IEEE 2017 NSS/MIC/RTSD ControlCenter

Online Program Overview Session: M-15

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

MIC Poster Session III

Shortcut: M-15
Date: Friday, October 27, 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: 3

Poster Number:

Suitability of a 280 ps-CRT non-DOI detector for the helmet-neck PET (#1149)

S. Takyu1, A. M. Ahmed1, E. Yoshida1, H. Tashima1, T. Yamashita2, T. Yamaya1

1 National Institutes for Quantum and Radiological Science and Technology (QST), National Institute of Radiological Sciences (NIRS), Chiba, Japan
2 ATOX Co., Ltd., Business Development, Tokyo, Japan


We are developing a compact and high-sensitivity brain-dedicated PET scanner, which consists of a hemi-spherically arranged detector unit and an add-on detector unit. Following our first idea of using a chin position as the location of the add-on detector (helmet-chin PET), a neck position has been selected as the location (helmet-neck PET). We have developed prototypes of both geometries using the 4-layered depth-of-interaction (DOI) detector (2.8 mm sized GSOZ scintillators and a 64ch PMT) which does not have time-of-flight (TOF) capability. On the other hand, a new TOF-PET detector module which consists of 4.1 mm sized LFS scintillators and multi pixel photon counters (MPPCs) is now commercially available (C13500-4075LC-12, Hamamatsu Photonics K.K.). Using one-to-one crystal-photodetector coupling, it is possible to achieve a 280 ps coincidence resolving time (CRT), although the Anger-type calculation, which is essential for the 4-layered DOI detector, cannot be applied. In this paper, therefore, we investigated the suitability of this TOF-PET detector module for the next helmet-neck PET. At first, the helmet-neck PET with 20 mm crystal length, which is a standard parameter of the module, was modeled by GEANT4, and performance results were compared with those of the current helmet-neck PET prototype. We observed a compromised image quality; the non-DOI capability counteracted the TOF gain. On the other hand, the helmet-neck PET with shortened crystal length (10 mm) outperformed the current helmet-neck PET prototype; the shortened crystal length reduced the parallax error, and TOF information compensated for the loss of sensitivity. Based on the GEANT4 simulation results, we fabricated a one-pair prototype of the TOF-PET module with 10 mm length, which yielded 11.6% energy resolution at 511 keV and 281.8 ± 9.6 ps CRT at FWHM. In conclusion, we confirmed suitability of the 280ps-CRT non-DOI detector for the next helmet-neck PET.

Keywords: depth-of-interaction, DOI, time-of-flight, TOF, positron emission tomography, coincidence resolving time
Poster panel
(face) ID: 6

Poster Number:

Concept study of Cherenkov radiation based 3D position sensitive detector (#1296)

R. Ota1, R. Yamada1, T. Moriya1, T. Hasegawa2

1 Hamamatsu Photonics K.K., Central Research Laboratory, Hamamatsu, Japan
2 University of Kitasato, School of Allied Health Science, Sagamihara, Japan


Cherenkov photons are emitted promptly when an energetic charged particle goes straight in an optically transparent medium faster than the speed of light. Nowadays Cherenkov radiation is drawing attention to the possibility that detecting prompt Cherenkov photons can remarkably improve timing performance of radiation detectors. Cherenkov radiation is distributed in a continuous spectrum ranging from the ultraviolet through the visible spectrum. Intensity of the emission is inversely proportional to the square of the wavelength. Therefore a Cherenkov radiator should be especially transparent toward the ultraviolet region. A PbF2 crystal is one of the suitable radiators for Cherenkov photons due to its high transparency below 300 nm and its high refractive index. About 30 ps/σ coincidence time resolution (CTR) of two annihilation γ-photons was already reported using 5mm thick PbF2 crystals and Micro-Channel-Plate photomultiplier tubes. It is, however, worthwhile investigating the potential of PbF2 as it doesn’t emit any scintillation photons; this means that we can focus on only the Cherenkov effect. In this report, we propose the concept of Cherenkov radiation-based 3D position sensitive detector using a monolithic PbF2 coupled to one photodetector array. Monte Carlo simulations using GEANT4 were carried out and the detector performances were investigated. The radiator thickness, photodetector readout pitch and single photon timing resolution (SPTR) were parameterized in the simulations. To get CTR less than 100 ps FWHM, SPTR should be approximately better than 40 and 30 ps/σ for 5 and 20 mm thick radiator, respectively. In addition, estimating the interaction position on the photodetector plane using only Cherenkov photons is feasible. Spatial resolution better than 0.4 mm FWHM is achievable with the photodetector array which can independently readout each signal with 100 um interval. Principal component analysis is available to estimate the depth of interaction.

Keywords: Cherenkov radiation, coincidence time resolution, single photon time resolution, PbF2
Poster panel
(face) ID: 9

Poster Number:

Impact of intense X-ray pulses on a NaI(Tl)-based gamma camera (#1478)

W. Koppert1, S. van der Velden1, L. Steenbergen1, H. W. A. M. de Jong1

1 University Medical Centre Utrecht, Clinical Physics/Radiology, Utrecht, Netherlands


Introduction: In SPECT/CT systems, X-ray and γ-ray imaging is performed sequentially. Simultaneous acquisition may have advantages, for instance in interventional settings. However, this may expose the gamma camera to high X-ray doses and deteriorate its functioning. We studied the response of NaI(Tl) crystals to X-ray pulses with a photodiode, PMT and a NaI(Tl) gamma camera, respectively.

Method: First, we exposed a NaI(Tl) crystal to 20 mGy X-ray pulses to study potential afterglow with a photodiode. Next, we exposed a NaI(Tl)-PMT assembly to 10 ms LED pulses (mimicking X-ray pulses) and probed its response with LED flashes (mimicking γ-rays). Similarly, we exposed the assembly to X-ray pulses to investigate the gamma responses in presence of afterglow. Finally, we studied the gamma camera performance in presence of X-ray pulses while exposed to γ-rays.

Results: The photodiode measurements revealed intense afterglow with 55 ms decaytime. The NaI(Tl)-PMT assembly showed disturbed responses shortly after 10 ms LED exposure, indicating diminished PMT functioning. When exposed to X-rays, the assembly response showed elevated baselines with 67 ms decaytime, resulting in biased gamma energies. PMT circuitry modification and afterglow correction succesfully reduced detrimental effects. X-ray exposure of the gamma camera corroborated with the PMT experiments. No events were registered 10-50 ms after X-ray exposure, followed by apparent gamma energy elevations up to 100 ms after exposure. Limiting the dose to 0.02 nGy/pulse, prevents detrimental effects.

Conclusions: Conventional gamma cameras exhibit substantial dead time and misregistration of photon energies when exposed to X-ray pulses of more than 0.02 nGy/pulse. This is due to PMT limitations, but mainly due to afterglow in the crystal. Afterglow effects can be compensated for by correction for its contributions, hence gamma cameras may be modified and larger X-ray doses can be handled without deterioration.

Keywords: Real-time Hybrid Imaging, Gamma Camera, SPECT, PMT, Afterglow, NaI(Tl), X-ray Fluoroscopy, Simultaneous Hybrid Imaging
Poster panel
(face) ID: 12

Poster Number:

Gradient Tree Boosting as a Positioning Method for Monolithic Scintillator Crystals - Initial Results (#1887)

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

1 RWTH Aachen University, Physics of Molecular Imaging Systems, Aachen, Germany


Monolithic scintillator crystals are object of research as an alternative to segmented scintillator arrays in positron emission tomography (PET). We present a new positioning method for incident gamma photons based on the supervised machine learning method of gradient tree boosting (GTB). The method builds a predictive model based on an ensemble of chains of binary decision organized as trees. Parameters, such as the size of the ensemble or the maximum number of decisions for each tree, influence the performance. In particular, the proposed algorithm is able to handle events with missing data. A white wrapped LYSO crystal of size 32 mm x 32 mm x 10 mm is optically coupled to a 64-pixel digital SiPM structured into 16 independently triggering dies. A pinhole collimator with a diameter of 0.5 mm is used in combination with a 22Na source. Calibration and test data are measured on a perpendicular and equidistant grid with a pitch of 0.75 mm at a tile temperature of 5 °C. An opposing LYSO crystal is used to filter out coincident events. After applying a photon cut to remove noise, 90 % of all measured coincidences are taken into account. On average, 8.7 of 16 available dies are read out per event. The gradient boosting models are trained on raw data as well as different combinations of additional features of 500 events for every position. Additional features are e.g. a principal component analysis, the center of gravity or the pixel number with the highest photon count. We calculate the spatial resolution as the full width at half maximum (FWHM), the bias vector and the 90-th percentile radius of the point spread function to quantify the results. For every input combination, a parameter search for the GTB is conducted and evaluated. A spatial resolution of 1.60 mm FWHM and a 90-th percentile radius of 3.54 mm are achieved taking true system condition into account. Further optimizations are in progress and the final status will be presented.

Keywords: monolithic scintillator detector, positron emission tomography (PET), gradient tree boosting, spatial resolution, digital SiPM
Poster panel
(face) ID: 15

Poster Number:

A Method to Measure the Intrinsic Detector Resolution on Monolithic Crystals (#2067)

A. González-Montoro1, G. Cañizares1, P. Bruyndonckx2, A. Aguilar1, E. Lamprou1, S. Iranzo1, R. Marti1, S. Sánchez1, F. Sánchez1, J. M. Benlloch1, A. J. González1

1 Institute for Instrumentation in Molecular Imaging, Valencia, Spain
2 Bruker Skyscan, Kontich, Belgium


The aim of this work is to provide a method to retrieve the intrinsic resolution of PET detector blocks based on monolithic crystals. Estimating the intrinsic detector resolution is typically challenging, this resolution is limited by several factors such as detector size or photon depth of interaction in the scintillator. These factors cannot easily be isolated and studied experimentally, for this reason many works focused on exploring the resolution limits through Monte Carlo simulations or other models.

The proposed method can be done in-situ, and it is validated against a more traditional approach of using bench-top set-up where a detector for software collimation is moved backwards. In our experimental set-up each detector block is composed of a monolithic LYSO crystal almost rectangular (48×48 mm2 entrance, 50×50 mm2 exit) and with a thickness of 10 mm. The entrance face of the detector under study is coupled to a retro-reflector (RR) layer and the exit face to an array of 16×16 SiPMs. A readout providing digitized information of the projected X and Y scintillation light distributions are used.

The proposed method includes both mechanical and software collimation. An empirical equation has been deduced to fit the experimental data in which the detector intrinsic resolution follows a Gaussian distribution whereas the source shape, given the small size of 0.25 mm in diameter, follows a Lorentzian profile. We have alternatively analyzed the measured source profiles for very tiny collimation angles, and the intrinsic detector resolution was deduced using Voigt functions (convolution of Gaussian and Lorentzian profiles).

The experiments resulted on a detector intrinsic spatial resolution better than 0.7 mm FWHM. We found a good agreement between the proposed method and the traditional one (distant detector), including the measured profile analysis. These tests show a method to determine the intrinsic resolution of monolithic-based detector blocks, with high accuracy.

Keywords: Photodetector technology, Scintillators, Monolithic Scintillators, Positron Emission Tomography, Intrinsic detector resolution
Poster panel
(face) ID: 18

Poster Number:

Design and Development of Detector Modules for High-Resolution and High-Sensitivity Brain PET (#2173)

J. Du1, X. Bai1, S. R. Cherry1

1 University of California, Davis, Department of Biomedical Engineering, Davis, California, United States of America


Depth-of-interaction detector modules designed for a high-resolution and high-sensitivity brain PET scanner have been built and evaluated.  The detectors are based on 16 x 16 SiPM arrays coupled to both ends of LYSO arrays with a pitch size of 0.75 mm and a length of 20 or 29 mm. The 16 x 16 SiPM arrays were fabricated by tiling multiple 8 x 8 arrays of through silicon via SiPMs. Each SiPM array has a surface area 35.7 mm x 35.7 mm and each pixel has a 2.0 mm x 2.0 mm active area on a 2.2 mm pitch. Custom front-end electronics using a simple signal multiplexing method was designed to reduce the 512 SiPM signals of each detector module to 32 column signals and 32 row signals using 192 passive comments (resistors and capacitors), and then to nine signals (eight for position information and one for timing information) by applying appropriate weights to each column/row signal. The results show the detector module and the readout electronics worked well and all the crystals in the LYSO array were resolved. Two detector modules are being built for evaluation in coincidence, and the performance of a PET scanner with eight detector rings (32 detector modules per ring) is being simulated.

Keywords: Positron emission tomography, Silicon photomultiplier, Brain PET
Poster panel
(face) ID: 21

Poster Number:

An Evaluation of a Quad-Modality X-ray Luminescence, Fluorescence, Scattering and Transmission Computed Tomography System for Monitoring Nanoparticle-Mediated Therapies (#2263)

J. George1, B. Quigley2, P. La Riviere2, L. - J. Meng1

1 University of Illinois at Urbana-Champaign, Nuclear, Plasma and Radiological Engineering, Urbana, Illinois, United States of America
2 University of Chicago, Department of Radiology, Chicago, Illinois, United States of America


The advent of nanoparticle (NP) therapy has pushed researchers to explore new ways to monitor their circulation in the body as well as their therapeutic efficaciousness. One such treatment method, X-ray-activated photodynamic therapy (X-PDT), shows promise in its ability to deposit therapeutic effects at greater depths than possible with traditional photodynamic therapy. The process that produces the therapeutic effects could also potentially generate X-ray fluorescence and/or X-ray luminescence, which provides the means to monitor the distribution of the NPs and their therapeutic activation respectively. Additionally, the use of X-rays implicitly allows for the potential of X-ray Compton Scattering (XS) to provide information on electron density and incident beam attenuation while conventional CT can give anatomical information. This gives room to the possibility of using a quad-modality X-ray fluorescence (XFCT), X-ray luminescence (XLCT), XS, and CT system for monitoring X-PDT NPs. 

This work experimentally investigates the ability of a XFCT/XS/XLCT/CT quad-modality imaging system for monitoring X-PDT NP’s, such as Y2O3:Eu3+ and LaF3:Tb3+, through the use of a geometry, which mimics that of the X-ray microbeam therapy environment. In such an environment, the intense highly collimated incident X-ray beam would produce the fluorescent and luminescent yields needed for XFCT and XLCT imaging. This work will specifically (1) provide 3D quad-modal images of a phantom containing X-PDT-containing agents, (2) examine the sensitivities and resolutions of the of each of the modalities and for each agent of interest, (3) investigate any the ways the modalities can supplement each other in image production.

Keywords: X-ray, Fluorescence, XFCT, Luminescence, Nanoparticle
Poster panel
(face) ID: 24

Poster Number:

TOF and DOI capable SiPM-PET detector using striplines that provide row and column based measurements (#2417)

X. Lyu1, F. Xu1, C. - M. Kao3, C. Tian2, J. Lyu2, Y. Hua2, Y. Wang1, H. Kim3, Q. Xie2

1 Huazhong University of Science and Technology, School of Optical and Electronic Information, Wuhan, China
2 Huazhong University of Science and Technology, Department of Biomedical Engineering, Wuhan, China
3 University of Chicago, Department of Radiology, IL, United States of America


We propose a highly multiplexing electronic readout for developing high-resolution, TOF and DOI capable PET detectors based on SiPMs. The detector will employ one-to-one scintillator-to-SiPM coupling and double-ended light readout for DOI measurement. The front and rear SiPM arrays are electronically read by using a single stripline (SL) for identifying the row and column numbers respectively. In the SL readout, which has been successfully demonstrated by us, multiple SiPMs are connected to an SL and the signal-generating SiPM is identified based on the arrival time difference of the signals at two ends of the SL. In this work, we modify the SL readout to identify the row or column number. The pixel in the 2d detector array is then uniquely determined by the row and column number identified by the front and rear SiPM arrays. We prototype such SL boards for use with SensL's 6x6 SiPM arrays (4.2 mm pitch) coupled to a 6x6 LYSO array (4.2 mm pitch, 20 mm thickness). The 72 SiPMs of this detector are read by using 2 SLs to yield 4 outputs only. Our initial results show that accurate pixel identification can be achieved. Due to poor light coupling, we currently obtain pixel-level energy resolution ranging from 20-50% while we expect ~15% based on our previous work. We have not measured the TOF and DOI resolutions, but based on previously reported results we expect ~300 ps TOF resolution and 3-5 mm DOI resolution. We will improve light coupling and report full experimental results at the meeting. This readout method can be applied to larger detector arrays. Recently, we are able to read 32 SiPMs on an SL while attaining ~300 ps TOF resolution. Based on this, it is possible to read a 32x32-pixel detector by using four acquisition channels. Thus, the proposed method is very promising for developing large-area, high-performance PET detectors at reduced cost.

