence of irregular and ill-defined structure in architecturally distorted areas, fractal features namely fractal dimension and lacunarity are considered in our work to differentiate the malignant architectural distortion from the benign architectural distortion. Our work can provide a second opinion to the radiologists in decision making. The performance of the proposed system has been evaluated with a dataset of total 19 mammograms with architectural distortions from the mini-MIAS database. The Mann Whitney Wilcoxon nonparametric test shows the statistical significance of fractal features in differentiating the abnormal mammograms from the benign ones. Based on the experimental results, it has been found that the combination of fractal dimension and lacunarity feature gives a prediction accuracy of 90%.

Keywords: Breast cancer, Mammograms, Fractal dimension, Lacunarity, Architectural Distortion
Poster panel
(face) ID: 281

Poster Number:

Texture Feature Analysis of Neighboring Colon Wall for Colorectal Polyp Classification (#3657)

M. J. Pomeroy1, 2, A. Abbasi2, K. Baker2, J. Liang2

1 Stony Brook University (SUNY), Biomedical Engineering, Stony Brook, New York, United States of America
2 Stony Brook University (SUNY), Radiology, Stony Brook, New York, United States of America


Colorectal cancer (CRC) remains one of the leading causes of cancer deaths today. Since precancerous colorectal polyps slowly progress into cancer, screening methods are highly effective in reducing the overall mortality rate of CRC by removing them before developing into later stages. Virtual colonoscopy has been shown to be a practical screening method and provide a high sensitivity and specificity for diagnosis between hyperplastic polyps and precancerous adenomas or adenocarcinomas through the use of texture feature analysis. We hypothesize that effects from non-hyperplastic polyps, such as angiogenesis from adenocarcinomas, may result in changes to the texture of the colon wall that could help with computer aided diagnosis of the colorectal polyps. Here we present the preliminary results of incorporating the texture features of neighboring colon wall tissue into the diagnostic classification. We use gray level co-occurrence matrices to calculate the established Haralick features and supplemental features for colorectal polyp regions of interest, as well as for the neighboring colon wall environment of the polyp. A random forest package was then used to perform the classification tests on different sets of features, with and without the inclusion of the environment to obtain an area under the curve (AUC) value of the receiver operating characteristic (ROC). The preliminary tests showed noticeable gain, and further investigation is needed.

Keywords: Colorectal cancer, Texture feature analysis, colon wall, Random forest, Receiver operating characteristic
Poster panel
(face) ID: 284

Poster Number:

Enhancement of mammographic images via cascaded multi-scale deep convolutional neural networks (#3930)

J. He1, 2, Y. Wang1, 2, S. Li1, 2, D. Zeng1, 2, W. Lyu1, 2, Z. Bian1, 2, J. Huang1, 2, J. Ma1, 2

1 Southern Medical University, School of Biomedical Engineering, Guangzhou, China
2 Southern Medical University, Guangzhou Key Laboratory of Medical Radiation Imaging and Detection Technology, Guangzhou, China


Mammography is the golden standard in the diagnosis of breast cancer. However, original mammographic images can be very difficult to interpret because of the low-contrast nature. Enhancement of the original mammographic images is indispensable when using mammography as a screening tool to diagnose the breast abnormities. In this paper, we present an effective deep convolutional neural network (CNN)-based method for the enhancement of mammographic images using large number of mammographic images as training data. The enhancement network is denoted as EhNet, which learned the direct mapping between the original mammographic images and the corresponding enhanced counterparts in the spatial domain. We adopted a cascade architecture of EhNets (called CaEhNet) to further improve the enhancement performance in a multi-scale way. In CaEhNet, the output of the EhNet in the coarser scale is used as part of inputs for the EhNet in the finer scale, which helps produce finer mammographic images. Clinical dataset was used to evaluate our method. The results demonstrate that our method can produce high-quality and high-resolution enhanced mammographic images.

Keywords: Mammography, deep convolutional neural networks, image enhancement, multi-scale
Poster panel
(face) ID: 287

Poster Number:

Dependency of tumor functional sphericity on PET delineation method (#4063)

M. Hatt1, B. Laurent1, H. Fayad1, V. Jaouen1, D. Visvikis1

1 INSERM, UMR 1101, LaTIM, Brest, France


Introduction: Sphericity has been recently highlighted as a new metric that could complement volume and SUV measurements to characterize functional tumor volumes from PET images. It was shown to have prognostic value in both head and neck and lung cancer. However its dependency on the tumor delineation method has not be thoroughly investigated yet. Such was the goal of the present study.

Materials and methods: A dataset of 168 PET images combining simulated, phantom and clinical tumors with associated ground-truth or surrogate of truth was assembled. The sphericity of the ground-truth was compared to the one calculated on the segmented volumes produced by 5 different methods: 2 thresholds at 40% and 50% of SUVmax, an ant colony optimization (ACO) algorithm, the fuzzy locally adaptive Bayesian (FLAB), and a gradient-aided region-based active contour (GARAC).

Results: The accuracy in retrieving the true sphericity was related to the segmentation accuracy, as the most accurate methods had also the lowest sphericity errors. However the correlation between the segmentation accuracy and the sphericity measurement error was not very high and varied amongst the segmentation methods.

Conclusion: The tumor functional sphericity is highly dependent on the segmentation method. In addition, even an accurate segmentation can lead to an inaccurate sphericity measurement.

Keywords: PET functional volumes, image segmentation, sphericity
Poster panel
(face) ID: 290

Poster Number:

Ability of Texture and Shape Features to Classify Indeterminate CT Lung Nodules with and without Contrast Enhancement (#4195)

W. Wu1, L. Pierce1, W. A. Chaovalitwongse1, S. Pipavath1, P. D. Lampe2, A. M. Houghton2, P. E. Kinahan1

1 University of Washington, Seattle, Washington, United States of America
2 Fred Hutchinson Cancer Research Center, Seattle, Washington, United States of America


Most indeterminate CT lung nodules are not cancerous. The choices for patient care when small nodules are found are typically either a biopsy of the lung nodule or continued surveillance by CT to monitor nodule growth. A lower-risk early assessment of the probability of a nodule being cancerous would likely improve patient outcomes and reduce cost of care.

Methods: We performed a retrospective study with non-small cell lung cancer (NSCLC) and control cases from a pulmonary nodule clinics. These were roughly divided into non-contrast and contrast-enhanced cases. The nodules were contoured by a radiologist and texture features of the lesion were calculated. Two radiologists evaluated consensus categorical shape features. Multiple machine learning algorithms (MLAs) were evaluated using a leave-one-out cross-validation to evaluate the diagnosis accuracy of the texture and shape features in classifying the nodules.

Results: In the CT contrast-enhanced group, the texture features yielded an overall diagnostic accuracy of 80% in classifying NSCLCs vs. controls. Adding the shape features increased this to 82%. In the non-contrast group, the diagnostic accuracy was 64%. and adding the shape features increased this to 66%. Switching the MLA from Random Forest to Adaptive Logistic Regression increased the diagnostic accuracy to 80% for the non-contrast group, however this decreased the diagnostic accuracy to 80% for the contrast group. Shape features by themselves were not as predictive as texture features.  However, when combined with texture features shape features improve the overall accuracies.

Conclusion: For the indeterminate nodules, the combined use of shape and texture features from contrast-enhanced and non-enhanced CT images have a diagnostic accuracy that warrants further study to evaluate the potential to impact patient management. Further information such as plasma biomarkers may further improve diagnostic accuracy.