Keywords: PET Detector, TOF and DOI capable, Strip-line, multiplexing readout
Poster panel
(face) ID: 27

Poster Number:

Timing Performance of A DOI Detector Using Crystal Bars with Subsurface Laser Engraving (#2537)

A. Mohammadi1, S. Takyu1, E. Yoshida1, F. Nishikido1, K. Shimizu2, T. Sakai2, T. Yamaya1

1 National Institute of Radiological Sciences, Department of Radiation Measurement and Dose Assessment, Chiba, Chiba, Japan
2 Hamamatsu Photonics K.K., Hamamatsu, Shizuoka, Japan


The quality and uniformity of positron emission tomography (PET) images can be improved significantly by time-of-flight (TOF) and depth-of-interaction (DOI) capabilities. We have already developed a DOI dual-ended readout detector using crystal bars segmented by applying a subsurface laser engraving (SSLE) technique. The DOI was determined by the ratio of detected light between two readouts using Anger calculations. In this study, we investigated the influence of the number of DOI segments on timing performance of a single-readout detector composed of a lutetium fine silicate (LFS) single crystal bar (3×3×20 mm3) having several DOI segments, made by applying the SSLE technique, and a Hamamatsu TOF PET module. The coincidence resolving time (CRT) was evaluated by facing the detector to another Hamamatsu TOF PET module coupled with the same LFS crystal bar as the first module. FWHM of the CRT of each crystal increased with the increased number of DOI segments and the best FWHM of the CRT of 256 ps was obtained for the crystal without DOI segment. The CRT for each segment of the crystal bars was obtained from irradiation of that segment and the CRT of the segment coupled to the MPPC was improved 10% compared with the top segment of the crystal. The CRT for individual segments were less than 300 ps for the detector composed of a crystal bar with four DOI segments. The improvement of timing performance of the detector with a dual-ended readout will be presented in the conference.

Keywords: timing performance, crystal bars, segmented crystals by SSLE, time-of-flight, depth-of-interaction, PET
Poster panel
(face) ID: 30

Poster Number:

Double-photon emission imaging of 111-In with a high-resolution Si/CdTe Compton camera (#2691)

S. Takeda1, T. Orita1, F. Moriyama1, H. Sugawara1, G. Yabu2, 3, H. Yoneda2, 3, S. Watanabe2, 3, T. Takahashi2, 3, Y. Mizumachi4, M. Uenomachi4, Y. Yoshihara4, H. Takahashi4, K. Shimazoe4

1 Okinawa Institute of Science and Technology Graduate University, Advanced Medical Instrumentation Unit, Onna-son, Japan
2 Japan Aerospace Exploration Agency, Institute of Space and Astronautical Science, Sagamihara, Japan
3 The University of Tokyo, Dept. of Physics., Tokyo, Japan
4 The University of Tokyo, Dept. of Nuclear Eng. and Management, Tokyo, Japan


We report on the first experimental results of double-photon emission imaging of a well-known and medical use 111-In radioisotope with the high-resolution Si/CdTe Compton cameras. The double-photon emission imaging has been expected to be a powerful method to address the major drawback of Compton imaging on 3-dimensional image space, that is, relatively large intersection of a surface of a cone (probability), derived by solving a Compton equation, with image boxels. If the cascade gamma-rays from single isotope decay are detected by Compton cameras, the intersection of probability drastically shrinks by means of ANDing multiple conical surfaces. However, experimental studies with radioisotopes commonly used in medical imaging (125-I, 111-In, 99m-Tc etc.) are still not reported, because there is no suitable candidate that emits cascade gamma-rays in an energy band of normal Compton cameras, typically above 300 keV up to MeV. The ISAS/JAXA group has developed a novel Compton camera which consists of high-resolution Si and CdTe semiconductor detectors for the next generation of gamma-ray astronomy. Original technologies of low-noise silicon devices, high resolution schottky CdTe diodes and dedicated low-noise analog ASICs allow tracking of low energy gamma-ray down to 60 keV. This capability, for the first time, enables us to perform the double-photon emission imaging of 111-In which emits cascade gamma-rays of 171 keV and 245 keV. In our experimental setup, two Si/CdTe camera modules were placed to oppose to each other. The cascade gamma-rays stemming from 111-In decay were clearly detected thanks to good energy resolution of Si and CdTe detectors. We confirmed drastic improvement of a point spread function by the double-photon reconstruction when compared to a traditional single-photon reconstruction, and then obtained the high-contrast image.

Keywords: Compton camera, CdTe, gamma-ray imaging
Poster panel
(face) ID: 33

Poster Number:

Development of X-ray Talbot-Lau imaging scanner using the large area Tb:GdAlO3/Al2O3 eutectic scintillator plate (#2947)

K. Kamada1, 2, H. Yamaguchi1, A. Yamaji1, S. Kurosawa1, Y. Shoji2, 1, Y. Yokota1, Y. Ohashi1, A. Yoshikawa1, 2

1 Tohoku University, Sendai, Please Select, Japan
2 C&A corporation, Sendai, Miyagi, Japan


X-ray Talbot-Lau interferometry has been developing for the last decade because of its attractive potential for a quantitative X-ray phase imaging techniques . X-ray phase imaging provides three images such absorption, differential-phase, and visibility-contrast images. This future causes to higher resolution density variations in the sample than that conventional absorption-contrast X-ray imaging. Medical and biological imaging is the main target of X-ray phase imaging, and several trials using synchrotron radiation sources and laboratory sources have been made. In X-ray Talbot-Lau interferometry, single or mulch absorption gratings between a sample and X-ray generate detector differential-phase, and visibility-contrast images. The absorption gratings generate Moire fringes and differential-phase, and visibility-contrast images are obtained from analysis of the spatial frequency of the Moire fringes. However absorption gratings absorbed the transmitted X-ray and sensitivity is degraded. So exposure dose can become a problem in medical and biological imaging.

In this study, large area growth technique of Tb doped GAP/ a-Al2O3 eutectic scintillator using the micro pulling down (m-PD) method was developed for designing direct X-ray phase imaging. Furthermore proto-type of imaging system was developed and X-ray phase imaging using the eutectic scintillator was evaluated.

Keywords: X-ray Talbot-Lau imaging, Scintillator, eutectic, CMOS
Poster panel
(face) ID: 36

Poster Number:

Extension of Bipolar Multiplexing Circuit to Achieve High Channel Reduction Ratio (64:2) (#2998)

Y. K. Kim1, Y. Choi1, K. B. Kim1, H. T. Leem1

1 sogang University, Electronic Engineering, Seoul, Seoul, Republic of Korea


We have reported the feasibility of novel multiplexing concept using bipolar pulse to improve the drawbacks of the conventional multiplexing methods suffered various problems of reduced SNR, position distortion, and complexity for PET system development. The proposed multiplexing method generates different bipolar pulses having different zero crossing points by adjusting the time constant values of high pass filter, as a function of channel position. It would allow to identify channel position by estimating the width difference between rising edge and zero crossing point of the bipolar pulse. The goal of this study was to extend the bipolar pulse based multiplexing front-end circuit and DAQ system for the development of a gamma imaging detector. The detector module was consisted of 2 x 2 detector block each of which was composed of a 4 × 4 matrix of 3 × 3 × 20 mm3 LYSO and a 4 × 4 array GAPD. The energy and timing information were measured using the sum signal of cathode output. The performance of PET detector using 64 channel multiplexing circuit was evaluated by measuring the energy resolution and time resolution and the values were 14% and 620 ps, respectively. The channel position was accurately identified for all measured channels. This study demonstrated that the bipolar multiplexing method could be a simple and effective approach to reduce the number of readout channels and to accurately identify pixel position for the development of gamma imaging system. In further study, coincidence timing resolution will be improved by optimizing the rise time of cathode pulse and by using FPGA-based TDC. Also, we will apply the proposed bipolar multiplexing method to develop a proof-of-principle PET.

Keywords: multiplexing, bipolar pulse, PET, gamma imaging
Poster panel
(face) ID: 39

Poster Number:

Calibration of Monolithic PET Detector Using Uncollimated Source (#3174)

P. Fan1, 2, Y. Liu1, 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


Monolithic PET detector design is promising for low cost, high packing fraction as well as intrinsic DOI capability compared to the pixelated detector. The key to gamma interaction position estimation of monolithic detector is to calibrate the light response function (LRF). Conventionally LRF is obtained with a tedious and time-consuming process by acquiring training datasets of gamma events with known interaction positions using collimated sources attached to robotic stages. In this work, we propose a calibration method to obtain LRFs for monolithic detector using uncollimated sources. With uncollimated sources, the gamma interaction position distribution (IPD) of all the events can be accurately calculated. The basic idea of the method is to find the appropriate LRF to match the IPD calculated from estimated LRF and the true IPD. LRF was iteratively estimated by minimizing the discrepancy between the two IPDs with a gradient descent algorithm. Monte Carlo simulation was performed to evaluate the feasibility of the calibration method. A training dataset using an uncollimated source was simulated to estimate the LRF and detector positioning performance was evaluated with both estimated and true LRFs. Detector positioning performance with the estimated LRF was slightly worse but comparable with that obtained with the true LRF. In x/y directions, the mean positioning resolution and absolute bias were respectively ~2 mm and 0.25 mm with estimated LRF and 1.61 mm and 0.16 mm with the true LRF. In z direction, the average misclassification rate of all the events was 27% with estimated LRF and 18% with the true LRF. Based on the simulation validation results, we believe that the proposed calibration method using uncollimated sources is accurate and feasible for monolithic detector design and could greatly simplify the calibration process of monolithic detector and potentially facilitate its application in real PET systems. Further experimental validation study is ongoing.

Keywords: Monolithic detector, Uncollimated source, Detector calibration, PET, DOI
Poster panel
(face) ID: 42

Poster Number:

Simultaneous MR-Guided Adaptive SPECT System based on Soft Robotics: An Experimental Feasibility Study (#3245)

E. M. Zannoni1, L. Zhou2, 3, Y. Zhang2, 3, L. - J. Meng2, 3

1 Universiy of Illinois at Urbana-Champaign, Bioengineering, Urbana, Illinois, United States of America
2 Universiy of Illinois at Urbana-Champaign, Nuclear, Plasma and Radiological Engineering, Urbana, Illinois, United States of America
3 Universiy of Illinois at Urbana-Champaign, Beckman Institute for Advanced Science and Technology, Urbana, Illinois, United States of America


Soft-robotics is an emerging field in which robots are typically constructed with soft deformable structures and their motion is driven and controlled by compressed air and liquid. Soft-robotics offers several attractive features, such as being light-weight and metal-free, assuring a safer interaction with the biological sample under examination and offering hyper-redundant structures with an almost unlimited number of degrees of freedom.

In this study, we specifically focus on the use of soft-robotics for positioning gamma camera modules inside MR scanners for simultaneous MR-guided adaptive SPECT imaging applications.

An adaptive SPECT imaging system that could be self-optimized in real time according to high-quality MR information and SPECT data being collected would allow an optimum SPECT imaging capability to focus target areas and critical organ under examination. The soft-robotics offers a unique and critical capability to facilitate such imaging system and to develop optimized geometry acquisition schemes. These schemes can dramatically improve imaging of organs and tissues with relatively low specific uptake, imaging of tracers that cannot reach high labeling efficiency, or imaging of fast dynamic processes.

The major objectives of this work are: (a) to propose the first prototype of MR-guided adaptive SPECT imaging system based on MR-compatible detector modules mounted on soft-robotic arms; (b) to perform experimental studies using phantoms filled with clinically relevant SPECT radioisotopes, such as I125 or Tc99m; (c) to quantify the potential benefit of the adaptive approach over the traditional stationary and not optimized architecture.

In the experimental studies, we experimentally simulate various acquisition geometries that could be realized with a full-ring adaptive SPECT system equipped with soft-robotic arms. We will discuss both the experimental and Monte Carlo results and how MRI-guided SPECT imaging can benefit from this innovative approach.

Keywords: soft-robotics, SPECT/MRI, adaptive imaging
Poster panel
(face) ID: 45

Poster Number:

First Results from CMOS-Integrated SPAD and SiPM (#3602)

S. Gnecchi1, D. J. Herbert1, C. J. Jackson1

1 SensL Technologies, Cork, Ireland


Time-of-Flight Positron Emissions Tomography (ToF-PET) systems seek to optimize the coincidence resolving time (CRT) between a pair of radiation detection elements in order to provide the best image quality. The CRT is dependent on many factors, but in recent years focus has been on maximizing the PDE (Photon detection efficiency) of the photodetectors. For the SiPM, the currently available PDE is close to its theoretical maximum and further development are unlikely to yield significant system level ToF-PET CRT improvements. SensL believe that future progress in the field of ToF-PET will be driven by improved system integration, incorporating active CMOS elements on the sensor chip to improve SiPM timing, particularly for larger area sensors. The best performing scintillators for ToF-PET emit in the blue portion of the spectrum and the challenge is to achieve CMOS integration while retaining sensitivity to blue photons, in line with SensL’s existing C-Series and J-Series products. To provide the best sensor performance possible, SensL have developed a process using dedicated layers for the SPAD and SiPM microcells. This process uses 0.35um CMOS technology and has been used to create a number of test chips that feature integrated analog test circuits, with P-on-N SPAD and SiPM sensors. A 14x14 microcell SiPM array has been produced with microcells of 10um, 20um and 35um to test basic functionality. We will describe the process development, the test circuits realized and present full characterization of the CMOS integrated SPAD and SiPM sensors, including dark count rates, PDE and timing characterization. Continuing development of the process aims to incorporate further functionality, including digital components such as Time to Digital Converters (TDC) 

Keywords: SiPM, SPAD, CMOS, ToF-PET
Poster panel
(face) ID: 48

Poster Number:

Scintillation Design Study for Prompt Gamma Photon Detection in Proton Therapy Monitoring (#3688)

Z. Hongyang1, 2, F. Peng1, 2, X. Tianpeng1, 2, T. Ma1, 2, L. Yaqiang1, 2, W. Zhaoxia1, 2

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


Proton therapy is currently one of the most advanced tumor radiotherapy technologies. Due to factors that introduce range error of proton beams, proton range monitoring becomes significant for proton therapy. In this work, a scintillation detector design consisting of BGO block (size: 3.5ⅹ3.5mm2) coupled to an 8ⅹ8 SiPM array (Sensl FJ30035)is proposed to realize online range monitoring based on prompt gamma imaging which is a popular range monitoring method with high space resolution.

However, because of the 2~8MeV high energy of prompt gamma, saturation effects of SiPM should be carefully evaluated. SiPM saturation effect when coupled with either BGO or LYSO was estimated by a numerical simulation, and BGO crystal was selected due to reduced SiPM saturation effect. Length of BGO crystal was 30mm providing an average detector efficiency of 55% on 2~8Mev prompt gamma. In order to avoid SiPM saturation and reduce cost of the detector, three designs of detector, namely coupling 8ⅹ8 (detector A), 12ⅹ12 (detector B) and 16ⅹ16 (detector C) BGO blocks with 8ⅹ8 SiPM array, were proposed and tested under 22Na source.

Detector A obtains the best positioning performance and energy resolution with the highest cost. Detector B provides a reasonable positioning performance, an acceptable energy resolution and, significantly, a relatively low price. But there are still defects at corners of detector B’s 2D map which demands further study. Despite the fact that there is blur on the 2D map of detector C, the question remains whether it works on high-energy prompt gamma.

In conclusion, the scintillation detector consist of an 8ⅹ8 FJ30035 SiPM array and a 12ⅹ12 BGO block (size: 3.5ⅹ3.5ⅹ30mm3) is a reliable and economical design through synthetical consideration and can be further used for the prompt gamma imaging system.

Keywords: scintillation detector, prompt gamma, BGO, proton therapy
Poster panel
(face) ID: 51

Poster Number:

Improving Crystal Identification in Flood Maps Using Communication Between Adjacent SiPM Arrays (#3723)

P. Peng1, J. Du1, X. Bai1, E. Berg1, S. R. Cherry1

1 University of California, Davis, Department of Biomedical Engineering, Davis, California, United States of America


Detectors used in PET are often modular and use light sharing to decode a smaller pixel crystal array with larger pixel photodetectors. However, this often leads to problems in resolving the edge crystals due to the truncation of light transport at the edges, and typically requires a custom light guide or variable light sharing to resolve all crystals in a flood map. This work describes a method to improve flood map crystal identification based on electronic communication between adjacent SiPM arrays in a large-area detector module. The detector used here consisted of a 14 x 56 array of 0.89 x 0.89 x 20 mm3 LYSO crystals (0.96 mm pitch). Each crystal was polished on all faces, and were optically separated by Toray diffusive film. The scintillation light from both ends of the crystal array was measured by two SensL ArrayJ-30035-16P SiPM arrays for dual-ended DOI estimation. All measurements were conducted in a temperature controlled dark box, with T = 25.0 ± 0.5 oC and with an SiPM overvoltage of Vover = 3.00 ± 0.02 V. By adding the signals of adjacent SiPM pixels in the row-column decoding algorithm, crystals on the edges were clearly resolved. The improvement in flood map quality was quantified as the fraction of events positioned to the correct crystal based on a 2D Gaussian fit of the segmented flood maps. The trigger sharing method developed here improved the flood map quality metric from 41.3% to 83.9%.  The energy resolution across all crystals varied from 13.0% to 27.5% with a mean of 16.7 ± 0.2%. The FWHM DOI resolution varied from 1.71 mm at both ends of the crystal, to 4.79 mm at the center of the crystal, with an average DOI resolution of 3.22 ± 0.40 mm. The timing resolution in coincidence with a PMT detector, and using leading edge discrimination was 2.79 ± 0.04 ns. Using this method, both the light collection efficiency and the energy resolution were improved.  This approach appears promising for a high resolution small-animal or brain PET scanner.

Keywords: positron emission tomography (PET), silicon photomultiplier(SiPM), light sharing, flood map quality
Poster panel
(face) ID: 54

Poster Number:

Experimental assessments of the timing performances of detectors constructed with LaBr3, CeBr3, LFS, LSO, LYSO, GAGG scintillators (#4134)

Q. Peng3, S. Xie1, M. Yang1, T. Sui1, 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, California, United States of America


The purpose of this study is to compare the timing performances of popular scintillators for time-of-flight (TOF) PET. Six types of scintillators, including LaBr3, CeBr3, LFS, LSO, LYSO (Ca doped) and GAGG, are fabricated in different shapes and different surface finishes, and constructed into detectors. The coincidence time resolution (CRT) of those detectors are measured.    

Findings and collusions in the studies are  (1) We have measured a 67.6ps FWHM CTR with a pair of detectors constructed with 6 mm LaBr3 cubes; (2) We have measured 88.1ps FWHM CTR with a pair of detectors constructed with 4 mm LYSO: Ce, Ca pyramids; (3) The two SiPMs from SensL and AdvanSiD have no significant differences in the timing performances; (4) the crystals with rough surfaces have a better timing than those with polished surfaces; (5)   LYSO: Ce, Ca from different venders and LSO from Siemens have similar performances; (6) 0.2%Ca co-doped LYSOs have a better CTR than 0.1% co-doped LYSO; (7) LYSO: Ce, Ca has a better timing performance than LFS; (8) The CRT measurements are related to the size of the crystals. 2 mm crystals has a better timing than 4 mm crystals. 4 mm crystals has a better timing than 6 mm crystals. (8) Timing performance can be improved by using a geometry such as a pyramid shape which can improve the light transportation in the crystals. (9) The CRT of 2 mm GAGG cubes is 193 ps. But note that the SensL’s SiPM has a much lower photo detection efficiency (PDE) @ 520nm compared to that @ 420nm. Thus, the measured light yield from GAGG is 47% lower than that of LYSO. The CTR of the 2mm GAGG cubes improves to 117ps, if the effect of the low PDE @ 520nm is compensated.

Keywords: Time Of Flight, Scintillator, LaBr3, CeBr3, LFS, LSO, LYSO (Ca doped), GAGG
Poster panel
(face) ID: 57

Poster Number:

Crystal identification for a dual-layer-offset LYSO based PET system via Lu-176 background (#1225)

Q. Wei1, T. Dai2, T. Ma3, T. Xu3, Y. Gu1, Y. Liu3

1 University of Science and Technology Beijing, Department of Automatic and Electrical Engineering, Beijing, China
2 China-Japan Friendship Hospital, Department of Radiation Oncology, Beijing, China
3 Tsinghua University, Department of Engineering Physics, Beijing, China


Crystal identification is a critical procedure for PET systems. In this paper, we proposed a crystal identification method for a dual-layer-offset LYSO based animal PET system via Lu-176 background radiation. Single event data of Lutetium background radiation were acquired in list-mode for one hour to generate a single event flood histogram (SFH). Coincidence events were retrieved from the same data using time information to generate a coincidence flood histogram (CFH). The coincidence flood map was employed to identify the peaks of the inner layer responses using average peaks calculation and mean-shift algorithm. The response of the inner layer was removed from the SFH using the CFH and then the peaks of the outer layer were also identified using mean-shift algorithm. At last, the crystal position map was generated using these peaks with the distance criteria. Results show that the proposed method can be employed for the dual-layer offset PET system to implement crystal identification instead of using external radiation sources.

Keywords: PET, crystal identification, Lutetium background radiation, coincidence
Poster panel
(face) ID: 60

Poster Number:

Development of Depth Encoding Small Animal PET Detectors by Using Dual-ended Readout of Pixelated Scintillator Arrays with SiPMs (#1371)

Z. Kuang1, X. Wang1, X. Fu1, N. Ren1, X. Deng1, K. Feng1, X. Zhang1, Y. Zheng1, Q. Yang1, Z. Hu1, Y. Yang1

1 Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China


LYSO arrays with 11×11 crystals and 11.6×11.6×20 mm3 outside dimension were made of ESR, Toray and BaSO4 reflectors. The LYSO arrays were read out with Hamamatsu SiPM arrays from both ends. The SiPM array is 4×4, with a pixel size of 3×3 mm2 and 0.2 mm gap in between. The performance of the detector was measured. The BaSO4 arrays provide the best and the ESR array provides the worst flood histograms. All crystals can be clearly resolved for all four arrays. The DOI resolution obtained from the DOI profiles of the individual crystals of the four array is from 2.1 to 2.35 mm for events with E>350 keV. The DOI ratio variation among crystals is bigger for the BaSO4 arrays as compared to the ESR and Toray arrays. The photopeak amplitude of the Toray array changes the most with depth; it provides the worse energy resolution of 21.3%. The photopeak amplitude of the BaSO4 array with a thin reflector thickness of 80 mm almost doesn’t change with depth; it provides the best energy resolution of 12.9%. Both the ESR and Toray arrays provide better packing fraction than BaSO4 arrays due to the thinner reflector thickness. Those detectors can be used to build a small animal PET scanner to simultaneously achieve uniform high spatial resolution and high sensitivity.

Keywords: depth of interaction, small animal PET, dual-ended readout, SiPM
Poster panel
(face) ID: 63

Poster Number:

Development of a Preclinical PET System Based on Pixelated LYSO Crystals and SiPM Arrays (#1800)

N. Zeraatkar1, S. Sajedi1, S. Kaviani1, M. Taheri1, H. Khanmohammadi1, S. Sarkar2, 3, M. R. Ay1, 2

1 Tehran University of Medical Sciences, Cellular & Molecular Imaging Research Center, Tehran, Iran (Islamic Republic of)
2 Tehran University of Medical Sciences, Department of Medical Physics & Biomedical Engineering, Tehran, Iran (Islamic Republic of)
3 Tehran University of Medical Sciences, Research Center for Science & Technology in Medicine, Tehran, Iran (Islamic Republic of)


PET imaging is playing an important role in the field of molecular imaging regarding its high sensitivity in detection of radiotracers. Hence, development of dedicated preclinical PET scanners with better performance parameters utilizing state-of-the-art technology has an ongoing trend. We recently developed a preclinical PET scanner based on pixelated lutetium-yttrium oxyorthosilicate (LYSO) crystals and Silicon Photomultiplier (SiPM) arrays. In this paper, we report on the system design properties together with calibration/correction algorithms, image reconstruction, and preliminary evaluation of performance parameters and animal imaging. The detection subsystem consists of 10 blocks forming a full-ring bore. Each block is composed of attachment of a 24x24 LYSO crystal array to a 12x12 SiPM array. After pre-processing, position extraction, and time-stamping in each block, a mainboard is responsible for acquiring the data from all the blocks, coincidence detection, and transferring them to a computer where other processing steps are performed. The initial assessments showed that the spatial resolution of the system at the center of field-of-view is about 1.8 mm while the sensitivity is around 1.4%. Also, reconstructed images of a rat scan after administration of 7.4 MBq 18F-sodium fluoride (NaF) revealed acceptable performance of the scanner in imaging of animal models. Time resolution and average energy resolution of the system after corresponding corrections were measured as 1.2 ns and 17.5%, respectively. The system demonstrated comparable performance in contrast to similar preclinical PET scanners. However, we aim to further assess the performance of the system using NEMA protocols to be able to provide a more comprehensive comparison with the other PET systems.

Keywords: PET, small-animal, preclinical, SiPM, LYSO
Poster panel
(face) ID: 66

Poster Number:

Imaging performance of the MOLECUBES β-CUBE – a dedicated small animal PET scanner using monolithic LYSO scintillator (#2426)

S. Krishnamoorthy1, E. Blankemeyer1, P. Mollet3, S. Surti1, R. Van Holen3, J. S. Karp1, 2

1 University of Pennsylvania, Dept of Radiology, Philadelphia, United States of America
2 University of Pennsylvania, Dept of Physics & Astronomy, Philadelphia, United States of America
3 MOLECUBES, Ghent, Belgium


The MOLECUBES β-CUBE scanner is the newest amongst commercially available PET scanners for rodent imaging. The scanner is lightweight, modular and utilizes a small footprint, enabling bench-top imaging. Its PET detector comprises of an 8-mm thick monolithic LYSO scintillator read-out by a 25 mm x 25 mm array of 3 mm x 3 mm Hamamatsu silicon photomultipliers. The monolithic scintillator not only improves scintillator packing fraction, but also enables continuous depth-of-interaction measurement that permits development of such a compact PET scanner. Arranged in 5 rings of 9 detectors, the scanner diameter of 8 cm and axial length of 13 cm make it suitable for imaging both whole body mice and rats. Initial scanner characterization demonstrates the scanner to have good sensitivity and excellent volumetric spatial resolution. NEMA point-source measurement provides a spatial resolution of 970 μm (785 μm with MLEM) and sensitivity of 12% (255 – 765 keV) at the center of the scanner, and 1.14 mm (930 μm with MLEM) at 30 mm away from the center of the FOV. In combination with the X-CUBE – a dedicated high-resolution small animal CT, it offers quantitative imaging with spatially co-registered fully 3D PET-CT images. Animal imaging studies demonstrating the quantitative imaging performance and NU-4 characterization of the PET scanner will be presented.

Keywords: small animal PET scanner, NU-4, high spatial resolution, MOLECUBES, β-CUBE PET scanner, Depth of interaction, DOI
Poster panel
(face) ID: 69

Poster Number:

Intercrystal Scatter Studies for a Clinical PET System with 1mm3 Resolution 3D Position Sensitive Scintillation Detectors (#2477)

D. F. - C. Hsu1, D. L. Freese1, D. R. Innes2, C. S. Levin1, 2

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


We are constructing the world’s first 1 mm3 resolution clinical PET system dedicated to locoregional imaging (e.g. breast, head/neck), which, when completed, will comprise two panels, each with sensitive area of 16x10 cm2. The edge-on layout of densely-packed 8x8 arrays of 0.9x0.9x1 mm3 LYSO crystal elements [two arrays in depth coupled to two position sensitive avalanche photodiodes (PSAPDs)] provides high sensitivity and the capability to position the individual 3-D coordinates of intercrystal scatter (ICS) interactions, which are events where the annihilation photon scatters from one scintillation crystal element to another, depositing a portion of its energy in each. In this work, we evaluate the capability of our system to detect ICS interaction pairs, which are defined as two ICS interactions originating from the same annihilation photon, and measure the contributions from these ICS events to coincidence timing resolution (CTR) and effective photon sensitivity. A 200uCi Na-22 line source was scanned for 20 minutes with a global hardware energy threshold (over 1536 PSAPDs/98,304 crystal elements) of 50 keV, and both energy and timing calibration were applied to this dataset. Prior to coincidence sorting, ICS interaction pairs can be seen in the energy histograms, while the measured time difference and average distance between ICS interactions over a subset of 672k events are 58.2 ns and 20.0 mm, respectively. Coincidence sorting was then performed with a timing window of +/- 100 ns, first limited only to photoelectric (PE) events (one interaction in one crystal element in one LYSO array that falls in the energy window), and then expanded to include ICS interaction pairs. The measured CTR for PE-PE/PE-ICS/ICS-ICS coincident events are 16.8/25.9/31.1 ns FWHM, respectively. By expanding coincidence sorting to include ICS events, the number of coincidence LORs increased from 4,621,873 to 5,835,381, corresponding to a 26.3% increase in effective photon sensitivity.

Keywords: PET, High-Resolution, Inter-crystal scatter, PSAPD, Locoregional, Limited-angle tomography
Poster panel
(face) ID: 72

Poster Number:

Fabrication and Preliminary Evaluation of Detectors for PET-Laparoscope System (#2662)

M. R. Liyanaarachchi1, K. Shimazoe1, H. Takahashi1, 2, E. Kobayashi3, I. Sakuma1, 3

1 The University of Tokyo, Department of Bioengineering, Bunkyo, Tokyo, Japan
2 The University of Tokyo, Department of Nuclear Engineering and Management, Bunkyo, Tokyo, Japan
3 The University of Tokyo, Department of Precision Engineering, Bunkyo, Tokyo, Japan


Positron Emission Tomography (PET) is used in preoperative evaluation of gastric cancer surgery and it could identify lymph nodes metastasis. But it is difficult to locate those lymph nodes metastasis during the surgery and this causes unnecessary tissue removal. The Quality of Life (QOL) of patients can be improved if intraoperative identification of lymph nodes metastasis is possible. PET-Laparoscope system consisting of an external fixed detector and a movable detector probe which can be inserted into the patient’s stomach has been proposed for intraoperative identification of lymph nodes metastasis during gastric cancer surgery. This paper presents the design of the first prototype of the PET-Laparoscope system and experimental results obtained from the detector system. GAGG scintillating crystals coupled with SiPMs (Silicon photomultipliers) are used in the detectors to identify 511 keV gamma photons. All the crystals in the fixed detector and the movable detector produced narrow photopeaks so that 511 keV photons can be well identified.