Keywords: Lung nodules, screening, texture, radiomics, quantitative imaging biomarkers, low-dose CT
Poster panel
(face) ID: 293

Poster Number:

ParaPET, A New Statistical Approach to Derive 3D Maps of FDG-PET Kinetic Parameters (#1335)

E. Colard1, L. Padovani2, S. Delcourt3, S. Thureau4, 1, B. Farman Ara3, P. Gouel5, A. Dumouchel5, I. Gardin5, 1, P. Vera5, 1, D. Taïeb3, 6, D. Barbolosi7, S. Hapdey5, 1

1 University of Rouen, LITIS QuantIF EA4108, Rouen, France
2 La Timone University Hospital, Department of Radiotherapy, Marseille, France
3 La Timone University Hospital, Department of Nuclear Medicine, Marseille, France
4 Centre Henri Becquerel, Department of Radiotherapy, Rouen, France
5 Centre Henri Becquerel, Department of Nuclear Medicine, Rouen, France
6 European Center for Research in Medical Imaging (CERIMED), Marseille, France
7 University of Marseille, SMARTc, INSERM, UMR 911 CR02, Marseille, France


We extended a statistical methodology recently proposed by Barbolosi et al. allowing the generation of 3D images of FDG kinetic parameters in PET: ParaPET. The major improvements of ParaPET are the inclusion of a new error model of PET measurement, a non-invasive blood activity measurement (NIBAM) and the determination of 3D kinetic parameters. Our approach only requires a late list-mode PET acquisition of 15 minutes centered over the lesion. This acquisition is reconstructed into 5 images of mean FDG activity concentration, associated with 5 images of its variance to model the errors of PET measurements at the voxel level. Regions of interest are manually drawn over blood compartment to extract the mean NIBAM and its associated variance. Our method was carried out on a voxel basis to derive 3D parametric PET images. ParaPET was evaluated and compared to Patlak analysis as a reference. The same comparison was performed for Hunter and Barbolosi’s methods (with blood samples or with NIBAM). We focused our evaluation on the net influx rate Ki. All methods were evaluated on 16 non-small cell lung cancer lesions. Results show that our approach presents an excellent Spearman correlation with Patlak (rSpearman = 0.99) compared to Hunter (rSpearman = 0.94), Barbolosi with blood samples (rSpearman = 0.97) or with NIBAM (rSpearman = 0.95) methods. The mean errors (± SD) in Ki estimate were -0.2% ± 10.2% for ParaPET (non-statistically significant, Wilcoxon signed rank test), 40.5% ± 34.6% for Hunter, -4.9% ± 10.0% for Barbolosi and -1.3% ± 15.5% for Barbolosi with NIBAM (non-statistically significant). ParaPET provides an accurate estimate of Ki. Based on a single late PET acquisition, ParaPET represents a non-invasive alternative to methods using blood samples. It also proposes to derive a set of kinetic parameters on a voxel basis, giving access to additional quantitative information.

Keywords: Dynamic PET, Parametric Imaging, FDG Kinetics, Lung Cancer, Quantification
Poster panel
(face) ID: 296

Poster Number:

Clinical whole body CBM parametric PET with flexible scan modes (#2146)

J. Hu1, V. Panin1, A. M. Smith1, W. Howe1, V. Shah1, F. Kehren1, M. Casey1

1 Siemens Medical Solutions USA, Inc., Molecular Imaging, Knoxville, Tennessee, United States of America


Clinical PET scanners with continuous bed motion (CBM) offers great flexibility to acquire dynamic data for parametric imaging in terms of scan range, scan direction, and scan speed compared to step and shoot acquisition. It is natural to implement multi-pass dynamic scans with sequential same direction, bi-direction, or mixed scan mode with CBM. Ideally, parametric images obtained by different scan modes should be identical since underlying physiology is independent of the scan. However, it is challenging to get consistent quantification results for parametric images among different dynamic CBM scan modes if scans are not accurately tracked and time information is not properly taken into account. In this work, we propose a method to calculate image slice reference time of different scan passes for parametric PET based on finely sampled bed tags which accurately encode position and time information of patient bed. For each image slice, we define its imaging start and end time as time points when the slice enters and leaves scanner field of view based on bed tags. The imaging duration of the slice is defined as the difference between the start and end imaging time. We then calculate image slice reference time as the time point when the average activity occurs due to decay while assuming no activity change due to tracer kinetics over that time duration. This slice reference time is used to obtain the blood input value and calculate the area under the curve from the fitted blood input function for whole body CBM parametric PET based on the Patlak model. This approach of tracking dynamic scans and calculating time information for parametric PET can be readily applied to different CBM scan protocols and solve the problem of non-uniform time sampling over different image slices. We validated the proposed method with dynamic patient data acquired on clinical scanners. We also demonstrated the potential advantages of parametric images over SUV images for tumor detection.

Keywords: Parametric Imaging, PET, Continuous bed motion, Tracer kinetic modeling
Poster panel
(face) ID: 299

Poster Number:

Kinetic Compressive Sensing (#2975)

M. Scipioni1, 3, M. F. Santarelli2, L. Landini1, 2, C. Catana3, 4, D. N. Greve3, 4, J. C. Price3, 4, S. Pedemonte4, 5

1 University of Pisa, Department of Information Engineering, Pisa, Italy
2 National Research Council (CNR), Institute of Clinical Physiology, Pisa, Italy
3 Massachusetts General Hospital, Athinoula A. Martinos Center for Biomedical Imaging, Boston, Massachusetts, United States of America
4 Harvard Medical School, Boston, Massachusetts, United States of America
5 Massachusetts General Hospital, Center for Clinical Data Science, Boston, Massachusetts, United States of America


Introduction: Parametric images provide insight into the spatial distribution of physiological parameters, but they are often extremely noisy, due to low SNR of tomographic data. Direct estimation from projections allows accurate noise modeling, improving the results of post-reconstruction fitting. We propose a method, which we name kinetic compressive sensing (KCS), based on a hierarchical Bayesian model and on a novel reconstruction algorithm, that encodes sparsity of kinetic parameters. 

Methods: The model has three key components: the model of the acquisition system; the kinetic model; and a Markov Random Field with an L1-norm cost function, defined in kinetic parameters domain. Parametric maps are reconstructed by maximizing the joint probability, with an Iterated Conditional Modes approach, alternating the optimization of activity time series (OSL-MAP-EM), and kinetic parameters (MAP-LM): a parallel GPU implementation allows synchronized update of all the voxels, computing the gradient of the log joint posterior at each iteration.

Experiments: 100 noise realizations of a simulated dynamic geometric phantom were generated using a 2TC irreversible model. A bias/variance study confirmed how direct estimates can improve the quality of parametric maps over a post-reconstruction fitting, and showed how the novel sparsity prior can further reduce their variance, without affecting bias. Real FDG PET human brain data (Siemens mMR, 40min) images were also processed. Results enforced how the proposed KCS-regularized direct method can produce spatially coherent images and parametric maps, with lower spatial and better tissue contrast. 

Conclusion: Combining sparse kinetic compressive sensing into a direct reconstruction framework can help generating high-quality images and parametric maps, both amenable for display and quantitatively more accurate than what a post-reconstruction fitting can achieve. A GPU-based open source implementation of the algorithm is provided.

Keywords: parametric images, PET, compartmental models, compressive sensing, hierarchical Bayesian model, sparsity, Markov Random Field, FDG, GPU
Poster panel
(face) ID: 302

Poster Number:

Simultaneous PET Imaging of Antibodies Labeled with Zr-89 and I-124 Based on Triple Coincidences (#3349)

J. L. Herraiz1, U. Mahmood2, J. M. Udías1

1 University Complutense of Madrid, Dep. Fisica Atomica, Molec. y Nuclear. Grupo de Fisica Nuclear, Madrid, Spain
2 Massachusetts General Hospital & Harvard Medical School, Division of Precision Medicine. Athinoula A. Martinos Center at MGH, Boston, Massachusetts, United States of America


Multiplexed PET (mPET) is a new imaging technique able to provide separated images of the biodistribution of two radiotracers based on their standard double coincidences and the triple coincidences generated by one of them. In this work, we evaluated the feasibility of using mPET to improve and facilitate the kinetic analysis of studies with monoclonal antibodies (mAb).  By simultaneously administering and imaging mAb labeled with either Zr-89 or I-124, the differences in the activity concentration in the tumor of each isotope (due to the efflux of the non-residualizing I-124 from the cells after being internalized) can be used to improve tumor detection and the estimation of the binding potential. This method allows considering the non-linear effect of the reduction of the available antigens on the surface of the target tumor caused by the mAb, which is not negligible with therapeutical doses of mAb. The combination of I-124 with Zr-89 can be also used to reduce the effect of non-specific uptake of Zr-89 in the liver, blood and spleen. We have evaluated the proposed method with dynamic realistic simulations of numerical mice phantoms performed with PeneloPET for the preclinical SuperArgus scanner. The activity concentration in each tissue and the acquisition protocol for each considered case were obtained from studies in which a pairwise comparison of specific mAb labeled with Zr-89 and I-124 were performed. The accuracy of the results was established by comparing parameters such as the efflux rate of I-124 from the tumor cells with the published values. The good agreement demonstrates that this in-silico approach can be used for testing different acquisition protocols before in-vivo acquisitions. It also shows that mPET can be an additional new tool for kinetic modelling of mAb studies.

Keywords: multiplexed PET, monoclonal antibody, kinetic model, Zr89, I124
Poster panel
(face) ID: 305

Poster Number:

Direct 4D Patlak 18F-FDG PET/MR for the multi-parametric imaging assessment of active cardiac sarcoidosis (#3552)

N. A. Karakatsanis1, M. G. Trivieri1, R. Abgral1, 2, M. R. Dweck1, 3, P. M. Robson1, V. Mani1, M. M. Padilla4, M. Miller5, A. Lala5, J. Sanz5, J. Contreras5, J. Narula5, V. Fuster5, J. C. Kovasic5, Z. A. Fayad1

1 Icahn School of Medicine at Mount Sinai, Translational and Molecular Imaging Institute, New York, New York, United States of America
2 University Hospital of Brest, Division of Nuclear Medicine, Brest, France
3 University of Edinburgh, British Heart Foundation/University Centre for Cardiovascular Science, Edinburgh, United Kingdom of Great Britain and Northern Ireland
4 Icahn School of Medicine at Mount Sinai, Division of Pulmonary, Critical Care and Sleep Medicine, New York, New York, United States of America
5 Icahn School of Medicine at Mount Sinai, Cardiovascular Institute, New York, New York, United States of America

This work was supported by NIH/NHLBI R01HL071021 grant


Cardiac involvement in sarcoidosis is currently under-diagnosed despite being the leading cause of death amongst sarcoidosis patients. Therefore, accurate and early diagnosis of subclinical but active cardiac sarcoidosis (ACS) is an important clinical goal. The recent advent of hybrid PET/MRI has enabled the multi-parametric and non-invasive evaluation of ACS in the myocardium at a significantly reduced radiation exposure. Although cardiac MRI with late gadolinium enhancement (LGE) may visualize the anatomical pattern of myocardial injury due to ACS, it cannot differentiate between active disease and old chronic scarring. On the other hand, 18F-FDG PET standard uptake values (SUVs), which are used to identify regions of increased myocardial inflammation in patients with ACS, may not discriminate ACS-related inflammatory FDG signal, from high non-specific physiological FDG uptake, often leading to false positive ACS interpretation. In this study, we introduce a clinically feasible dynamic 18F-FDG PET/MR cardiac imaging protocol enabling the 4D analysis of the 18F-FDG myocardial uptake pattern, coregistered with the LGE MR images. In addition, we employ advanced direct 4D Patlak parametric PET image reconstructions to deliver, beside standard semi-quantitative SUVs, highly quantitative images of the physiological parameters of FDG uptake rate (Ki), obtained directly from the complete 4D PET acquisition data, for enhanced robustness to statistical noise and Ki quantification. We believe that our proposed framework of multi-parametric 4D 18F-FDG PET/MR cardiac imaging may substantially improve ACS diagnosis by i) allowing a more accurate identification of positive ACS patterns in matched myocardium regions across LGE, SUV as well as Ki images and ii) enabling the quantitative Ki-driven differentiation between true- and false-positive ACS indications for enhanced specificity.

Keywords: PET/MR, sarcoidosis, cardiac, parametric, direct, 4D, Patlak
Poster panel
(face) ID: 308

Poster Number:

Astatine-211 imaging by a Compton camera for targeted α-particle radiotherapy (#1244)

Y. Nagao1, 2, M. Yamaguchi1, S. Watanabe1, N. S. Ishioka1, N. Kawachi1, H. Watabe2, 3

1 National Institutes for Quantum and Radiological Science and Technology (QST), Takasaki Advanced Radiation Research Institute, Takasaki, Gunma, Japan
2 Tohoku University, Graduate School of Biomedical Engineering, Sendai, Miyagi, Japan
3 Tohoku University, Cyclotron and Radioisotope Center (CYRIC), Sendai, Miyagi, Japan


Astatine-211 is a promising radionuclide for targeted α-particle radiotherapy of cancers. It is required to image the distribution of targeted radiotherapeutic agents in a patient’s body for optimization of treatment strategies and determination of the suitability of a given agent for a particular patient. The 211At and its daughter radionuclide of 211Po emit high-energy photons (570, 687, and 898 keV) at the total intensity of 0.9%. These high-energy photons are not substantially attenuated in the body, and hence, Compton cameras are suitable for visualizing 211At distribution noninvasively. We have developed a Compton camera to image high-energy photons. It consists of two detectors: a scatterer and an absorber. The scintillator material of both the detectors is Ce-doped Gd3Al2Ga3O12. The thicknesses of the scintillators for the scatterer and the absorber are 5 mm and 10 mm, respectively. We used a silicon photomultiplier for the scatterer and a flat panel type multianode photomultiplier tube for the absorber. A conventional data acquisition system for a positron emission tomography camera was diverted. An imaging experiment of a point-like 211At source using the Compton camera was performed to test the 211At imaging capability of the camera, and the source was successfully imaged. Future research plans include the optimization of some parameters such as the energy windows and the coincidence time window.

Keywords: astatine-211, Compton camera, targeted radiotherapy
Poster panel
(face) ID: 311

Poster Number:

Point-spread function modelling for prompt gamma imaging (#1545)

H. - M. Huang1

1 Chang Gung University, Institute for Radiological Research, Taoyuan County, Taiwan


A Compton-based prompt gamma imaging system has been developed to reconstruct the three-dimensional (3D) distribution of prompt gammas. However, the performance of s Compton cameras (CC) is limited by spatial resolution and energy resolution. To improve the image quality of a Compton camera, we investigated an iterative algorithm to incorporate point spread function (PSFs) into the image reconstruction from CC data. We conducted Monte Carlo simulations to estimate the parameters of the PSFs using point sources placed at different positions and to acquire data from a two-stage CC positioned right of and orthogonal to a proton pencil beam irradiating a water phantom. It was found that the PSF-based reconstruction method can improve the quality of prompt gamma imaging. Moreover, the PSF-based reconstruction method can provide accurate falloff position estimates.

Keywords: Prompt gamma Imaging, Compton camera
Poster panel
(face) ID: 314

Poster Number:

Luminescence imaging of water during carbon-ion irradiation for range estimations (#1965)

S. Yamamoto1, M. Komori1, T. Akagi2, T. Y. Tomohiro Yamashita2, T. Toshito3

1 Nagoya University Graduate School of Medicine, Nagoya, Aichi, Japan
2 Hyogo Ion Beam Medical Center, Ako, Hyogo, Japan
3 Nagoya Proton Therapy Center, Nagoya, Aichi, Japan


We previously reported that the luminescence imaging of water during proton irradiation is possible and can be used for range estimation. However, since it remains unclear whether such luminescence imaging of water is possible for carbon-ion irradiation, we conducted luminescence imaging during carbon-ion irradiation and estimated the ranges. We placed a pure-water phantom on a patient couch with a carbon-ion therapy system and measured the luminescence images with a high-sensitivity cooled charge-coupled device (CCD) camera during 240MeV/n carbon-ion irradiation. We also conducted imaging of the phantoms of tap water, an acrylic block, and a plastic scintillator and compared their intensity and distribution with those of a pure-water phantom. The luminescence images of the pure-water phantoms during the carbon-ion irradiation showed clear Bragg peaks with higher distribution in the shallow part of water, and the measured carbon-ion ranges from the images were almost the same as those obtained with a simulation. The image of the tap-water phantom showed almost the same distribution as the pure-water phantom. The acrylic phantom’s luminescence image produced seven times higher luminescence and had a 13% shorter range than that of water; the range in the acrylic phantom generally matched the calculated value. The plastic scintillator showed ~15,000 times higher light than that of water. Luminescence imaging during carbon-ion irradiation of water is possible and has the potential to become a new method for range estimations in carbon-ion therapy.