Keywords: Gastric cancer, Intraoperative, Laparoscopic surgery, Lymph nodes metastasis, Positron Emission Tomography
Poster panel
(face) ID: 75

Poster Number:

easyPET: characterization of a concept prototype following the NEMA NU 4-2008 standards (#2822)

P. M. M. Correia1, J. Menoita1, A. L. M. Silva1, N. O. Romanyshyn1, I. F. C. Castro2, F. D. S. Alves2, P. M. M. C. C. Encarnação2, F. Rodrigues2, A. C. Santos3, C. Ramos3, F. Caramelo3, N. C. Ferreira3, 4, D. A. Sá3, 4, N. Matela5, P. Almeida5, P. M. Sá5, J. F. C. A. Veloso1

1 i3N - Departamento de Física da Universidade de Aveiro, Aveiro, Portugal
2 RI-TE - Radiation Imaging Technologies, Lda, IEUA Edifício 1, Aveiro, Portugal
3 IBILI/FMUC - Instituto de Biofisica/Biomatemática, Faculdade de Medicina da Universidade de Coimbra, Coimbra, Portugal
4 ICNAS, Universidade de Coimbra, Coimbra, Portugal
5 Instituto Biofísica e Engenharia Biomédica, Universidade de Lisboa, Lisboa, Portugal


EasyPET is a new concept of a PET scanner using an innovative acquisition method based on two rotation axes for the movement of detector pairs. The concept allows achieving high and uniform position resolution over the whole field of view (FoV), by eliminating parallax errors due to the depth of interaction (DoI), which are typical of ring based PET systems. The absence of DoI effect, does not impose limitations on the proximity of the detector elements to the FoV and therefore favours the sensitivity of the system. Full axial imaging is possible with easyPET using only a small number of detector elements. A small concept demonstrator for 3D imaging with 16 + 16 detector cells was built, based on LYSO scintillators coupled to SiPMs, covering a 50 mm diameter × 34 mm long field of view. The technology used in the detector cells allows operation in strong magnetic fields, in the Tesla scale region. A comparison between simulation results obtained with GATE and experimental results will be presented in terms of system performance, namely position resolution and sensitivity, following NEMA NU 4-2008 standards

Keywords: PET, SiPM, DOI, pre-clinical
Poster panel
(face) ID: 78

Poster Number:

Time Calibration of phenoPET based on the Lu-176 Background of LYSO (#3181)

M. Streun2, K. Borggrewe2, A. Chlubek2, D. Durini2, A. Erven2, C. Hinz2, L. Jokhovets2, R. Metzner2, H. Nöldgen2, D. Pflugfelder2, J. J. Scheins2, S. Völkel2, S. Jahnke2, U. Schurr2, S. van Waasen2, 1

1 University Duisburg-Essen, Duisburg, Germany
2 Forschungszentrum Jülich GmbH, Jülich, North Rhine-Westphalia, Germany


The digitally generated timemarks from more than 2000 individual detector elements in the phenoPET scanner show individual time offsets. These offsets (or skew) are in the range of +/‑1 ns while the actual time resolution of each element is below 250 ps. Thus, the time resolution of the entire system will be highly improved by correcting these offsets. This paper presents a precise correction method based on measurements of the LYSO background radiation. The Coincidence Resolving Time for the entire scanner could be improved from 1.45 ns to less than 300 ps.

Keywords: dSiPM, Lu176, DPC, Skew, CRT
Poster panel
(face) ID: 81

Poster Number:

Development of Quantitative Approaches for Behavioral Neuroimaging in Birds (#3407)

M. Salerno1, E. Ferrer2, S. Wei1, A. Balanoff2, P. Vaska1

1 Stony Brook University, Department of Biomedical Engineering, Stony Brook, New York, United States of America
2 American Museum of Natural History, New York, New York, United States of America


Birds are of scientific importance in fields such as ecology and especially paleontology, given their evolutionary link to dinosaurs.  We are studying birds to understand brain mechanisms related to flight and the implications for the evolution of flight in extinct species. While there are abundant anatomic, genetic, and behavioral data in birds, there are only a handful of studies on brain function. FDG PET imaging is a promising approach, but there are multiple technical challenges in this species, and here we address how to best capture brain activity during flight and how to quantify the regional metabolic rates of glucose.  We initially studied starlings flying in a wind tunnel and used our RatCAP portable PET scanner for brain imaging, and are now utilizing pigeons with a more natural flight behavior.  FDG PET in anesthetized birds demonstrated peak brain uptake similar to that of rodents (~0.5%ID/cc) despite much higher endogenous blood glucose levels.  However, an unusual observation was that brain uptake of FDG steadily decreased in contrast to mammals.  To better understand this effect, we developed appropriate blood sampling methods and performed the first ever quantitative PET studies in birds.  Two tissue compartment modeling indicates that brain time activity curves are not consistent with the irreversible model that is usually assumed for mammals, yet gray matter metabolic rates are similar to mammals despite substantial physiological differences.

Keywords: brain imaging; PET; behavior; glucose metabolic rate; kinetic modeling
Poster panel
(face) ID: 84

Poster Number:

Optimized Image Reconstruction and Preliminary Clinical Applications of the MRI-compatible VersaPET System (#3535)

S. Wei1, D. Ouellette1, M. Salerno1, T. Cao1, J. S. Karp2, P. Vaska1, 3

1 Stony Brook University, Biomedical Engineering, Stony Brook, New York, United States of America
2 University of Pennsylvania, Radiology, Philadelphia, Pennsylvania, United States of America
3 Stony Brook University, Radiology, Stony Brook, New York, United States of America


The VersaPET scanner is a versatile high-resolution MRI-compatible PET system capable of rodent imaging and now being adapted for clinical applications including image-derived input function from arteries in the leg during brain PET/MRI studies, and simultaneous PET/MRI of the breast for cancer studies.  It has demonstrated good NEMA performance and MRI compatibility in previous measurements.  However, image reconstruction is a challenge due to gaps between blocks in both the transaxial and axial directions, which results in image artifacts if open-source packages such as STIR are used.  Currently, we have developed an on-the-fly fully 3D MLEM image reconstruction for VersaPET which is based on the STIR projectors but accounts for the scanner’s block-based geometry and gaps.  To optimize imaging performance, in part because the imaging FOV extends close to the detectors, we are also incorporating aspects of the point spread function (PSF) into the system model including the parallax effect due to the crystal penetration and inter-crystal scattering via a Monte-Carlo-generated sinogram-based blurring kernel. The blurring kernel was incorporated in the MLEM reconstruction using GATE-simulated data based on the exact VersaPET geometry.  Preliminary phantom and human input function imaging using our customized reconstruction approach demonstrate high-quality, artifact-free images and results from PSF modeling are promising.

Keywords: Positron Emission Tomography (PET), PET/MRI, Maximum Likelihood Expectation Maximization (MLEM), Point Spread Function (PSF) Modeling, Monte-Carlo Simulation, Sinogram-based Blurring Kernel
Poster panel
(face) ID: 87

Poster Number:

First experimental evaluation of MAPSSIC prototype, a potential novel CMOS intra-cerebral beta+ probe for deep brain imaging in awake and freely moving rat. (#3895)

L. Ammour1, J. Heymes5, M. Bautista6, S. Fieux4, F. Gensolen6, M. Kachel5, F. Lefebvre1, F. Pain1, P. Pangaud6, L. Pinot1, J. Baudot5, P. Gisquet-Verrier3, P. Laniece1, C. Morel6, L. Zimmer4, M. - A. Verdier2

1 Univ. Paris Sud, Université Paris-Saclay, CNRS/IN2P3, IMNC, Orsay, France
2 Univ Paris-Diderot, Sorbonne Paris Cité, CNRS/IN2P3, IMNC, Orsay, France
3 Univ Paris Sud, Université Paris-Saclay, CNRS/INSB, NeuroPSI, Orsay, France
4 Univ Lyon 1, CNRS/INSERM, CRNL, Bron, France
5 Univ Strasbourg, CNRS/IN2P3, IPHC, Strasbourg, France
6 Aix Marseille Université, CNRS/IN2P3, CPPM, Marseille, France


Among the numerous methods developed to address neuroscience research needs, the combination of pre-clinical PET with behavioral studies has been recently pointed out as a potential key breakthrough to go further in the understanding of functional processes in the brain.
Achieving such a combination is difficult. Anaesthesia or restraints inherent to micro-PET preclude its use for behavior studies. In that context, we have presented an original strategy using intracerebral probes with submilimeter pixels to directly measure positrons inside the rat brain. Integrated electronics and wireless communication system allow fully freely-moving rats experiments. Moreover, the small detection volume resulting  from the low range of positrons is comparable with rat brain loci sizes (inferior to 1 mm).
Previous intracerebral probes have shown promising results but have suffered from various limitations in terms of mechanical and electronic robustness. A new generation of probes called MAPSSIC benefits from innovative CMOS MAPS sensors to overcome these limitations. Following Monte-Carlo studies that have evidenced high positrons sensitivity and low gamma rays background capabilities, we present here the first experimental results obtained with a prototype sensor. The CMOS sensor features a matrix of 16×128 pixels, which are 30×50 μm² large, the sensitive layer is a thin 18 μm high-resistivity epitaxial layer.
Experiments were conducted with ¹⁸F, ²⁰⁴Tl and ⁵⁵Fe sources over a large range of activities and different integration times, using various attenuation mediums. Measurements show a good sensitivity to positrons: 3.58×10² events/MBq/s for a 7 mm remote, 3 mm thick ¹⁸F source in air, while a low sensitivity to annihilation gamma rays (no significant difference were measured when adding a gamma source) and a low background noise (9.0×10⁻⁴ events/s).
These results confirm our confidence in CMOS MAPS technology for intracerebral positrons detection applications.

Keywords: brain, rat, awake, freely-moving, CMOS, MAPS, direct positrons detection
Poster panel
(face) ID: 90

Poster Number:

LOR-PET: a novel PET camera constructed with a monolithic scintillator ring (#4058)

S. Xie1, Z. Zhao2, M. Yang1, F. Weng2, Q. Huang2, J. Xu1, Q. Peng3

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


Conventional PET cameras are constructed with detectors which consist of arrays of discrete scintillators and photosesnsors. Arrays of very fine discrete crystals have been used in high-resolution small animal imaging.  However, it is very challenging to machine and assemble discrete crystals with a size smaller than 1 mm. Thus, the arrays of fine crystals are usually very expensive. And the quality of these arrays, which is heavily depended on the vendor’s skills, does not always meet the requirements.

In this paper, we present a novel PET system constructed with a monolithic scintillator ring. This PET system is very different from any PET systems which have been constructed so far. It is constructed from a single piece of monolithic crystal. And the lines of response (LOR) are calculated directly from the light distribution around the crystal ring, instead of being calculated from the measurement of the positions of interaction (POI) of the two gammas. Thus, we name it LOR-PET.

We have designed and constructed the first LOR-PET with a monolithic LYSO scintillator ring. The inner and outer diameters of the ring are 48.5 mm and 58.5 mm respectively. The length and thickness of the ring are 25.12mm and 5 mm respectively. Twenty-three 4x1 SiPM arrays are fabricated and mounted inside of the monolithic LYSO crystal ring. A low-cost high-performance electronics system (named Pico-PET electronics) was built to read out the 92 channels of SiPM in parallel. The results of initial tests shows that the electronics system works very well. We conclude that the LOR-PET system is a unique low-cost PET system that has the potential to achieve good sensitivity and spatial resolution in small animal or organ specific imaging. We are performing system characterization and phantom study to further assess the performance of the LOR-PET system.

Keywords: Line of Response, PET, Small animal, Low cost
Poster panel
(face) ID: 93

Poster Number:

Design and Characterization of 4x1 MD-SiPM Small-Animal PET Building Block (#4114)

S. Mandai1, E. Venialgo1, E. Charbon2

1 Delft University of Technology, Applied Quantum Architectures, Delft, Netherlands
2 École polytechnique fédérale de Lausanne, Delft, Switzerland


In high-resolution small-animal PET scanners, CRT (coincidence resolving time) is a critical parameter. The gamma-photon timemark accuracy determines the capacity to reject spuriously detected random coincidences, which limit the maximum achievable NECR (noise equivalent count rate) of the scanner. In addition, high spatial granularity is desirable for high-resolution small-animal PET images.

In this work, we design and implement an optimized small-animal PET detector module that is the key building block for a high spatial and timing resolution scanner. The module is composed of 4x1 monolithic MD-SiPM array chips, which features and optimized packing fraction with respect to the previous 4x2 version. In addition, the monolithic arrays of 9x18 MD-SiPMs are readout independently in order to boost the single-event detection throughput of the scanner.

We present the full small-animal PET detector module characterization including: energy resolution, CRT, spatial resolution and uniformity, sensitivity, thermal, dead time, and calibration stability characterizations.

Keywords: small-animal PET, MD-SiPM, Multiple timestamps
Poster panel
(face) ID: 96

Poster Number:

Simulation study on a stationary SPECT system with multi-pinhole collimators (#1075)

H. Kubota1, N. Motomura2, K. Ogawa1

1 Hosei University, Department of Applied Informatics, Tokyo, Japan
2 Toshiba Medical Systems, Tochigi, Japan


Recently, semiconductor detectors are used to detect gamma rays instead of scintillation detectors in single photon emission CT (SPECT). In these systems, dedicated collimation systems are adopted to reconstruct a myocardium, and data samplings in the reconstruction area are inhomogeneous. This sometimes results in artifacts depending on the size of a patient and location of the myocardium. To avoid these artifacts, we have studied a multi-pinhole SPECT system with a triple head gamma camera. In this paper, we investigated the relationship between the geometrical efficiency of multi-pinhole collimators determined by the diameter of holes and the quality of reconstructed images with Monte Carlo simulations. In these simulations we assumed 19 pinholes and changed the diameter of pinhole collimators; 2, 3, and 5 mm. We calculated projection data with three conditions: a stationary data acquisition with three angular positions, a semi-stationary data acquisition with six angular positions and 12 angular positions over 360 deg. The results of simulations showed that the semi-stationary data acquisitionwith a small pinhole diameter could high quality images.

Keywords: pinhole collimator, SPECT, triple head gamma camera
Poster panel
(face) ID: 99

Poster Number:

Performance Evaluation of A Prototype PET Detector With High-Growth-Rate Scintillator and Wavelength-Shifting fibers (#1467)

K. Fujihara1, Y. Emoto1, H. Ito1, H. Kawai1, S. Kimura1, A. Kobayashi1, H. Matsunaga2, T. Mizuno1, T. Nakamura1, T. Tanaka2, T. Yuzawa2

1 Chiba University, Graduate school of science, Chiba, Japan
2 Chiba University, Department of Physics, Faculty of Science, Chiba, Japan


Existing PET (Positron Emission Tomography) systems make unclear images in real diagnosis. It is considered that this problem arise from Compton scattering in detector. When scattering occurs, scintillation light accrue from plural crystal. Therefore, reconstructed image with centroid method becomes unclear. Owing to identify scattering, We are developing a new PET system with plate-like high-growth-rate (HGR) La-GPS ( [Ce0.01La0.24Gd0.75]2Si2O7) scintillators and wavelength-shifting fibers (WLSF) linked to PMTs. HGR scintillators are lower cost and have lower transparency because of microbubbles. This is useful for reducing cost and identifying scattering. In this system, emission positions are measured by WLSFs’ sheets on top and bottom of scintillators. If transparency is moderately low, scintillation light don’t spread in scintillator; nevertheless, sufficient quantity of light is ensured. This produces good position resolution discrimination ability of scattering in one scintillator. In this study, a preliminary test was performed for developing the PET detector by using a 22Na sources, where the consistency was demonstrated by using an numerical simulation, Monte Carlo code: GEANT4. In experimentation, We used trigger counter and PMTs linked to WLSF sheets. We adopt events when trigger PMT's value is in 511-keV peak. In result, the number of photoelectron with 460-560 keV gamma rays is estimated to be 80.0 ± 11.9 . Therefore, energy resolution is about 36% and we can measure 50-keV deposit. Position resolution is during measuring. In simulation, parameters are as follows: 0.2 mm width resolution, 1 mm depth resolution, 15% energy resolution, and no error time resolution. In calculation, if the event is regarded as scattering event, nearest emission point to body is adopted as the first emission point. In result, this system identified 3 mm cancer with less exposure.

Keywords: PET, Wavelength sifting fiber, high-growth-rate sintillator, La-GPS
Poster panel
(face) ID: 102

Poster Number:

Scalability Test of Hybrid Charge Division Multiplexing Using A SiPM-based Modular PET Detector (#1825)

H. Park1, 2, J. - W. Son1, 2, G. B. Ko1, 2, S. Seo3, J. S. Lee1, 2

1 Seoul National University, Department of Nuclear Medicine, Seoul, Republic of Korea
2 Seoul National University, Department of Biomedical Sciences, Seoul, Republic of Korea
3 Gachon University, Department of Neuroscience, Incheon, Republic of Korea


One of the major challenges in SiPM-based PET system is the management of a large number of individual output channels. Therefore, charge division multiplexing is widely used in PET detector to reduce the number of output channels of SiPM in a cost-effective manner. However, the conventional resistive and capacitive charge division multiplexing methods suffer from an unwanted pulse shape distortion, undermining the performance of PET detector. To address this problem, we have recently proposed and validated the concept of a hybrid charge division multiplexing method using discretized positioning circuit (DPC) combined with a Hamamatsu 16-channel SiPM. The aim of this study is to investigate the scalability of the hybrid DPC (64:4 multiplexing ratio) by implementing LSO/SiPM modular PET detectors.