Keywords: Luminescence, imaging, water, carbon-ion, range estimations
Poster panel
(face) ID: 317

Poster Number:

A novel approach to contamination suppression in upstream detectors for radiotherapy (#2509)

L. Beck1, R. F. Page1, J. J. Velthius1, C. T. Nsangu1, R. P. Hugtenburg2

1 University of Bristol, Physics, Bristol, United Kingdom of Great Britain and Northern Ireland
2 Swansea University, Swansea, United Kingdom of Great Britain and Northern Ireland


As intensity modulated radiotherapy treatments are becoming more complex, new methods are required to monitor them. Downstream beam monitoring is difficult as the beam is distorted by the patient, whilst upstream monitoring attenuates the beam. A novel method for real-time upstream beam monitoring is presented using monolithic active pixel sensors. These can be made as thin as 20 microns, which gives a beam attenuation of less than 0.1%. The challenge is to separate the response from therapeutic photons and from contamination electrons, due to Compton scattering of the photons in the air or accelerator, or from pair-production. The signal from photons is mainly produced via Compton scattering in the sensor and is dependent on the thickness of the bulk layer when the device is illuminated from the back. When etching a pattern of strips into the top bulk layer of the sensor, only the signal from photons should be modulated by the strips. Hence the photon contribution can be determined by taking the difference between the signal under a strip and under a trench. Knowledge of the trench depth allows the dose to be determined without relying on simulation. A proof of principle study was performed using a range of field sizes delivered by a linear accelerator. As a prototype, 50 micron deep trenches were etched in a 300 micron thick piece of silicon and placed on top of the MAPS. A fit was made to the signal response using a second order polynomial for the background field shape and a sinusoidal wave for the modulation of the photon signal caused by the strips. The pitch of the strips was determined through analysis to be 398.8±0.2 microns whilst the true strip pitch is known to be 400±10 microns. The photon contribution to the signal was found to be constant with field size and hence the dose to the patient can be extracted making this a successful method of beam monitoring.

Keywords: monolithic active pixel sensor, intensity modulated radiotherapy, beam monitoring, MAPS, upstream, compton scattering, silicon, etching, contamination suppression, modulation
Poster panel
(face) ID: 320

Poster Number:

A novel method of measuring nuclear reaction cross sections by using Cherenkov radiation toward high-precision proton therapy (#2728)

T. Masuda1, J. Kataoka1, M. Takabe1, M. Arimoto1, T. Nishio2, K. Matsushita3, T. Miyake4, S. Yamamoto5, T. Inaniwa6, T. Toshito7

1 Waseda University, Graduate School of Advanced Science and Engineering, Shinjuku-ku, Tokyo, Japan
2 Tokyo Women’s Medical University, Department of Radiation Oncology, Shinjuku-ku, Tokyo, Japan
3 Kyoto Prefectural University of medicine, Department of Radiology, Kyoto-shi, Kyoto, Japan
4 Rikkyo University, Graduate School of Science, Toshima-ku, Tokyo, Japan
5 Nagoya University, Graduate School of Medicine, Nagoya-shi, Aichi, Japan
6 National Institute of Radiological Sciences, Department of Accelerator and Medical Physics, Chiba-shi, Chiba, Japan
7 Nagoya Proton Therapy Center, Nagoya-shi, Aichi, Japan


We propose a novel method of measuring nuclear reaction cross sections that produce positron emitters. They are essential to estimate the dose delivered by protons. The key idea reported here is the use of Cherenkov radiation emitted from relativistic positrons generated via β+ decay. We irradiated a quartz glass SiO2 with 70-MeV protons provided by a cyclotron of the National Institute of Radiological Sciences (NIRS), Japan. After the proton irradiation, we measured the temporal evolution of the Cherenkov radiation along the proton path using an electron multiplying charge-coupled device (EM-CCD) camera. We acquired the cross sections of 16O(p,X)15O, 16O(p,X)13N, and 16O(p,X)11C, and compared them with the archival database provided by the National Nuclear Data Center (NNDC). The overall trend of our cross sections is consistent. In some of the data, the cross sections had many more data points and smaller error values compared to the other data. Furthermore, we implemented the cross sectional data into a Monte Carlo (MC) simulation, PHITS, to validate the results. We compared the simulation with the observations from the positron emission tomography (PET) measurements of water that was phantom irradiated with the same 70-MeV protons of the NIRS. Our MC simulation and PET measurements were in good agreement with each other, in terms of both the spatial and temporal evolutions of the PET image. Thus, our proposed method is viable to measure various cross sections with a high accuracy and wide energy range, and helpful for future proton therapy.

Keywords: proton therapy, PET, nuclear reaction cross section, MC simulation, Cherenkov radiation
Poster panel
(face) ID: 323

Poster Number:

Characterization and Modeling of Knife-Edge Slit Collimator Response for MeV Prompt Gamma Photons in Proton Therapy Monitoring (#2996)

W. Lu1, 2, P. Fan1, 2, Y. Liu1, 2, Z. Wu1, 2, S. Wang1, 2, T. Ma1, 2

1 Tsinghua University, Department of Engineering Physics, Beijing, China
2 Ministry of Education (Tsinghua University), Key Laboratory of Particle & Radiation Imaging, Beijing, China


The Gamma camera imaging process can be well characterized and modeled by the response of the imaging system and collimator response is one of most important parts. In previous research, many models were proposed to formulate analytically the pinhole or parallel-hole collimator response for conventional gamma cameras. However, few studies take the photons of MeV energy range into consideration. In proton therapy monitoring, the prompt gammas emitted during proton therapy delivery have high energy with a range 2~10 MeV, which makes the collimator response dramatically different from that of conventional gamma cameras. Accurate analytical modelling of collimator response for high energy photons is important to generate the system matrix within a reasonable time and improve the accuracy of the reconstructed images in proton therapy monitoring. The knife-edge slit collimator based imging system is one of the most promising systems for proton therapy monitoring. However, few models were proposed for knife-edge slit collimator response in the existing literature.

We performed MC simulation using Gate and generated the projections of primary, penetrated and scattered photons. By analyzing the projections of the three kinds of photons, we characterized the knife-edge slit collimator response for prompt gammas dependent on the photon energy, point source position and geometric design of knife-edge slit. Meanwhile, fitting based modeling of primary and penetrated projection profiles were developed and verified. Preliminary results show good fitting of the simulated projections in limited cases. More investigations will be done to analyze the influence of various factors including photon energy, point source position and collimator design on the scattered photon projections quantitatively. We aim at proposing a general applicable semi-empirical equations or index look-up tables to calculate the collimator response for high energy photons.

Keywords: collimator response model, knife-edge slit, prompt gamma, proton therapy monitoring
Poster panel
(face) ID: 326

Poster Number:

Optimizing energy window for Yttrium-90 bremsstrahlung imaging: Monte-carlo simulation (#3218)

M. H. Kim1, 2, J. K. Bae1, W. Lee2, K. M. Kim1

1 Korea Institute Radiological and Medical Sciences, Divion of Medical Radiation Equipment, Seoul, Republic of Korea
2 Korea University, Dept. of Bio-convergence engineering, Seoul, Republic of Korea


Yttrium-90 (Y-90) is useful radioisotope in tumor treatment because of emitting high energy electron (Emax= 2.27 MeV, Emean = 0.9367 MeV). This physical characteristic makes Y-90 possible to being imaging during therapy. However, it is difficult to determine proper collimator and energy window because of its continuous bremsstrahlung spectra.

In this study, we aimed to better energy window for Y-90 bremsstrahlung imaging when using conventional gamma camera by using Monte-carlo simulation.

We used GATE package in the Monte-Carlo simulation. We realized a commercial clinical SPECT camera (SYMBIA-T2, Siemens) with HEGP (High Energy General Purpose) collimator. For confirmation generation of bremsstrahlung spectra, we realized 1.1 cm radius water sphere phantom and 0.1 mm radius point source located center of phantom. And we also simulated 10cm radius water sphere phantom to confirmation changing bremsstrahlung spectra caused by scattering material. A human-like voxelized phantom (KTMAN-2) was used to realize Y-90 bremsstrahlung image.

First we confirmed bremsstrahlung spectra in full energy range (50 to 500 keV) of SYMBIA both water sphere and KTMAN-2 phantom. We analyzed Y-90 bremsstrahlung energy spectrum by separating primary and scattering spectrum. After that we defined favorable energy window by signal to noise pattern, scatter fraction and windowing AUC (Area Under Curve).

Bremsstrahlung energy spectrum cause by Y-90 showed continuous shape as expected. As the thickness of scattering medium became thicker, scatter ratio became higher compared with primary. Primary to scatter ratio was changed by the scattering medium. In KTMAN-2 phantom simulation, the lower energy range, the lower scatter fraction, but the count in images increased dramatically. And primary to scatter ratio was increased until 200 keV.

The energy window as low as possible in the range below 200 keV would be advantageous at Y-90 bremsstrahlung gamma imaging.

Keywords: Yttrium-90, Bremsstrahlung imaging, energy window, Monte-Carlo simulation
Poster panel
(face) ID: 329

Poster Number:

Characterization of a Compton camera setup with monolithic LaBr3(Ce) absorber and segmented GAGG scatter detectors (#3467)

S. Liprandi1, S. Takyu2, S. Aldawood1, 4, T. Binder1, G. Dedes1, K. Kamada3, R. Lutter1, M. Mayerhofer1, 5, A. Miani1, 6, A. Mohammadi2, F. Nishikido2, D. R. Schaart7, I. I. Valencia Lozano1, E. Yoshida2, T. Yamaya2, K. Parodi1, P. G. Thirolf1

1 Ludwig Maximilians Universitaet Muenchen, Department of Physics, Munich, Germany
2 National Institutes for Quantum and Radiological Science and Technology (QST), National Institute of Radiological Sciences (NIRS), Chiba, Japan
3 C&A corporation, Sendai, Japan
4 King Saud University, Department of Physics and Astronomy, Riyadh, Saudi Arabia
5 University of Hamburg, Department of Physics, Hamburg, Germany
6 Universita degli Studi di Milano, Department of Physics, Milano, Italy
7 Delft University of Technology, Radiation Science and Technology, Delft, Netherlands


Prompt-γ based medical imaging, in particular exploiting the Compton scattering kinematics in a Compton camera detector arrangement, is one of the presently explored techniques targeting an in-vivo particle beam range monitoring in hadron therapy. To evaluate the performance of different Compton camera setup options, we characterized a detector system consisting of a segmented array of GAGG scintillation crystals as scatterer and a monolithic LaBr3(Ce) scintillator acting as absorber, chosen due to its excellent energy and time resolution, already demonstrated at LMU Munich in a separate study. 
The scintillator is coupled to a 256-fold segmented position-sensitive multi-anode photomultiplier. The GAGG detector array consists of 22x22 individual crystals (0.9x0.9x6 mm3), read out by a 64 ch MPPC SiPM array. Distances of 5 cm and 20 cm, respectively, between the Compton camera scatterer and absorber components were investigated. A 137Cs calibration source (Eγ = 662 keV) was used for offline characterization of the Compton camera, placed in a distance of 4.5 cm from the scatterer in three different horizontal positions. 
After having selected the coincidence events between scatterer and absorber, the primary photon interaction position in the GAGG array was determined using an Anger logic calculation. 
The position determination in the monolithic absorber was performed using the “Categorical Average Pattern” version of the k-Nearest-Neighbors algorithm. 
In a first approach, photon source image reconstruction was performed based on the Compton cone defined by the interaction positions and energy deposits in the two detector components. The status of the characterization analysis will be presented.

This work is supported by the NIRS International Open Laboratory, the DFG Cluster of Excellence Munich Centre for Advanced Photonics (MAP) and King Saud University (KSU).

Keywords: Compton camera, monolithic scintillator, hadron-therapy, gamma-ray medical imaging
Poster panel
(face) ID: 332

Poster Number:

A Feasibility Study of Prompt Gamma Imaging during Boron Neutron Capture Therapy using a Gamma Camera: GATEv7.1 Simulation Study (#3704)

H. G. Kang1, J. K. Bae2, M. H. Kim2, S. J. Hong1, 3, K. M. Kim2

1 Eulji University, Department of Radiological Science, Daejeon, Republic of Korea
2 Korea Institute of Radiological and Medical Science, Seoul, Division of Medical Radiation Equipment, Seoul, Republic of Korea
3 Eulji University, Department of Radiological Science, Seongnam, Democratic People's Republic of Korea


Boron capture neutron therapy (BNCT) is a biochemically targeted radiotherapy technique which can deliver an optimal dose to the tumor by using nuclear capture and fission reactions. As thermal neutrons are irradiated into the boron which is accumulated in tumor, nuclear capture and fission reactions are occurred, then high linear energy transfer (LET) alpha particle (4He) and prompt gamma (PG) photon of 478 keV are produced. The aim of this study is to investigate the feasibility of dose monitoring of BNCT by detecting the PG ray with a gamma camera in GATEv7.1 Monte Carlo simulation. Two boron spheres with a diameter of 10 mm were embedded inside the center of a cylindrical water phantom (D=100 mm, L=100 mm). The thermal neutron of 106 events was irradiated into the boron spheres. In order to detect the PG photon produced by the boron neutron capture reaction, a gamma camera array was placed on top of the water phantom. Each gamma camera consists of a parallel square-hole tungsten collimator (length=30 mm, septum thickness = 0.12 mm, hole diameter = 1.0 mm) and a monolithic GAGG crystal (48 × 48 × 5 mm3). In order to check the energy spectrum of the gamma camera, the degradation of energy resolution in detector components was assumed to be negligible. The PG of 478 keV can be found on the energy spectrum of the gamma camera. The production of alpha particle and lithium ion can be verified by using the “ProductionAndStoppingActor” in GATE. The dose distribution inside the boron spheres showed good correlation with the alpha particle and lithium ion distribution. The optimal collimator length for the PG detection was 150 mm. In future, the collimator design and scintillation crystal length will be optimized for the BNCT-SPECT imaging.

Keywords: Boron neutron capture therapy (BNCT), GATE simulation, Gamma camera, Radiotherapy
Poster panel
(face) ID: 335

Poster Number:

Towards a MAPS sensor for in vivo beam monitoring of MRI guided radiotherapy (#4082)

J. J. Velthuis1, 2, R. Page1, L. Wilkins1, R. P. Hugtenburg2, 1, L. Beck1

1 University of Bristol, physics, Bristol, United Kingdom of Great Britain and Northern Ireland
2 Swansea University, Medical School, Swansea, United Kingdom of Great Britain and Northern Ireland


An important trend in cancer treatment is the increased use of intensity modulated radiotherapy (IMRT), where the LINAC beam is dynamically shaped using multileaf collimators (MLC) to conform to the tumour shape and optimize the dose distribution. This way the dose to healthy tissue is minimized while the dose to the tumour is maximized. Since the MLC move, monitoring of the leaf positions and verification of the treatment is essential. We have already shown that we can achieve 52±4μm precision on the leaf positions using only 0.1 second of data. However to deliver the therapy effectively, not only does the beam need to be monitored with great precision but so does the tumour position. One way to monitor the tumour position in real time is to combine an IMRT machine with an MRI machine. This allows real time tumour position measurements and treatment verification. We have tested our beam monitoring system inside the bore of an MRI-IMRT machine and proved that it still works well inside the high magnetic field needed to do MRI. The sensor response shows the expected asymmetry due to electrons generated in the build up above the sensor being swept in one direction by the magnetic field. This is proved by increasing the amount of build up and showing that the steepness of the penumbra changes accordingly. This and more results of our tests will be presented.