Prior to the experiment, we performed PSPICE simulation for the feasibility assessment of the 8 x 8 hybrid DPC. We repeated the simulation by recursively replacing the resistor-capacitor pairs of the hybrid DPC until we achieve the uniform pulse shape distribution throughout the multiplexing network. The selected passive components were subsequently used for the implementation of SiPM-based PET detector module for the real measurement.

A designed modular PET detector supports four detector blocks: each block detector consists of a 8 x 8 array of 3.95 x 4.4 x 20 mm3 LSO crystal block and a 2 x 2 array of Hamamatsu 16-channel SiPM. The PET signals were processed within the FPGA-based TDC and ADC, and the data were acquired in a real-time via custom-built software SOPHIA. The result indicated that our designed modular PET detector was able to resolve 64 crystals, demonstrating the feasibility of 8 x 8 hybrid DPC in the real measurement. Our next milestone will be the construction of a full-ring PET system by arranging the modular PET detectors. In addition, further optimization of the hybrid charge division multiplexing network also will be proceeded as a future work.

Keywords: Hybrid Charge Division Multiplexing, Charge division Multiplexing, Silicon Photomultiplier (SiPM), Positron Emission Tomography (PET)
Poster panel
(face) ID: 105

Poster Number:

Improved Point Spread Function Modeling for a Multi-Pinhole SPECT Camera Using Higher Order Polynomial Terms. (#2107)

S. Banerjee1, A. Könik1, J. M. Mukherjee2, K. S. Kalluri1, J. C. Goding1, L. Caucci3, G. Zubal4, L. R. Furenlid3, M. A. King1

1 University of Massachusetts Medical School, Radiology Nuclear Medicine, Worcester, Massachusetts, United States of America
2 The Mathworks, Inc., Natick, Massachusetts, United States of America
3 University of Arizona, Center for Gamma Ray Imaging, Tucson, Arizona, United States of America
4 Z-Concepts, LLC, East Haven, Connecticut, United States of America


Herein we report on the mathematical modeling of the simulated point spread functions (PSF) of pinhole apertures for clinical I-123 DaTscan imaging on a dual-head SPECT system consisting of fan and multi-pinhole (MPH) collimators on separate heads. The PSF can be measured sparsely by translating a point source within the volume of interest (VOI). These PSFs are generated using GATE Monte Carlo simulation software and then modeled using standard 2D Gaussian having 6 parameters, and three other models using higher order polynomial terms as well as cross terms in the exponential. The goal is to efficiently store the parameters of the modeled PSF measured across the VOI and then interpolate them on the fly during reconstruction. It has been shown that MPH reconstruction can be improved with accurate modeling of the PSF. However, for our application it has been determined that improved accuracy in PSF modeling (reduced NRMSE) can be obtained by incorporating more polynomial terms in the exponential than employed by the standard 2D Gaussian, especially with increased pinhole angulations. In this paper we introduce higher order polynomial terms (degree 3 and 4) in addition to the Gaussian. The error for the most oblique pinholes was reduced by up to 27%. These added extra terms could model the penetration and depth of interaction effects of the oblique pinholes. For the direct pinhole, the additional terms did not appear to improve the accuracy as these effects have less impact when the pinhole is perpendicular to the detector surface. We also investigated different pinhole aperture sizes ranging from 1mm to 5mm radius. The 3rd and 4th degree polynomial terms could model the PSF obtained from the 3mm pinhole to a great extent, but not so well for the 5mm radius pinhole. Hence, we are investigating different approaches for very large aperture sizes. Also, linearity in parameters across VOI obtained from the model would facilitate easy and efficient interpolation.

Keywords: SPECT, DaTscan, Multi-pinhole (MPH), Point Spread Function (PSF), Gaussian
Poster panel
(face) ID: 108

Poster Number:

Development of a PET-MR insert for small animals with high resolution and sensitivity, using depth-encoding detectors (#2254)

A. R. Selfridge1, S. R. Cherry1, M. Judenhofer1

1 University of California, Davis, Biomedical Engineering, Davis, California, United States of America


Preclinical PET-MR has proved to be a flexible research tool, facilitating longitudinal study of subtle changes in animal physiology and biochemistry. The continuous push towards high resolution MR-compatible PET systems has opened new and exciting applications. We are constructing a system which will continue to push the frontiers of spatial resolution, but will also achieve exceptional sensitivity by increasing the axial field-of-view, and using thick scintillating crystal blocks to improve photon stopping power. To maintain spatial resolution despite positioning errors common to preclinical geometries, we are incorporating a dual-ended crystal readout to provide measurement of the photon’s depth-of-interaction (DOI). Furthermore, we are incorporating robust, self-contained readout electronics, resulting in a system which is compact, easy to use, and reliable for the day-to-day rigors faced by biologists and imaging scientists.

System development is subdivided into three aims: (1) design and evaluation of a modular block detector, (2) development of system readout electronics, and (3) system assembly and characterization. The first of these three aims is complete, and an optimized design for a modular detector has been selected. These results of a variety of tested designs are presented here.

The completed and selected detector module has an energy resolution of 14 ± 1 % and DOI resolution of 3.8 ± 0.6 mm. Furthermore, the proposed module can resolve 1 mm crystals with good accuracy, without introducing excessive dead space in the assembled system. Experiments to understand noise contributions from different photodetector readout configurations are ongoing, and will allow us to decide on how the readout electronics can be incorporated into the overall system.

Keywords: PET, MRI, instrumentation, depth-of-interaction
Poster panel
(face) ID: 111

Poster Number:

Attenuation correction for fixed MR components in a simultaneous PET/MR system (#2730)

H. Xie1, Y. Lv2, Y. Dong2, J. Zhao1

1 Shanghai Jiao Tong University, School of Biomedical Engineering, Shanghai, China
2 Shanghai United Imaging Healthcare Co., Ltd., MI Recon Dept., Shanghai, China


Attenuation correction for fixed MR components like MR table and coils in a simultaneous positron emission tomography and magnetic resonance (PET/MR) system is mandatory for accurate quantification of activity distribution. Conventional method to derive the attenuation maps of MR table and coils includes CT scan and transmission scan.  Both methods have their advantages and disadvantages. CT scan method is straightforward, however, the images obtained from CT scan have severe metal artifacts due to wires and circuits,and the attenuation coefficients transformation from X-ray energy to Gamma ray energy needs to be remodeled. Transmission scan method avoids the transformation between different energy levels but needs an extra setup of a rotating rod source (usually 68Ge) and its resolution is poor.  In this work, we aim to combine the merits of these two methods. Instead of complex rotating rod source device, a 68Ge cylinder source was placed at the center of field of view (FOV) as a stationary emission source.  By doing so, the full scan reconstruction problem turned to be a limited angle reconstruction problem. Modified ordered subset expectation-maximization (OSEM) algorithm incorporated with Split Bregman algorithm was employed to reconstruct the attenuation map directly from the emission data.  An additional CT scan provided the structure information of the MR table and coils, which was used as prior knowledge to confine the region of attenuation coefficients. The proposed method was validated by simulations and real data acquired with a simultaneous PET/MR system.

Keywords: Attenuation correction, MR components, PET/MR, OSEM
Poster panel
(face) ID: 114

Poster Number:

A Proof-of-Principle Time-of-Flight (TOF) PET Utilizing a Capacitive Multiplexing Circuit (#2849)

H. - J. Choe1, Y. Choi1, D. J. Kwak1, H. Yan2, H. Ye2, J. Yan2, Y. Wang2

1 Sogang University, Electronic Engineering, Seoul, Republic of Korea
2 Minfound Medical Systems Co., Ltd., Zhejiang, China


A capacitive multiplexing circuit is an attractive solution to achieve fast coincidence resolving time (CRT) while effectively reducing the number of readout channels since it provides fast rise time and low baseline fluctuation. Compared to an individual channel readout or ASIC based readout, it can be a cost-effective way in the development of time-of-flight (TOF) PET system. The purpose of this study was to develop a proof-of-principle TOF PET utilizing the capacitive multiplexing circuit having excellent CRT performance and cost-effectiveness.

The proof-of-principle TOF PET consisted of two signal processing boards with a motorized rotation stage. Each signal processing board was composed of two major parts: capacitive multiplexing board and data acquisition (DAQ) board. The multiplexing board was designed to reduce the number of readout channels of a detector module which consisted of an 8×8 array of 3×3 mm2 SiPMs (MicroFJ-30035-TSV, SensL, Ireland) and an 8×8 array of 3×3×20 mm3 LYSOs (CPI, USA) using a 64:4 capacitive multiplexing circuit. Two detector modules were installed to the multiplexing board. The multiplexed outputs were sent to the DAQ board equipped with an eight-channel high-speed ADC and FPGA in which data processing modules and tapped-delay-line TDCs were implemented. Event packets were transferred from the FPGA to PC using high-speed serial communication via a DisplayPort cable. Energy and timing performance was evaluated by measuring flood histogram, energy resolution, and CRT of the detector module.

The average energy resolution of the detector module was 11.8±1.0% after energy calibration. The average CRT of 565±34 ps FWHM was measured with a reference detector. PET image acquisition using the TOF PET is in progress.

The TOF PET exploiting the capacitive multiplexing circuit showed good energy resolution and CRT. Further performance evaluation will be conducted by assessing PET image acquired using the TOF PET developed in this study.

Keywords: TOF, PET, Capacitive, Multiplexing
Poster panel
(face) ID: 117

Poster Number:

Integrated Level set-based Deep Learning for Pancreas Segmentation (#3012)

A. Taneja1, P. Ranjan2, A. Ujlayan3, R. Janardhanan4

1 Amity University, Amity Institute of Information Technology, Noida, Delhi, India
2 Amity University, Noida, Delhi, India
3 Gautam Budha University, Greater Noida, Delhi, India
4 Amity University, Amity Institute of Public Health, Noida, Delhi, India


Early computer diagnosis is the necessary stage for quantitative analysis in computer vision-based cancer image analysis which requires the segmentation as an essential stage. Due to the shape, size and the location of pancreas in abdominal Magnetic Resonance Imaging (MRI), the segmentation process is considered as the difficult task. Besides, the variations in appearances, consistency / inconsistency and the features have the great impact on the classification. Hence, the visual clues extraction related to the pixel level labelling is considered as the major task in diagnosis applications. High-anatomical variations in the pancreas affects the visual clues-based segmentation methods and limits the accuracy compared to other organ segmentation from abdominal images. This paper proposes the integrated framework of Gaussian with the level set formulation to avoid the probable mask creation. Initially, the proposed work employs the Laplacian formula to filter the noise regions from the images and the second-order formulation to sharpen the edges of images. Then, the integral framework of Gaussian with the level set formulations segment the pancreas from the images. Then, the N-ternary-based methodology extracts the different patterns in the images. Finally, the training of Convolution Neural Network (CNN) from the results from the N-ternary pattern modules depicts the labels for classification. The comparative analysis between the proposed CNN-based learning models with the various existing methods regarding the performance parameters assure their effectiveness in early diagnosis models.

Keywords: Pancreas, Segmentation, Convolution Neural Network (CNN), Levelset, Ternary Patterns
Poster panel
(face) ID: 120

Poster Number:

High density interconnect NUV-HD SiPMs: performance assessment (#3284)

A. Ferri1, F. Acerbi2, A. Gola2, D. Krabe1, T. Lichtenegger1, G. Paternoster2, C. Piemonte2, M. Stich1

1 Avago Technologies GmbH, a Broadcom Limited company, IFPD, Regensburg, Bavaria, Germany
2 Fondazione Bruno Kessler (FBK), Centro Microsistemi e Materiali, Trento, Italy


One challenge for the next generation of PET scanners is to cover large areas minimizing the inactive space. The best approach is to cover the PET ring with tiles composed of tightly packed single SiPMs. The inactive area can be minimized with the through silicon via (TSV) technology that enables to completely remove the bonding wire and solder the sensors similarly to surface mount components. A 16-channel tile with a fill factor of 93 % has been manufactured. We describe the main features of the packaging technology and analyze its effect on the overall performance of the device, with special focus on electro-optical characteristics and timing resolution with commercial LYSO scintillators. The TSV process did not introduce additional noise sources (primary and correlated), and maintained a very high PDE at 420 nm, above 55% at 5 V OV. The best measured timing resolution is 111 ps with commercially available 3x3x5 mm3 LYSO crystals.

Keywords: SiPM, PET, TSV
Poster panel
(face) ID: 123

Poster Number:

Adaptive Detector Positioning for a RoomPET System (#3703)

B. K. Byrd1, W. Xi2, A. G. Weisenberger2, B. Kross2, S. Lee2, J. E. McKisson2, J. McKisson2, C. Zorn2, M. F. Smith3

1 Christopher Newport University, Department of Physics, Newport News, Virginia, United States of America
2 Thomas Jefferson National Accelerator Facility, Nuclear Physics, Newport News, Virginia, United States of America
3 University of Maryland School of Medicine, Department of Diagnostic Radiology & Nuclear Medicine, Baltimore, Maryland, United States of America


Investigating human brain function in response to environmental stimuli provides valuable information on brain physiology. Ideally, studies would be performed with freely moving humans without attached instrumentation. Covering the surface area of a room with fixed PET detectors is too expensive and has prompted us to consider a concept called RoomPET, in which moving PET detectors adaptively move to follow a subject’s head. RoomPET contains an optical head tracking system and PET detectors that move to keep the head in the field-of-view. As proof of principle, a small box-size prototype was built with gantry mechanics, tracking capabilities, and control systems to enable horizontal translations to keep a target in the PET field-of-view. A four-camera infrared optical vision system used fiducial markers and provided 3D positional information that is used to adaptively position the PET detectors. The positional accuracy of the driving mechanism was evaluated by moving the detectors a controlled increment across the field-of-view. The measured translation was within 1% of the expected value at all locations. Positional precision was assessed with stationary experiments at different positions across the field-of-view. The standard deviation of 3D position measurements averaged less than 20 micrometers in each dimension. The uncertainty in positional data caused by a moving target was investigated by collecting data for a static target and for motion at three different constant velocities. The deviations of points from fitted trendlines provided RMS position error estimates. These averaged to 1.2, 3.9, 5.2 and 9.1 micrometers for the speeds of 0, 3.87, 7.75 and 11.63 cm/s, respectively. These positioning errors are much smaller than PET system resolution and indicate the promising potential of adaptive detector motion for a full-size RoomPET system.

Keywords: PET, Brain, optical vision
Poster panel
(face) ID: 126

Poster Number:

Performance Demonstration for PET Detector Using 0.2 mm phi Wavelength-Shifting Fibers (#4019)

A. Kobayashi1, H. Ito2, H. Kawai2

1 Chiba University, Graduate School of Science and Engineering,, Chiba-shi, Chiba, Japan
2 Chiba University, Graduate School of Science, Chiba-shi, Chiba, Japan


Recently we are developing a new gamma ray detector for Positron Emission Tomography (PET). It is expected the position resolution and the energy resolution are a few mm and approximately 10%, respectively. The detector consist of six plate detector (the effective area is 300 mm times 300 mm) setting up around the human body. The plate detector consists of 8 layers. Each layer consists of 8 times 8 scintillator plates. The size of each scintillator plate is 34 mm times 34 mm times 3.4 mm. 160 wavelength-sifting fibers (WLSFs) cover each 34 mm times 34 mm surface of each plate along each axis (x or y). Ends of each fiber are connected to silicon photomultipliers (SiPMs). On each side (surface of 34 mm times 3.4 mm) of scintillator plate, 10 SiPMs are glued. When a gamma ray incidents to the scintillator plate and deposit energy, the light is emitted. The light leaked out from the scintillator plate enters to WLSFs and reemission is occurred in WLSFs. The light of reemission propagates to ends and detected by SiPMs connected end of the fiber. By a distribution of reemitting WLSFs, the incident position of gamma is measured. On the other hands, the light satisfied total reflection condition detected by glued SiPMs on side, The deposit energy and incident time of gamma are also measured.

In this study, the detector performance for 511-keV gamma ray is estimated by an experiment. Its setup consist of a La-GPS [(La,Gd)2Si2O7(Ce)] scintillator plate of 34 mm times 34 mm times 3.4 mm, WLSFs, SiPMs. We made a narrow 511-keV gamma beam from 22Na enter the setup and reconstruct incident position and energy by analyzing signal of SiPMs. As a result of analysis, we prove that our detector has the position resolution of 5.4 mm and the energy resolution of 10.08%. In addition, we are conducting more accurate experiment. Its result also will be shown in this conference.