Keywords: IMRT, MRI, image-guided treatment, radiation hardness, x-ray imaging
Poster panel
(face) ID: 338

Poster Number:

Full dynamic brain simulation using GATE in a high-performance computer (#1155)

L. Caldeira1, 2, S. Lalitha1, M. Lenz1, 2, R. Deepu3, W. Klijn3, N. J. Shah1, 4, U. K. Pietryzk1, 2

1 Forschungszentrum Jülich GmbH, Institute of Neuroscience and Medicine, Medical imaging Physics, Jülich, North Rhine-Westphalia, Germany
2 University of Wuppertal, Faculty of Mathematics and Natural Sciences, Wuppertal, Germany
3 Forschungszentrum Jülich GmbH, Simulation Lab Neuroscience - Bernstein Facility for Simulation and Database Technology, Institute for Advanced Simulation, Jülich Aachen Research Alliance, Jülich Supercomputing Centre, Jülich, North Rhine-Westphalia, Germany
4 RWTH Aachen University, Department of Neurology, Faculty of Medicine, JARA, Aachen, Germany


Dynamic PET brain studies are common in research and are becoming common in clinical applications. Simulation of dynamic PET is an important step to validate techniques and methods. This study compares a GATE simulation running on a CPU versus running on a GPU. We simulated the 3T Siemens MR-BrainPET with dynamic brain activity. The results show close agreement in the number of coincidence events, including phantom scatter. The code will be made available to the GATE community and the simulation data is available to the interested researchers.

Keywords: Dynamic PET, Simulation, High-Performance Computing
Poster panel
(face) ID: 341

Poster Number:

Development of helium CT (HeCT) imaging: Monte Carlo simulation and first experimental results (#1892)

P. Piersimoni1, V. A. Bashkirov2, C. A. Collins Fekete3, B. A. Faddegon4, R. P. Johnson5, C. E. Ordoñez6, J. Ramos Méndez4, R. W. Schulte2, L. Volz1, 7, J. Seco1, 7

1 German Cancer Research Center - DKFZ, Medical Physics in Radiooncology, Heidelberg, Baden-Württemberg, Germany
2 Loma Linda University, Radiation Research, Loma Linda, California, United States of America
3 Université Laval, Physics Department, Quebec, Québec, Canada
4 University of California San Francisco, Radiation Oncology, San Francisco, California, United States of America
5 University of California Santa Cruz, , Institute for Particle Physics, Santa Cruz, California, United States of America
6 Northern Illinois University, Center for Research Computing and Data, DeKalb, California, United States of America
7 Universität Heidel­berg, Physics Department, Heidelberg, Baden-Württemberg, Germany


The purpose of this work is to evaluate the accuracy of relative stopping power (RSP) measured with helium CT (HeCT) using a Monte Carlo (MC) simulation and experimental data. A prototype proton CT (pCT) scanner, able to track single particle with a frequency of 1 MHz, was installed on the beam line at the Heidelberg Ion-Beam Therapy (HIT) facility, Germany. The scanner, extensively tested for proton beams in the past, was for the first time used for HeCT. The pCT scanner was simulated using the TOPAS tool to study the accuracy of the pCT reconstructed RSP values, obtained with 200 MeV/u proton or helium ion beams. An ideal setup was implemented (flat beam source and ideally totally absorbing energy-range detector) so that the theoretically best achievable RSP accuracy could be estimated. Different phantoms were used: a cylindrical water phantom with inserts of different materials, sizes and positions (simulation only); a Catphan phantom with high resolution and sensitometry modules (both simulation and experiment); a voxelized head phantom (simulation only); the CIRS HN715 pediatric phantom (experiment only). The accuracy for the simulated ideal setup was better than 0.8% for the water phantoms with all the different inserts. The Catphan phantom were reconstructed and an average RSP accuracy better than 0.6% was obtained with a spatial resolution of 4 lp/cm. The digitized phantom was reconstructed with HeCT, and all structures were well recognizable in the reconstructed image and no artifacts were visible. The main tissue materials were well identifiable in the reconstructed RSP-volume distribution. Results for all the phantoms are in close agreement with those obtained for pCT. The reconstructions of the CIRS phantom are without artifacts and all structures are well visible. In conclusion MC simulation was used to demonstrate accurate image reconstruction using helium beams. The prototype scanner installed at the HIT facility proved to work well with helium beams.

Keywords: pCT, Topas, helium imaging, Geant4
Poster panel
(face) ID: 344

Poster Number:

Multiplexed Pinholes with Circular Apertures and Elliptical Ports for I-123 DAT Imaging (#2082)

A. Könik1, J. D. Beenhouwer2, Y. He1, 4, T. Fromme1, 4, L. R. Furenlid5, M. A. King1

1 UMMS, Radiology, Worcester, Massachusetts, United States of America
2 University of Antwerp, iMinds-Vision Lab, Antwerp, Belgium
4 Worcester Polytechnic Institute, Mechanical Engineering, Worcester, Massachusetts, United States of America
5 University of Arizona, Center for Gamma-Ray Imaging, Tucson, Arizona, United States of America


Earlier, we proposed an inexpensive method to improve the performance of the conventional dual-head SPECT systems for I-123 dopamine transporter (DAT) imaging of Parkinson Disease. In this approach, the collimator on one head is replaced with a multi-pinhole (MPH) collimator while retaining the conventional collimator on the other head, thus enabling combined MPH/Parallelbeam (or Fanbeam) acquisition. Our original MPH design is based on 9 pinholes with square apertures and rectangular exit/entrance ports covering a cylindrical region around the striatum without multiplexing. While the rectangular ports provide the best coverage of brain volume, the imaging of striatum, which is the main goal of the MPH component, can be further improved with the usage of other pinhole designs and multiplexing. In this work we propose the usage of pinholes with circular apertures and elliptical ports (Elliptic). Circular apertures provide uniform resolution, more accurate point spread function modeling and fabrication while the elliptical ports provide more flexibility in the detector coverage compared to circular ports. We modeled the MPH system with 4 additional pinholes (total of 13) using the GATE Monte Carlo simulation package and obtained projections of XCAT brain phantom with I-123 DAT distribution for three different pinhole shapes: square aperture with rectangular ports (Trapezoid); circular aperture with circular ports (Conic); and Elliptic. We will extend this work for 8 additional pinholes (total of 17 pinholes) and simulate the variation in patient positioning as well as variation in anatomy. We will also investigate the temporal extent of opening (multiplexed data) and closing (non-multiplexed data) of these extra pinholes to obtain artifact-free reconstruction. The usage of non-multiplexed Fanbeam data will allow us to employ a greater percentage of the time with the multiplexing of MPH and hence maximizing the count sensitivity.

Keywords: SPECT, Multipinhole, Simulation, GATE, Multiplexing
Poster panel
(face) ID: 347

Poster Number:

Towards system matrix incorporating efficient detector modeling: A small animal SPECT study on several strategies (#2215)

B. Auer1, 2, F. Boisson1, 2, V. Bekaert1, 2, D. Brasse1, 2

1 Université de Strasbourg, Institut Pluridisciplinaire Hubert Curien (IPHC), Strasbourg, France
2 CNRS/UMR7178, Strasbourg, France


Nowadays, the use of gold standard Monte Carlo Simulation (MCS) modeling of the acquisition process in a way to compute the System Matrix (SM) is one of the well-established methods that has been used in small animal SPECT image reconstruction.

However, MCS requires extensive computation time to obtain a low noise SM. Such reconstruction methods are therefore largely penalized by the huge time consumption required for the SM generation since a large number of photons has to be generated: an improvement in simulation speed is thus mandatory. Simplified analytical approach has the potential to lead to efficient SM computation, in a reduced time while requiring neither particular computing skills nor heavy informatics resources (computational grid).

In this work, we proposed to evaluate several modeling types (analytical and MCS) of the acquisition process of a pinhole SPECT system available at our institute. Secondly, various complexity degrees of an efficient and simplified analytical modeling of the physical effects occurring into the detector during SPECT examinations will be investigated.

The two-developed analytical modeling of detector response represent some fast and efficient alternatives strategies to the implemented MCS based one. Although, even if they are less accurate, they allow coherent estimation while overcoming the disadvantages of the MCS like excessive computation time, high technical complexity and heavy computation infrastructures. However, the performance obtained, both qualitatively and quantitatively, do not allow their use in a quantitative reconstruction process. Nevertheless, recovery coefficient divergences with respect to the MCS reference modeling are on average of the order of ∼ 3%.

Keywords: Small animal SPECT, Monte Carlo Simulation, System response modeling, System Matrix
Poster panel
(face) ID: 350

Poster Number:

Optical Monte-Carlo Simulation to Evaluate Monolithic PET Detector Concepts (#2418)

J. Grahe1, F. Müller1, D. Schug1, P. Hallen1, V. Schulz1, 2

1 University of Aachen, Department of Physics of Molecular Imaging Systems, Institute for Experimental Molecular Imaging, Aachen, North Rhine-Westphalia, Germany
2 Philips Research, Department of Molecular Imaging Systems, Aachen, North Rhine-Westphalia, Germany


A variety of monolithic scinillator setups for positron emission tomography are currently investigated as an alternative to pixelated scintillators. We present a framework, which is based on Geant4, to evaluate the spatial and energy resolution of different detector geometries. Gamma particles generated by the decay of 22Na pass a pinhole collimator with a diameter of 0.5 mm before interacting with the 32 x 32 x 10 mm3 monolithic LYSO crystal. 92% of the gamma particles reaching the crystal have an energy of 511 keV, 6% have a lower energy and 2% have higher energies up to 1.27 MeV. The scintillator is coupled to an 8 x 8-pixel digital SiPM, whose readout characteristics are replicated in the simulation. The scintillation properties of the crystal were determined by comparing simulated and experimental energy spectra obtained for different positions and, additionally, with and without highly-reflective Teflon wrapping. Best accordance is obtained for a light yield of 33000 photons per MeV, an intrinsic energy resolution of 14% full width at half maximum (FWHM) and a mean free path of 18 cm for optical photons in LYSO. The reflectance of Teflon is determined as 92.5%. For the Teflon wrapped crystal, the incident beam position was varied along a perpendicular, equidistant grid with a pitch of 0.75 mm. The energy resolution over the whole crystal without applying filters is 30% (FWHM) in experiment and simulation. The positional dependence shows good agreement with measurements. After discarding 10% of the events with the lowest photon count as noise, we use gradient tree boosting with 50 estimators and depth 13 to position the remaining data. This method yields an overall spatial resolution of 1.6 mm (FWHM) in the simulation, which agrees well with measured data. The use of an intrinsic retro-reflector with the same reflectivity as Teflon improves the average spatial resolution down to 1.0 mm (FWHM) but reduces the overall, unfiltered energy resolution to 39% (FWHM).

Keywords: monolithic scintillator, positron emission tomography, optical simulation, Geant4, digital SiPM
Poster panel
(face) ID: 353

Poster Number:

Simulation of a two-panel head and neck dedicated PET system (#2491)

M. Li1, S. Abbaszadeh1

1 University of Illinois, Nuclear, Plasma, and Radiological Engineering, Urbana, Illinois, United States of America


Whole body (WB) positron emission tomography/computed tomography (PET/CT) has become a standard clinical imaging modality in patients with head and neck cancers (HNC). PET/CT is used to establish how far the tumor has invaded locally and to guide the decision to resect a tumor rather than irradiate and give chemotherapy. For example, a tumor along the mucosal surface of the larynx is resectable, unless it invades a few mm deeper and invades into the thyroid cartilage. Therefore, a high spatial resolution is especially important in the neck because layers of tissues are so thin. Current WB PET spatial resolution is about 4-6 mm. In this study, we are investigating a dedicated two-panel HNC system that complements the WB PET. We perform a Monte Carlo simulation on the two-panel system. The scintillator was Lutetium-yttrium oxyorthosilicate and the pixel size was 0.5 x 0.5 x 20 mm3. The small crystal size better enables inter-crystal Compton recovery. A cylinder phantom with 4 hot rods whose diameters were 1mm, 2mm, 3mm and 4mm was used. The total activity of F-18 was set to 10 MBq and the activity ratio of hot rods to background was 4:1. The contrast-to-noise ratio (CNR) of the four rods in the two-panel system was calculated. It was shown that recovering the inter-crystal scatter that undergo one Compton scatter followed by a photoelectric event will improve the CNR by a factor of ~2 and the sensitivity by a factor of ~2.5. A GE Discovery 690 WB PET will be simulated with the same phantom and the benefit of the dedicated system will be evaluated.

Keywords: Head and neck cancers, dedicated positron emission tomography
Poster panel
(face) ID: 356

Poster Number:

MCNP dosimetry study for prediction of Sn-117m radiopharmaceuticals injection dose (#2859)

S. J. Jeon1, B. I. Kim1, H. Youn2, J. G. Kim1

1 Korea Institute of Radiological and Medical Sciences, RI Fusion Division, Seoul, Republic of Korea
2 Pusan National University Yangsan Hospital, Department of Radiation Oncology, Yangsan, Republic of Korea


The simulation based small animal dosimetry methods are commonly used for the evaluation of new radionuclide therapy. In this study, we have investigated therapeutic internal radiation dose effects on Sn-117m rituximab, and calculated radiation doses on mouse tumor model for prediction of its injection dose.

The absorbed doses for internal radionuclides (Sn-117m and Iodine-131) were calculated using mouse digital phantom and Monte Carlo simulation code, MCNPX V2.7. The mean dose per cumulated activity, S-values (mGy/Bq/s) were simulated for different source target organs for a number of radio-nuclides of interest organ regions. The voxel based mouse digital phantoms were designed using mouse whole-body(MOBY) phantom, and the xenograft tumor mass of 1053 mg were additionally coded to the right flank of the voxel based mouse model. The physiologically based bio-kinetic factor results were obtained by actual bio-distribution tumor xenograft mouse measurement data at 1,2,3,5,7,14d post injection(n=5). The doses of 111 kbq /200ug/head were bolus injected to normal and tumor xenograft mousses. The average doses for organs itself and others were calculated as a function of fractional energy depositions.

The results showed the effective doses of 10 organs and tumor for mouse dosimetry of Sn-117m radionuclide therapy. Also, we calculated the S-value ratio of Sn-117m and I-131 including 10 organs, the blood, xenograft tumor, and remained body.

In this study, we can support to estimate Sn-117m radiopharmaceuticals production yield. The calculated S-values of Sn-117m compared with I-131. It can further be applied to evaluate pharmacokinetics and pharmacodynamics research with dose calculation. We expect that the results can be used to determine proper administrative injection dose for radiopharmaceuticals.

Keywords: MCNP dosimetry, Sn-117m radiopharmaceuticals, prediction of injection dose
Poster panel
(face) ID: 359

Poster Number:

Monte Carlo-based reconstruction of focusing collimators and application to liver dosimetry (#3266)

M. M. A. Dietze1, R. Bastiaannet1, M. G. E. H. Lam1, M. A. Viergever1, H. W. A. M. de Jong1

1 University Medical Center Utrecht, Radiology and Nuclear Medicine, Utrecht, Netherlands


Monte Carlo (MC) based reconstruction schemes have been successful in the realization of quantitative SPECT images, which is especially relevant in dosimetry for radionuclide therapy. One particular example is liver dosimetry in the context of radioembolization, requiring high accuracy, ideally combined with fast scanning to allow for dosimetry during the intervention.