Keywords: PET, Wavelength-shifting fiber
Poster panel
(face) ID: 129

Poster Number:

Design of a 2-bit sigma-delta modulator for scintillation waveform capture (#4193)

M. Ruiz-Gonzalez1, 2, L. R. Furenlid1, 2

1 University of Arizona, Center for gamma-ray imaging, Tucson, Arizona, United States of America
2 University of Arizona, College of Optical Sciences, Tucson, Arizona, United States of America


Sigma-delta modulation is a technique used in analog-to-digital conversion that consists of oversampling an input signal and encoding it into a 1-bit data stream. It has been shown that it can be implemented in a field-programmable gate array (FPGA) using minimal external analog components. Zhao et al. demonstrated that energy of scintillation pulses can be computed by adding up all the elements on the output from a conventional 1-bit sigma-delta modulator. However, for parameters selected for a suitable resolution and dynamic range, the system passes through a period of saturation. During saturation, information about the shape of the pulse is lost even though the integrated energy remains preserved. We present the design of a 2-bit version of the sigma-delta modulator that makes use of an adaptive 1-bit digital-to-analog converter (DAC) that keeps the system unsaturated. Each output bit is scaled by one of the four possible values the DAC can take, resulting in a 2-bit modulation. We have demonstrated that the 2-bit modulation retains information about the pulse shape. One important advantage of having information about the shape of the pulse is the identification of events inconsistent with a forward model, such as having pulse pile up, when using maximum-likelihood energy- and time-estimation methods.

Keywords: analog-to-digital conversion, electronics, sigma-delta modulation, scintillation pulse, waveform capture
Poster panel
(face) ID: 132

Poster Number:

Comparing Different Preprocessing Methods in Automated Segmentation of Retinal Vasculature (#2178)

M. Tavakoli1, P. Kelley1, A. Golestaneh2, F. Kalantari3

1 Indiana University-Purdue University, Physics, Indianapolis, Indiana, United States of America
2 Arizona State University, Electrical Engineering, Pheonix, Arizona, United States of America
3 University of Texas Southwestern Medical Center, Radiation Oncology, Dallas, Texas, United States of America


Many systemic diseases change the vascular network and could be diagnosed through the eye. Normally evaluation of retinal vascular network to find abnormality are done by ophthalmologists, which is time consuming and associated with error and fatigue. One possible solution for these problems is to use Computer Assisted Diagnosis systems. Here we applied two different preprocessing methods (Illumination equalization along with contrast enhancement, and top-hat transform) separately and compare the results of vessel segmentation in each of these two ways. The final objective of this study is to apply three retinal vessel segmentation methods, Laplacian of Gaussian edge detector (uses the second-order spatial differentiation), Canny edge detector (estimate the gradient intensity), and Matched filter edge detector to preprocessed images for segmentation of retinal vessels either in normal fundus images or in presence of retinal lesion like in diabetic retinopathy.

In general, the steps for our segmentation methods are in following: (1) Smoothing: suppress as much noise as possible, without destroying the true edges, (2) Enhancement: apply a filter to enhance the quality of the edges in the image (sharpening), (3) Detection: determine which edge pixels should be discarded as noise and which should be retained by thresholding the edge strength and edge size, (4) Localization: determine the exact location of an edge by edge thinning or linking.

From accuracy view point, according to manual segmentation by ophthalmologist, for all retinal images of test set (240 images from 2 public databases), by using the first preprocessing, Illumination equalization along with contrast enhancement, for Laplacian-of-Gaussian, Canny, and Match filter our segmentation results were 89.12%, 85.97%, and 90.54% respectively. Also by using the second preprocessing, top-hat, for Laplacian-of-Gaussian, Canny, and Match filter our segmentation results were 88.09%, 85.17%, and %91.04 respectively.

Keywords: retinal vessel, Laplacian of Gaussian, Matched filter, Canny edge detector, Illumination equalization, contrast enhancement, top-hat transform
Poster panel
(face) ID: 135

Poster Number:

Multi-Materials Decomposition using clinical Dual-energy CT (#3650)

T. Zhao1, K. Kim2, D. Wu2, M. K. Kalra3, G. E. Fakhri2, Q. Li2

1 Tsinghua University, Department of Engineering Physics, Beijing, China
2 Massachusetts General Hospital, Gordon Center for Medical Imaging, Boston, Massachusetts, United States of America
3 Massachusetts General Hospital, Department of Radiology, Boston, Massachusetts, United States of America


In medical application of CT, the scanned body is composed of several types of materials, such as bone, soft tissue, fat and blood, and we want to use the dual-energy CT to get the material information. The multi-materials decomposition method has been proposed to achieve this goal. A projection-domain decomposition is firstly performed to estimate the attenuation coefficient function of the scanned object. Then the estimated attenuation coefficient function is decomposed with the attenuation coefficients of the basis materials. The problem is that in the projection-domain method, the spectrum of the x-ray source or additional calibration data is necessary but may not be acquired in practice. We propose an image-domain multi-materials decomposition method. The basic idea is the same with the original multi-materials decomposition method.  The difference is that we use the direct reconstruction of DECT projection data to replace the projection-domain decomposition and the direct reconstructed value is used to replace the attenuation coefficient functions.  Accordingly, the attenuation coefficient of the basis materials should be carefully selected for the best performance of the decomposition because the direct reconstructed value is the effective attenuation coefficient. We use the clinical DECT data to conduct experiments. Without knowing exact spectrum, the proposed method decomposed the materials accurately.  The attenuation coefficients of basis materials are chosen empirically. An automatic method is a future work. 

Keywords: Dual-energy CT, Multi-materials decomposition
Poster panel
(face) ID: 138

Poster Number:

Linear Attenuation Coefficient Measurements of Low Atomic Number Materials by Energy-resolved Computed Tomography using a transXend Detector (#1660)

I. Kanno1, Y. Maruyama1, T. Hamaguchi1, T. - H. Tsai1

1 Graduate School of Engineering, Kyoto University, Kyoto, Japan


This paper shows the method of response function estimation for the segmented detectors of a transXend detector, which is employed for energy-resolved X-ray computed tomography (CT). We performed Monte Carlo simulation calculations using GEANT 4 code for the response function estimation with energy bin width 1 keV from 11 to 120 keV. Energy-resolved CT measurement was performed for a 30 mm diameter acrylic phantom with four 5 mm diameter resin rods: polypropylene, acrylonitrile butadiene styrene, polycarbonates, polyvinylidene difluoride. Energy-resolved CT images give linear attenuation coefficients as a function of X-ray energy. The obtained linear attenuation coefficients of 5 resins were estimated within 5 % of relative error to the theoretical ones in the energy range 30-100 keV. This method is useful for tissue segmentation and effective atomic number measurement.

Keywords: X-ray, computed tomography, energy-resolved, linear attenuation coefficient
Poster panel
(face) ID: 141

Poster Number:

Quantitative decomposition for three materials in spectral computed tomography (#1811)

B. Jo1, D. Kim2, H. - M. Kim2, H. - J. Kim1, 2

1 Yonsei University, Department of Radiological Science and Research Institute of Health Science, Wonju, Gangwon-do, Republic of Korea
2 Yonsei University, Department of Radiation Convergence Engineering, College of Health Science, Wonju, Gangwon-do, Republic of Korea


A spectral computed tomography (CT) system based on an energy-resolved photon-counting Cadmium Zinc Telluride (CZT) detector with a dual energy technique can provide spectral information and possibly distinguish between two or more materials with a single X-ray exposure using energy thresholds. This work provides the potential three-material decomposition of vulnerable plaques using two inverse fitting functions. Additionally, there exists the possibility of using gold nanoparticles as a contrast agent for the spectral CT system in conjunction with a CZT photon-counting detector. In this simulation study, we used fan beam CT geometry that consisted of a 90 kVp X-ray spectrum and performed calculations by using the SpekCal program (REAL Software, Inc.) with Monte Carlo simulations. A basic test phantom was imaged with the spectral CT system for the calibration and decomposition process. This phantom contained three different materials, including lipid, iodine and gold nanoparticles, with six holes 3 mm in diameter. In addition to reducing pile-up and charge sharing effect, the photon counting detector was considered an ideal detector. Then, the accuracy of material decomposition techniques with two inverse fitting functions were evaluated between decomposed images and reference images in terms of root mean square error (RMSE). The results showed that decomposed images had a good volumetric fraction for each material, and the RMSE between the measured and true volumes of lipid, iodine and gold nanoparticle fractions varied from 12.51% to 1.29% for inverse fitting functions. The study indicated that spectral CT in conjunction with a CZT photon-counting detector in conjunction with a dual energy technique can be used to identifying materials and may be a promising modality for quantifying material properties of vulnerable plaques.

Keywords: Photon counting detector, CZT, material decompostion, spectral computed tomography
Poster panel
(face) ID: 144

Poster Number:

MicroPattern Gaseous Detector for Energy Resolving X-ray Imaging (#2393)

L. F. N. D. Carramate1, S. C. Sousa1, A. L. M. Silva1, C. D. R. Azevedo1, S. G. Monteiro1, J. F. C. A. Veloso1

1 I3N, Department of Physics, Aveiro, Portugal


A MicroPattern Gaseous Detector (MPGD) operating in a sealed mode was developed and studied in order to be applied for energy resolved X-ray imaging.

The detector is operating in pure Kr at 1 bar in sealed mode with a closed purification system based on getters to remove the gas impurities continuously and allow a constant detector performance. The first testes revealed stable performance in terms of charge gain and energy resolution over time.

The detector has been tested for X-ray transmission imaging of some samples (with different attenuation coefficients) and the energy information of individual photons has been used to enhance image quality or extract more information than conventional systems. This work intends to present some energy resolved X-ray transmission images and energy resolved Computed Tomography images reconstructed with different methods, such as integrating, counting or applying the Energy Weighting Technique and an analysis and comparison between them.

Acknowledgements: This work was partially supported by project PTDC/FIS-NUCL/2525/2014 and FCT (Lisbon) programs. L. F. N. D. Carramate, A. L. M. Silva and C. D. R. Azevedo are supported by the FCT (Lisbon) scholarships SFRH/BD/71429/2010, SFRH/BPD/109744/2015 and SFRH/BPD/79163/2011, respectively.

Keywords: Energy Resolving Detectors; Spectral X-ray Imaging; Spectral CT
Poster panel
(face) ID: 147

Poster Number:

Application of Deep Learning in Multi-Material Decomposition of Spectral CT (#2843)

Z. Chen1, L. Li1

1 Tsinghua Uvniersity, Engineering Physics, Beijing, China


Material decomposition is an important application of spectral Computed Tomography (CT). However, traditional post-process material decomposition algorithms are based on pixel-local usually, which cannot describe the scanned object exactly. In order to enlarge the visual field instead of considering the neighborhood of pixel only, we adopt deep learning technique to solve the multi-material decomposition problem. We build a convolutional neural network (CNN), who is a simplified version of VGG16 net, receiving a patch cut from the reconstruction image and predicting the multi-material decomposition result of its central pixel. Then we simulate some reconstruction images of spectral CT to generate the training data to train the network. After been trained by plenty of samples, by sliding the CNN “window” pixel by pixel, we can form the material decomposition result. Compared to the results of solving linear equations, the CNN method can reduce the MSE by 1~2 orders in the test samples. As the conclusion, we think CNN shows its effectiveness to solve multi-material decomposition problem, but there still remains many researches to work, such as how to get the source of training data, and how to balance between the priori knowledge and measurement.

Keywords: deep learning, spectral CT, multi-material decomposition, CNN
Poster panel
(face) ID: 150

Poster Number:

Material decomposition using the PIXSCAN-FLI spectral micro-CT (#2987)

N. Ducros2, O. Pivot2, M. Dupont1, O. Kochebina2, J. - M. Létang2, S. Rit2, J. Abascal2, F. Peyrin2, C. Morel1, Y. Boursier1

1 Aix-Marseille Univ, CNRS/IN2P3, CPPM, Marseille Cedex 9, France
2 Univ Lyon, INSA Lyon, UCBL, UJM Saint Etienne, CNRS, INSERM, CREATIS UMR 5220 U1206, Lyon, France


The advent of hybrid pixels, a technological breakthrough that makes possible to capture spectral information in X-ray tomography, paves the way to material decomposition and quantification. In this paper, a two-material decomposition approach based on a calibration step is implemented. We present the first results obtained exploiting real data acquired by the PIXSCAN-FLI, a spectral micro-CT scanner that is equipped with an in-house photon-counting detector. Separation of Aluminium and PMMA in a small animal phantom is shown.

Keywords: Spectral CT, material decomposition, real data, PIXSCAN-FLI
Poster panel
(face) ID: 153

Poster Number:

Signal-preserving non-local noise suppression for photon-counting CT (#3182)

J. Harms1, L. Zhu1

1 Georgia Institute of Technology, Nuclear Engineering and Medical Physics Programs, George W. Woodruff School of Mechanical Engineering, Atlanta, Georgia, United States of America


With advancements in detector technology, photon-counting CT (PCCT) in a clinical setting will soon become a reality. Photon-counting detectors typically split an x-ray spectrum into discrete bins and then reconstruct multiple images independently, reconstructing each image with a fraction of the number of photons of an energy-integrated CT scan, leading to higher image noise.

We aim to suppress this noise by extracting the redundant structural information provided by PCCT via weighted averaging of pixels of similar materials. For each reconstructed CT image, an exponential model is used to calculate a similarity coefficient between pixels based solely on their CT values. These similarity values are calculated individually for each image and stored in a similarity matrix. A weighted average of these similarity matrices is then calculated, with weights proportional to the noise variance of the corresponding image. Noise suppression is achieved for each energy channel via multiplying the image vector by the similarity matrix, preserving the signal level of each image. We call this the noise variance weighted similarity method, or NVWSM.

For evaluation of NVWSM, noise-suppressed low dose images are compared to full dose images. For evaluation of CT value accuracy and CNR, NVWSM is compared to another contrast-enhancing method for multi-energy CT, the CNVRW method. Simulation studies show that NVWSM enhances low-dose images to equal or greater quality than images acquired with roughly 6 times greater dose. In experimental phantom studies, spatial resolution of 7 lp/cm is preserved, even at a noise reduction factor of 2.2, relative to high-dose image. Additionally, NVWSM shows an RMSE of 7 HU at a noise STD reduction factor of 1.8 relative to the high-dose image, while CNNVRW has an RMSE of 159 at the same noise level. Additionally, a NVWSM shows a 91% increase in average CNR for 6 materials as compared to the high-dose image.

Keywords: energy-resolved CT, photon-counting CT, non-local filtration, low-dose CT, noise suppression
Poster panel
(face) ID: 156

Poster Number:

Fully Data-Driven Respiratory Motion Correction for ECT under Non-Stationary Breathing (#1239)

R. Bastiaannet1, M. Verra1, M. M. A. Dietze1, M. A. Viergever1, H. W. A. M. de Jong1

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


Underlying current approaches to respiratory motion management is the implicit assumption that there is no change in breathing dynamics over time (i.e. no drifts or inter-cycle variance). This assumption is known to be invalid from respiratory motion modelling in radiotherapy planning optimization. It has been shown that breathing dynamics may change significantly, even on short time-scales. As a result, current motion management strategies may result in sub-optimal respiratory motion compensation.

As an alternative, we propose to explicitly model the dynamics of respiratory motion during the acquisition process as a time-series and optimize its parameters as part of an iterative reconstruction. The goal of this study was to explore the feasibility of optimizing a simple explicit motion model under realistic, non-stationary respiratory dynamics, using only raw PET data, and compare it to retrospective gating and normal reconstructions.

Our results show that optimizing this motion model is feasible, as the objective function is smooth and convex, with a global minimum. Secondly, the reconstructions show both low noise and little residual motion blurring. This is contrary to both the gated (high noise) and uncorrected (motion blur) reconstructions.

These results may provide a starting point for a novel and practical approach to respiratory motion management.

Keywords: respiratory motion correction, data-driven, motion compensation, pet
Poster panel
(face) ID: 159

Poster Number:

Novel material identification method using three energy bins of a photon counting detector taking into consideration Z-dependent beam hardening effect correction with the aim of producing an X-ray image with information of effective atomic number (#1518)

N. Kimoto1, H. Hayashi2, T. Asahara1, E. Tomita3, S. Goto3, Y. Mihara1, Y. Kanazawa2, T. Yamakawa4, S. Yamamoto4, M. Yamasaki4, M. Okada4, D. Hashimoto5

1 Tokushima University, Graduate School of Health Sciences, Tokushima-shi, Tokushima, Japan
2 Tokushima University, Graduate School of Biomedical Sciences, Tokushima-shi, Tokushima, Japan
3 Tokushima University, School of Health Sciences, Tokushima-shi, Tokushima, Japan
4 Job Corporation, Yokohama-shi, Kanagawa, Japan
5 R&D Direct, Sakai-shi, Osaka, Japan


Currently, practical methods for producing an image having effective atomic number information is being developed in the X-ray diagnostic region, and a few studies have utilized for dual energy CT equipment. We focused our attention on plain X-ray examination in which an effective atomic number image is hoped to be used to enhance its value and to carry out more precise clinical diagnosis. In this study, we aimed to propose a novel material identification method using the three energy bins of a photon counting detector. First, we simulated response functions of a multi-pixel-type CdTe detector using the Monte-Carlo simulation code EGS5. The ideal X-ray spectra were folded with the calculated response functions. Then, the X-ray spectra were divided into three energy bins and the products of linear attenuation coefficient and material thickness for low, middle, and high energy bins were derived from the difference of counts before and after the X-ray penetrated the material. In order to accomplish accurate material identification, beam hardening corrections for derived attenuation factors were individually applied; in order to consider the dependence of atomic numbers on the beam hardening corrections, we propose a new method in which the relationship between mass thickness multiplied by mass attenuation coefficient and attenuation factor was used. Then, using the corrected attenuation factors, normalized linear attenuation coefficients for lower and higher energy bins were calculated, and they were converted to the effective atomic numbers using the theoretically determined relationships. In the present study, ZLow and ZHigh were individually determined from the analysis of lower and higher energy bins. Here, if beam hardening corrections were properly corrected, ZLow and ZHigh become an equal value which is also consistent with the true value. Using virtual materials of Z=5-13 with mass thickness of 1-10 [g/cm2], we confirmed that our method works properly.

Keywords: Beam hardening correction, Photon counting technique, Material identification
Poster panel
(face) ID: 162

Poster Number:

Simultaneous Correction of motion and metal artifacts in head CT scanning (#1607)

T. Sun1, J. Nuyts1, R. Fulton2, 3

1 KU Leuven, Department of Imaging and Pathology, Leuven, Belgium
2 University of Sydney, School of Physics, Sydney, Australia
3 Westmead Hospital, Department of Medical Physics, Westmead, Australia


Often patients are found to have metal implants or devices in a head CT scan. The high-density materials create severe artifacts because of beam hardening and scattering effects. Further complication happens when the patient moves at the time of the scan, even if only slightly. The compound artifacts in the reconstructed images are likely to degrade the accuracy of the diagnosis. In this study we propose an approach to correct these compound artifacts based on the knowledge of the measured projections and known x-ray spectrum. We did this in two steps: (1) in an iterative scheme, rigid motion was estimated and a preliminary reduction of the metal artifacts was performed; (2) the estimated motion was modeled in a full polychromatic reconstruction, in which the starting image was the preliminary corrected image from step 1. The polychromatic reconstruction has the capability to further reduce the compound artifacts. We demonstrate  the proposed approach in a simulation study. The image was improved significantly after correction, compared with the one with no correction.

Keywords: metal artifacts, motion correction, helical CT
Poster panel
(face) ID: 165

Poster Number:

Comparison of Partial Volume Correction Techniques for Lesions near High Activity Regions (#1786)

M. I. Akerele1, P. Wadhwa1, S. Vandenberghe2, C. Tsoumpas1

1 University of Leeds, Div. of Biomedical Imaging, Leeds, United Kingdom of Great Britain and Northern Ireland
2 Ghent University, Dept. of Medical Imaging, Gent, Belgium


Partial volume effect (PVE) is a major factor affecting PET quantification as it leads to uptake miscalculation in small lesions, thereby resulting in inaccurate diagnosis and staging. Many techniques are being used to correct for PVE but there is still need for better and more accurate results in high activity regions. This study aims at investigating and correcting spill-out from high activity regions to the surrounding lesions. A hot region was simulated using the XCAT2 phantom with different activities allocated to simulate uptake increase during a typical FDG PET examination. Spherical lesions with diameters 6-12mm and fixed activity were randomly placed at variable distances from the hot region to investigate spill-out effect as a function of lesion size and distance from the hot region. Validation was done using a physical phantom with cardiac insert and synthetic lesions were added for quantitative evaluation. Analytical simulations were carried out using Software for Tomographic Image Reconstruction (STIR) package. All reconstructions were done using OSEM, incorporating all corrections. PVE correction was done using three techniques: (1) incorporating point spread function in the reconstruction (PVC_PSF), (2) Geometric Transfer Matrix technique (PVC_GTM) and (3) a recently proposed reconstruction-based background correction (PVC_BC). Region of Interest (ROI) analysis was carried out using standardized uptake values, contrast-to-noise ratio and recovery coefficient as figures of merit. The study revealed a significant spill-out effect for small lesions and those relatively close to the hot region. The variation in lesion uptake is 61.82% without correction, 42.47% with PVC_PSF, 16.96% with PVC_GTM and 1.51% with PVC_BC.  There is a great uptake underestimation in lesions less than 12mm but with correction, there is a better recovery from lesion size 8mm and less bias. Hence, PVE correction enhances stable quantification, with PVC_BC giving the best result.

Keywords: PET, Reconstruction, Partial volume effect, Correction
Poster panel
(face) ID: 168

Poster Number:

Agarose/sucrose gel breast phantom for validation of quantitative methods in DWI MR (#2040)

F. Gallivanone1, M. Interlenghi1, I. Carne2, D. Fantinato2, I. Castiglioni1

1 IBFM - CNR, Milano, Italy
2 IRCCS Fondazione S. Maugeri, Medical Physics Unit, Pavia, Italy


Physical and digital phantoms are useful instruments for the validation of image processing methods allowing the simulation of various realistic scenarios with the knowledge of gold standard (GS) parameters, such as oncological lesion characteristics. This work aims presenting a strategy to create disposable realistic breast phantoms, suitable for Magnetic Resonance - Diffusion Weighted Image acquisitions (MR-DWI) and characterized by synthetic lesions with irregular shape and signal. Diffusion materials were prepared using a mixture of agarose and sucrose solidified gels. Diffusion characteristics were evaluated in cylindrical phantoms. Realistic oncological lesions were created with irregular shapes using 3D-printed plastic molds filled with the gel mixture at a sucrose concentration mimicking the diffusion characteristics of high cellularity tissues. Breast cancer phantoms were manufactured suspending synthetic lesion in a uniform gel mixture background, in order to simulate realistic oncological condition.  As proof of concept of the usefulness of this phantom, validation of two different segmentation methods was performed.

Keywords: breast phantom; MR-DWI
Poster panel
(face) ID: 171

Poster Number:

Evaluation of Respiratory Gating Schemes for Cardiac SPECT Using a Population of Phantoms (#2384)

D. Zhang1, M. Ghaly2, Q. Zhang1, G. S. P. Mok1, 3

1 University of Macau, Biomedical Imaging Laboratory (BIG), Department of Electrical and Computer Engineering, Faculty of Science and Technology, Macao, China, Macao Special Administrative Region
2 Johns Hopkins University, The Russell H Morgan Department of Radiology and Radiological Science, Baltimore, Maryland, United States of America
3 University of Macau, Faculty of Health Sciences, Macao, China, Macao Special Administrative Region


Respiratory gating was proposed to reduce motion blur in cardiac SPECT. This study aims to evaluate 3 respiratory gating methods for cardiac SPECT. We used a population of 4D XCAT phantoms to simulate 10 patients injected with Tc-99m-MIBI, 6 with normal heart and 4 with heart defects. Phantoms varied in gender, body size, heart size, activity distribution and breathing pattern. For each phantom, 3 respiratory cycles were divided into 288 frames which were grouped to 6 gates with: (1) equal amplitude gating (AG); (2) amplitude gating with equal counts (CG) and (3) equal time gating (TG). Average activity and attenuation maps in each gate represented gated SPECT and CT. The end-expiratory and end-inspiratory phases were used as references. Both noise-free and realistic noisy 120 projections were generated from RAO to LPO using an analytical LEHR projector and reconstructed by OS-EM with 200 and 30 updates respectively, using gated CT for attenuation correction. Reconstructed images in each gate were registered to references using affine+b-spline method, then averaged and reoriented to generate short-axis images and polar plots. Relative difference (RD) of the average image intensity was computed at each segment in 17-segment analysis based on the references. A region-of-interest was chosen in the uniform septal region of the polar plots from each gate to calculate the normalized standard deviation (NSD) as the noise index. AG and CG showed less defect blurring from visual assessment. For all phantoms, the maximum RDALL of AG, CG and TG were 5.54±3.75%, 5.47±4.05% and 9.09±4.73% for noisy images registered to end-expiration, and were 5.36±3.14%, 6.80±3.16% and 9.55±5.75% when registered to end-inspiration. The NSD fluctuation in different gates was less for CG and TG than AG. We conclude that the CG shows better performance in motion reduction and noise. Using end-expiration as the registration reference is slightly better than end-inspiration.

Keywords: SPECT, Myocardial perfusion, Respiratory gating
Poster panel
(face) ID: 174

Poster Number:

Seated vs. supine: optimum measurement pose for brain-dedicated PET (#2576)

Y. Iwao1, H. Tashima2, E. Yoshida1, H. Wakizaka1, F. Nishikido1, T. Yamashita1, T. Yamaya1

1 ATOX Co. Ltd, Tokyo, Japan
2 National Institute of Radiological Sciences, Chiba, Japan


In some recently developed brain-dedicated PET systems, a seated position is selected for the patient. The PET scanners with the seated measurement style have a merit of allowing system downsizing. However, the effect of the seated position for head motion suppression compared with the conventional supine position is not clear. Although motion correction methods have been developed, they are not in practical use at a routine level. Even when using the motion correction method, it is desirable to make the head motion as small as possible. In this study, we developed a contactless head motion tracking system, and we conducted a volunteer study of the head motion tracking to assess the optimum measurement position for the brain PET. We developed the tracking system using the Microsoft Kinect sensor as a range sensor. As a result of the accuracy evaluation, translation accuracy of about 1 mm and rotation accuracy of less than 1 deg were achieved. In the volunteer study, we measured the head motion of volunteers with supine, normal sitting, and reclining. Additionally, we measured these positions with and without head fixation. As a result, the normal sitting position without head fixation had the largest head motion, and the reclining position had smaller motion than the supine position. The head fixation was effective for motion suppression in all cases. The lowest head motion was achieved with the reclining and supine positions (average displacement in 60 s was about 0.5 mm). In order to assess the effect of the head motion, we added the motion information obtained with the volunteer study to brain phantom data measured with our helmet-chin PET prototype. In the case of the reclining with head fixation, we could obtain reconstruction images of equal quality to the case without motion even though we did not apply any motion correction. We concluded that the reclining position had a high head motion suppression effect equal to or better than that of the supine position.

Keywords: Positron emission tomography, brain PET, head motion tracking, optimum measurement pose
Poster panel
(face) ID: 177

Poster Number:

New Imaging Method of Positrons Leaving the Source - Application for PET/MR hybrid Scanners – (#2818)

J. J. Scheins1, L. Tellmann1, C. Lerche1, N. J. Shah1

1 Forschungszentrum Jülich GmbH, Institute of Neuroscience and Medicine (INM-4), Jülich, North Rhine-Westphalia, Germany


The 11CO2 tracer is of high interest to study carbon transport in living plants. However, conventional PET images of plant leaves suffer from the poor positron annihilation probability in the targeted structures, thus giving strong bias in reconstructed images. Here, PET/MR hybrid scanners offer a new method to detect escaping positrons as 2D projections by exploiting the directed travelling of positrons parallel to the present B0 field of the MR magnet system. Due to the Lorentz force all emitted positrons propagate on spiral trajectories parallel to the B0 field with tiny gyroradii. The introduction of a thin absorber screen within the field-of-view, forces the annihilation process for almost all incoming positrons. The annihilation photon pairs can be registered as coincidences and thus directly indicate the positron impact point on the screen by interpolating the Lines-of-Response onto the screen. In this way, 2D images are obtained and the amount of positrons leaving the source can be dynamically quantified without tomographic reconstruction.

In this paper, we present preliminary results as proof of concept for the new method. We have applied a first test measurement on the Siemens 3T MR/BrainPET system using a 18F-labeled sheet of paper. Since the positron source is a thin sheet of paper, a large fraction of positrons can leave the source without undergoing an annihilation. These positrons can be imaged according to the proposed method. The obtained image nicely represents the original distribution. Thus, the new method supplements the standard PET method to achieve an unbiased quantification (when combining both methods) of thin objects. Also dynamical studies of plant metabolism are possible.The resolution of the images are basically limited by the resolution of the BrainPET (approx. 3mm), whereas the mean gyroradii are much smaller (approx. 0.6 mm for B0=3T). First, experiments using plant leaves using 11CO2 as tracer are in preparation.

Keywords: PET quantification, 2D Direct Positron Imaging, PET/MR hybrid systems
Poster panel
(face) ID: 180

Poster Number:

Human Receiver Operator Characteristic Confirmation of Potential for Radiation Dose Reduction with Improved Reconstruction for Cardiac SPECT (#3008)

P. H. Pretorius1, A. J. Ramon2, M. A. King1, A. Könik1, S. T. Dahlberg1, M. Parker1, K. L. Johnson1, J. Jang2, M. N. Wernick2

1 University of Massachusetts Medical School, Department of Radiology, Worcester, Massachusetts, United States of America
2 Illinois Institute of Technology, Department of Electrical and Computer Engineering, Chicago, Illinois, United States of America


Dose fractionation is a popular proposed method to lower the radioactive exposure to patients undergoing myocardial perfusion imaging (MPI) SPECT/CT. Recently, we optimized reconstruction strategies employed during dose fractionation in MPI using polar maps in combination with the calculation of total perfusion deficit (TPD) scores employing hybrid defects of various sizes. Although observed to agree well with experienced observers, TPD scores have not been used to judge the impact of reduced dose SPECT imaging for standard two-headed SPECT systems. Thus, the aim of this study was to confirm that the optimized reconstruction strategies are indeed ranked accordingly by TPD in comparison to human observers reading hybrid cardiac defects studies with known truth. We setup our human observer study with 126 test cases (63 with defects inserted) and 24 training cases (12 with defects). Four observers participated by reading three different reconstruction strategies, ordered-subset expectation maximization with the original dose, (OSEM100), OSEM with only 25% of the dose (OSEM25), and filtered backprojection with the original dose (FBP100). Before commencing the reading of the actual test cases, training (24 image sets) with feedback were done for each of the reconstruction strategies. The test cases were read in nine randomized sessions (42 test cases per session, half with defects), three for each reconstruction strategy with 8 training images as a warm-up with feedback (50 image sets per reading session). The order of reading the different sets as well as the order of the test cases were different for each observer. ROCKIT (University of Chicago) software was used to analyze the observer data.  Areas under the curve (AUC) values of 0.823, 0.769, and 0.656 were recorded for OSEM100, OSEM25, and FBP100 respectively. The ranking of the reconstruction strategies is the same as for the TPD scoring method, while the observers fared worse scoring FBP100

Keywords: SPECT/CT, dose optimization, cardiac perfusion imaging
Poster panel
(face) ID: 183

Poster Number:

Real-Time Accurate Rebinning of PET Data Based on the Pseudo-Inverse of the Axial System Matrix (#3149)

A. López Montes1, P. Galve1, J. M. Udías1, J. L. Herraiz1

1 University Complutense of Madrid (UCM), Nuclear Physics Group (GFN), MADRID, Spain


Real-time Positron Emission Tomography (PET) imaging is a very useful tool for many different applications, providing first shot images, medical interventions guidance, information about patient position and motion, and useful feedback during the  acquisition. Fully-3D iterative reconstruction methods in PET provide accurate results but they cannot be used for real time imaging due to their long computational time. On the other hand, analytical methods are fast enough for real time applications, but they exhibit poor resolution and artifacts with noisy or incomplete data. We propose an alternative reconstruction method based on the pseudoinverse of the System Response Matrices (SRM), which can be fast and accurate. Due to the large size of the SRM, instead of pseudoinverting the fully 3D-SRM, only the axial part of the SRM is used to rebin the 3D data into 2D datasets with accurate resolution recovery in the axial direction. The pseudoinverse of the axial SRM was regularized with different methods, including one equivalent to the linear iterative algorithm SART (Simultaneous Algebraic Reconstruction Technique). The resulting rebinned2D datasets can be reconstructed with standard analytical methods such as Filtered Back-Projection (FBP), or using another pseudoinverse.Our results show how this method provides better results (1.2 mm of axial resolution) than the standard Single Slice ReBinning (SSRB), even for centered sources, and with computational time (500 ms) compatible with real-time applications.