To achieve faster scanning, focusing collimators can be used. However, the reduced field-of-view currently limits their application to brain applications. Although challenging to implement, multifocal collimators combine a larger field-of-view with increased performance. This is now common practice for cardiac applications, but has never been employed in larger organs such as the liver. The purpose of this simulation study is to explore if fast and accurate SPECT based liver dosimetry can be realized by combining MC based reconstruction with a converging collimator acquisition.

A MC simulator of parallel hole, cone beam and a multifocal collimator (both with focal length 50 cm) was implemented using an extended and spatially varying version of the convolution based forced detection algorithm. Projections were tested on a NEMA IQ phantom and liver SPECT simulation studies using the realistic XCAT phantom with liver tumors for varying acquisition times per projection angle. MC-based reconstructions were additionally compared to a standard OSEM reconstruction.

MC-based reconstruction led to a contrast to noise ratio in the tumor that was more than two times higher as compared to standard OSEM. Compared to the parallel hole collimator, focusing collimators reduced acquisition times by 33%. Truncation may occur when using conebeam collimator, which is solved with the multifocal collimator.

MC-based reconstruction with focusing collimators allows for fast and accurate SPECT also in the liver, which is crucial to realize dosimetry in time critical procedures such as radioembolization.

Keywords: quantitative imaging, SPECT, focusing collimators, dosimetry, monte carlo, iterative reconstruction
Poster panel
(face) ID: 362

Poster Number:

Light Distribution Models for Compact Gamma Camera (#3550)

A. Bricou1, 3, C. Trigila1, M. - A. Verdier2, M. - A. Duval1, F. Lefebvre1, L. Pinot1, A. Dubois1, L. Menard2, Y. Charon2

1 IMNC CNRS/IN2P3, Univ. Paris Sud et Paris Diderot, 91405 Orsay Cedex, France
2 Univ. Paris-Diderot, Sorbonne Paris Cité, IMNC CNRS/IN2P3, 91405 Orsay Cedex, France
3 Department of Gynecology and Obstetrics Jean Verdier Hospital/AP-HP, The University Paris Nord, Bondy, France


Compact gamma cameras are very attractive to provide more efficient localization of pathological structures to be resected during radio-guided cancer surgery. These devices have especially a great interest in the case of the SNOLL (Sentinel Node and Occult Lesion Localization) protocol for breast cancer. We successively developed two light hand-held gamma cameras with dedicated readout electronics providing accurate images and short exposure time capabilities based on a continuous scintillator and pixelated photo-detector assembly.

We report here results of a study of light distribution models models: the standard Scrimger-Baker model and a customized model taking into account reflected light in the crystal on the opposite face of the photo-detector. The goal is to improve our understanding of the camera and increase the efficiency of position of interaction reconstruction. We compare experimental measurements of the light density profile to both optical Monte Carlo simulation with the Gate platform and Monte Carlo simulation based on these models. We also report results of simulation for the measurement of energy resolution and depth of interaction with hese models, as well as their impact on the intrinsic performances of the camera when used for position of interaction reconstruction.

Keywords: Gamma Camera, Monte Carlo, Light Distribution
Poster panel
(face) ID: 365

Poster Number:

Evaluation of Geant4 Monte Carlo Toolkit Physics Models for Use in Heavy Ion Therapy (#3962)

M. Safavi-Naeini1, 2, A. Chacon2, 1, H. Rutherford2, 1, S. Guatelli2, 3, A. Mohammadi4, M. Nitta4, F. Nishikido4, Y. Iwao4, H. Tashima4, E. Yoshida4, T. Yamaya4, T. Hofmann5, M. Pinto5, K. Parodi5, M. - C. Gregoire1, A. Rosenfeld2

1 Australian Nuclear Science and Technology Organisation, Human Health, Lucas Heights, NSW, Australia
2 University of Wollongong, Centre for Medical Radiation Physics, Wollongong, NSW, Australia
3 University of Wollongong, Illawarra Health and Medical Research Institute, Wollongong, NSW, Australia
4 National Institutes for Quantum and Radiological Science and Technology, National Institute of Radiological Sciences (NIRS), Inage-ku, Chiba, Japan
5 Ludwig-Maximilians-Universitat Munchen, Medical Physics, Munchen, Bavaria, Germany


Carbon ion therapy is a highly conformal radiotherapy modality which can deliver a therapeutic radiation dose to a tumour while limiting significant damage to surrounding healthy tissue or organs. However, treatment with carbon ions is highly sensitive to positioning error compared to photon therapy, due to the steep dose profile, which can result in under-dosing the tumour and/or overdosing healthy tissue. One method for verifying the dose distribution delivered to the patient is to image the distribution of short-lived positron-emitting fragmentation products generated as the beam passes through the patient using PET. In order to develop and test an accurate model relating the delivered dose to the PET image, it is necessary to have a reliable Monte Carlo simulation model which can accurately reflect the physics of the interaction of a carbon ion beam with human tissue. Unfortunately, modelling nuclear interactions and the resulting secondary particle production is highly complex, because it involves high-energy interactions of a diverse range of nuclei, for which no fully validated models currently exist. In this work, we present the result of our experimental  quantification of the production of positron-emitting nuclei during carbon ion therapy across a range of homogeneous phantoms (water, polyethylene, poly(methyl methacrylate)) using a high resolution in-beam PET scanner, with a resolution of 1.5 mm. Experimentally estimated yields for each identified positron-emitting fragmentation product are compared with those predicted by a range of physics models available in the Geant4 Monte Carlo Toolkit.

Keywords: Heavy ion therapy, Geant4, Monte Carlo Simulation, in-beam PET, carbon therapy, quality assurance
Poster panel
(face) ID: 368

Poster Number:

GATEv6.2 optical simulation of a pixelated LYSO array coupled with SiPM (#4160)

H. G. Kang1, S. H. Song1, Y. B. Han1, K. M. Kim2, S. J. Hong1, 3

1 Eulji University, Department of Senior Healthcare, Daejeon, Republic of Korea
2 Korea Institute of Radiological & Medical Science (KIRAMS), Molecular Imaging Research Center, Seoul, Republic of Korea
3 Eulji University, Department of Radiological Science, Seongnam, Republic of Korea


The spatial resolution of positron emission tomography (PET) is mainly determined by the cross section of scintillation crystal. When it comes to a crystal with a small cross section (~1.5 mm2), the light guide inserted between the crystal and SiPM plays on an important role for the identification of the individual crystal in 2D flood histogram. The aim of this study is to investigate the effect of light guide thickness on the flood map quality of a pixelated LYSO array coupled to a SiPM using the GATEv6.2 optical simulation. The optical properties of LYSO and SiPM (Hamamatsu, S11064-050P, Japan) such as the emission spectrum, photon detection efficiency (PDE) as a function of wavelength, were implemented into the GATEv6.2 optical simulation. The PET module consists of a 9 × 9 array of pixelated LYSO crystals (1.5×1.5×7.0 mm3), a 1 mm thick light guide, and a SiPM (S11064-050P) which has an array of 4 × 4 channel (each channel has a sensitive area of 3 × 3 mm2).  In order to investigate the effect of light guide thickness on the quality of the flood map, various light guide thicknesses of 1, 2, and 3 mm were used. A plane rectangular source (18 mm × 18 mm) with the energy of 511 keV, was used to irradiate the gamma photon into the LYSO crystal array. The optical simulation output file was analyzed using a custom-made MATLAB code for the generation of flood map. The energy resolution and photo-peak value were calculated after the crystal segmentation using Voronoi diagram. The light collection efficiency and energy resolution were improved as the light guide thickness was increased from 1 mm to 3 mm. However, the flood map quality was deteriorated in terms of the crystal separation. The GATEv6.2 optical simulation could be used to model and optimize the scintillation detector performance. In future, electronic noise of SiPM and preamplifier will be implemented for a more realistic simulation of the pixelated scintillation detector.

Keywords: GATE optical simulation, LYSO, SiPM, Scintillation.