Keywords: Positron Emission Tomography, Data rebinning, Pseudo-inverse, Real-Time
Poster panel
(face) ID: 186

Poster Number:

Crystal-based Deadtime Correction for Positron Emission Tomography (PET) in Continuous Bed Motion (CBM) Studies (#3235)

M. Aykac1, V. Y. Panin1

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


It is essential to have proper data correction methods for normalization and scatter prior to the reconstruction process to obtain accurate quantitative PET images. Normalization includes deadtime in the detector and the coincidence electronics as part of its components. The scope of this study is to investigate a new deadtime model at the crystal level applicable for both step-and-shoot and CBM acquisitions. Deadtime correction model for commercial PET scanners is implemented typically based on a block or group of blocks due to its simplicity. However, the count performance of each crystal within the detector block differs due to detector design and scanner electronics. In our investigation, we would like to discuss a method which calculates the crystal count rate from randoms sinogram and characterizes a coincidence response to accommodate the deadtime in the electronics. By using the symmetries within the detector, one can create simplified deadtime model at the crystal level.

Keywords: Deadtime, PET
Poster panel
(face) ID: 189

Poster Number:

Improving Scatter Correction for Ga-68 PSMA PET Studies (#3380)

I. Hong1, C. Michel1, S. Nekolla2

1 Siemens Medical Solutions USA, Inc, Molecular Imaging, Knoxville, Tennessee, United States of America
2 Technische Universität München, Dept. of Nuclear Medicine,, München, Germany


Ga-68 Prostate Specific Membrane Antigen (PSMA) is currently used in prostate cancer PET imaging. The resulting images corrected for scatter using standard/distributed software show high uptakes in the kidneys and the bladder which could produce a halo- artifact which potentially masks metastasis lesions at the level of any tracer accumulating structure such as kidney or bladder. This artifact was attributed to scatter over-correction when neglecting the prompt gamma contribution. In previous work, we have shown clinically relevant improvement in selected clinical studies, when adding a prompt gamma contribution to our standard single-bed position 3D scatter correction. The latter is based on relative scatter estimation using single scatter simulation followed by axial scaling. The axial scaling uses scatter tail fitting in an attempt to compensate both the effect of ignoring multiple scatter and out-of-axial-FOV activity.

This work demonstrates further reduction in residual artifacts by replacing the one-bed SSS with a whole-body (multi-bed) one. When considering a long virtual scanner, the activity and the attenuation distributions are almost inside the axial FOV and the axial scaling accounts only for neglecting multiple scatter. The timing performance of the new algorithm is presented. An advantage of the multi-bed scheme is the uniformity of the 3D plane scaling factors especially in the bed overlap region. The new method is tested on clinical data with a Siemens mCT system and acquired by Technische Universität München, Germany.

Keywords: PSMA, Ga-68, Prompt gamma, PGC, Scatter
Poster panel
(face) ID: 192

Poster Number:

Investigation of Rodent Models for Applications in Model-based, Awake Animal Motion Tracking (#3678)

P. Soubiran1, C. Parl1, B. Hirt2, B. J. Pichler1

1 Eberhard Karls University of Tuebingen, Werner Siemens Imaging Center, Tuebingen, Baden-Württemberg, Germany
2 Eberhard Karls University of Tuebingen, Institute for Clinical Anatomy and Cell Analytics, Tuebingen, Baden-Württemberg, Germany


Preclinical non-invasive imaging of freely moving rodents is of particular interest in neurology, behavioral science and pharmacokinetics of imaging tracers. However, to achieve valid quantification of the imaging data, an accurate tracking system must be developed that can provide real-time motion data, and also be integrated into the imaging data. Unlike marker-based tracking, which affects the rat, model-based methods depend on the accuracy and articulate information of the 3D model. 3D scanners, which rely on static targets and enable texture, offer no internal details (articulate motion and anatomy). As live animals are mobile, this study evaluates the feasibility of using fixated rats for 3D imaging as well as building a model which integrates internal structures based on medical imaging.

Formalin and plastinated rats were studied using magnetic resonance imaging (MRI) and computed tomography (CT); a polygonal 3D-model was then created. Analysis was done on internal structure segmentation and accuracy. Segmentation was determined by the ability to discern three internal structures: the skeleton, brain, and heart. The accuracy was measured by comparing the distances between three landmarks on the 3D-models with those on the rat, considered to be the ground truth.

Sufficient segmentation of the internal structures was only possible for both rats using CT imaging. The CT 3D-models were most accurate, with an average error of 3.9%, compared to 8.2% for MRI. The plastinated rat was most accurate in both imaging modalities, with an average error of 3.7% compared to 8.3% for formalin.

This study aimed to determine the best combination of rat and imaging modality to produce a 3D- model with internal structure and articulate motion. This model will be used with 3D scanners to help train a model-based markerless motion tracking system. After analyzing structure segmentation and accuracy, it was determined that a CT imaged plastinated rat produces the best 3D-model. 

Keywords: awake animal, motion tracking, medical imaging, MRI, CT
Poster panel
(face) ID: 195

Poster Number:

Developing an Expert System to Improve Lesion Quantification for Personalized PET Imaging (#3870)

Y. Li1, M. E. Daube-Witherspoon1, S. Matej1, S. D. Metzler1

1 Penn, Philadelphia, PA, United States of America



The measurements of lesion standardized uptake value (SUV) in clinical PET studies are affected in a complicated way on many aspects of the data, including---but not limited to---count level, lesion size, lesion shape, lesion location, background level and structure, and patient size. Optimized reconstruction algorithms and their parameters can provide substantial improvements in lesion quantification for personalized PET imaging. Full optimization of reconstruction parameters can be too complicated/tedious for human beings, thus we propose an expert system to optimize these parameters and to improve lesion quantification. Two core techniques are explored: 1) synthetic lesion embedding technique and 2) system local impulse response modeling. We experimentally acquire list-mode data from physical radioactive spheres of known activity concentration with different sizes at different locations using the same PET scanner as was used for acquiring the patient data. We can then synthetically embed the spheres by merging the sphere list-mode data into patient data. We also use statistical bootstrapping method to generate multiple replicates of patient data with embedded lesions to determine the variance of the estimated SUV. We also use local impulse response (LIR) to characterize the local properties and responses of an imaging system, which can be used as a criterion for optimizing the reconstruction parameters. We have performed simulated phantom studies to confirm the use of bootstrapping of clinical data to determine the variance of SUV, and the use of LIR to determine the recovery coefficients and the bias of SUV of each lesion with different size. We are currently integrating the expert-system tools with graphical user interface to allow the execution and visualization of ensembles of lesions with the use of different reconstruction algorithms and parameters. The end result will be patient- and lesion-specific corrected SUVs with estimated uncertainty.

Keywords: Expert system, lesion embedding, local impulse response (LIR), image reconstruction, positron emission tomography (PET)
Poster panel
(face) ID: 198

Poster Number:

Fan-beam X-ray Fluorescence Computed Tomography (XFCT) with Gold Nanoparticles (#4016)

W. Fang1, L. Li1, S. Zhang1

1 Tsinghua University, Engineering Physics, Beijing, China


X-ray fluorescence computed tomography (XFCT) is a promising imaging modality. It calculates the spatial distribution and concentration of various elements by detecting the fluorescent signals excited by incident x-rays. XFCT has a promising future in the fields like molecular imaging in fundamental studies, drug development and clinical experiments. XFCT was used to be performed on synchrotron radiation source. However, the synchrotron radiation facility is not available for most researchers. So people are exploring the feasibility of performing XFCT imaging using benchtop X-ray tube under a typical laboratory setting. In this paper, we demonstrated a simulation study of XFCT reconstruction. We use the SpekCalc software to simulate the polychromatic source and a phantom composed of PMMA and four gold nanoparticles solver with different concentrations was imaged. We use photon counting detectors to detect the fluorescent signals. The scatter correction information is accessed by scanning the phantom with gold nanoparticle solution removed a second time.

Keywords: X-ray fluorescence computed tomography (XFCT), scatter correction, attenuation correction, image reconstruction
Poster panel
(face) ID: 201

Poster Number:

Use of characteristic functionals to analyze molecular images in targeted cancer therapy (#4226)

H. H. Barrett1, 2, K. J. Myers4, E. Clarkson1, 2, N. P. Henscheid3

1 University of Arizona, Dept. of Medical Imaging, Tucson, Arizona, United States of America
2 University of Arizona, College of Optical Sciences, Tucson, Arizona, United States of America
3 University of Arizona, Program in Applied Mathematics, Tucson, Arizona, United States of America
4 Food and Drug Administration, Center for Devices and Radiological Health, Silver Spring, Maryland, United States of America


Images from positron emission tomography, single-photon emission computed tomography and other molecular imaging modalities are commonly used in clinical oncology, but it is not obvious which studies are most informative, how best to extract information from the images or how to relate the results to therapeutic efficacy.  The objective of this paper is to present a unified mathematical framework for answering these questions.  Our premise is that the underlying object that produces a molecular image is (a) related to the patient’s physiology; (b) a function of position within the body and time; (c) spatially and genetically heterogeneous, and (d) random (unpredictable). We refer to such objects as physiological random processes. All statistical properties of a random process are captured in its characteristic functional, which can be regarded as the infinite-dimensional counterpart of the characteristic function, well known in probability theory.  In principle we can acquire several molecular images on the same patient, thereby observing multiple physiological random processes that can interact with each other.  To describe these interactions, we have derived evolution equations that show how the characteristic functionals of these processes evolve jointly in time.  The forward problem, given these joint characteristic functionals, is to compute the statistics of the molecular images.  The inverse problem, new in the field of image science, is to start with a collection of molecular images and estimate moments or other properties of the random physiology. If the object is a tumor and the image data include dynamic images of a radiolabeled drug or a validated surrogate, our equations provide a patient-specific method for estimating the probability of tumor response as a function of the administered dose of the drug.

Keywords: Random process, characteristic functional, emission computed tomography, chemotherapy
Poster panel
(face) ID: 204

Poster Number:

Evaluation of a strategy to find personalized, patient-specific injected activity levels for SPECT-MPI (#1931)

A. Juan Ramon1, Y. Yang1, P. H. Pretorius3, P. Slomka2, K. L. Johnson3, M. A. King3, M. N. Wernick1

1 Illinois Institute of Technology, Medical Imaging Research Center, Chicago, Illinois, United States of America
2 Cedars-Sinai Medical Center, Los Angeles, United States of America
3 University of Massachusetts Medical School, Department of Radiology, Worcester, United States of America


Reduction of radiation dose in SPECT-myocardial perfusion imaging (MPI) has become an important objective. This paper reports a new strategy to find the optimal dose level for a given patient so that he receives the minimum radiation dose required to maintain the diagnostic accuracy obtained at current clinical dose levels. The ultimate aim is to provide an initial framework for prediction models that can estimate personalized dose levels based on patient attributes, prior to the imaging procedure in the clinic. Using clinical SPECT-MPI data with realistic simulated lesions we first computed perfusion-defect detection performances using the same dose level for all patients (global dose). For that, each study was evaluated using the total perfusion deficit (TPD) score provided by the clinically validated Quantitative Perfusion SPECT (QPS) software package and the defect-detection performances were obtained from a receiver operating characteristics (ROC) study based on the patient scores. Then, personalized dose levels were found for each patient by determining at what reduced dose level the TPD score deviated from the full clinical-dose TPD. Finally, the perfusion-defect detection performance was computed using the personalized dose levels and compared to those obtained from global dose. Preliminary results suggest that further tracer dose (and hence radiation dose) reduction can be achieved by personalizing the dose levels compared to the global (non-patient specific) dose reduction approach.  We found this to be the case for both filtered backprojection (FBP) and ordered subsets expectation-maximization (OS-EM).

Keywords: SPECT-MPI, dose reduction, personalized dose
Poster panel
(face) ID: 207

Poster Number:

Imaging Salt Transport in Plants Using PET: A Feasibility Study (#2485)

G. Ariño-Estrada1, G. S. Mitchell1, P. Saha2, A. Arzani3, 4, S. R. Cherry1, E. Blumwald3, A. Z. Kyme5

1 University of California Davis, Department of Biomedical Engineering, Davis, California, United States of America
2 University of Pennsylvania, Philadelphia, Pennsylvania, United States of America
3 University of California Davis, Department of Plant Sciences, Davis, California, United States of America
4 Isfahan University Technology, Department of Agronomy and Plant Breeding, College of Agriculture, Isfahan, Iran (Islamic Republic of)
5 University of Sydney, School of AMME, Faculty of Engineering & IT, Sydney, Australia


We report on the dynamic study of sodium transport in plants with PET. PET is a good candidate to complement the study of salt transport and regulation mechanisms in plants as it offers uptake information from intact plants over several time points and a precise map of the sodium distribution in the plant. We have incubated 24 green millet plants from two different varieties in a growth solution with 22Na and have imaged them with a small animal PET scanner at 5 different time points after the start of the incubation. We also measured the activity in each plant and time point with a well counter. We overlaid the PET images with photographs from the plants and the 22Na distribution maps match well with the physical attributes of the plants. Applying ANOVA analysis on the data demonstrated a clear correlation between activity in the plants and time and plant variety and activity (both p-values below 0.01). These results prove that we can evaluate the uptake rate of each plant and that we can recognize different plant varieties based on the uptake rate. Based on these results we conclude that PET can become a unique tool to help to understand salt transport mechanisms in plants and, hence, facilitate the recognition of salt tolerant varieties of plants that fit best to high salinity soils that currently cannot be used for agricultural purposes.

Keywords: PET, sodium transport, plant imaging, sodium-22
Poster panel
(face) ID: 210

Poster Number:

Comparison of Time Resolution Measurement Methods (#3626)

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

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


Recently a method has been proposed to measure time resolution of PET scanners from NEMA NEC or scatter phantom so that time resolution measurements can be standardized. In this article we evaluate this method by comparing the time resolution measurements obtained using this method with those obtained from line-source, for a Siemens next generation SiPM PET/CT prototype scanner with high time resolution. We verified that the two methods used to measure time resolution yielded very similar results, The Siemens SiPM PET/CT prototype scanner showed a very small change in time resolution with count rate. Time resolution measured at low count rate was about 250 ps for the SiPM PET/CT prototype scanner, using either method. Based on this work, we believe that the time resolution of a TOF PET scanner can be correctly measured using data acquired during NEMA NEC measurements.

Keywords: NEMA; Time Resolution; PET scanners
Poster panel
(face) ID: 213

Poster Number:

Assessment of PET Amyloid Image Quantification Differences by Varying the Reconstruction Protocol (#3931)

A. M. Smith1, D. Matthews2, R. D. Andrews2

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


One goal of PET amyloid imaging is to provide a quantitative estimate of amyloid plaques that are often present in patients with Alzheimer’s disease (AD) and other forms of dementia.  This information is used to both stage and monitor patient treatment.  The standard uptake value ratio (SUVR) is the most often used estimate, and is defined as the PET activity in a target region divided by the PET activity in a reference region.  Several factors influence the SUVR, and this work focuses on how changes in the reconstruction protocol affect the PET activity in both the target and reference regions.  Methods:  PET/CT florbetapir data from 15 subjects who had varying levels of cortical amyloid deposits were reconstructed using 84 different protocols.  The reconstruction variable space that was spanned included the algorithm (FBP, OSEM and PSF), image voxel size, filter kernel size, TOF off or on, and in the case of the iterative algorithms, the number of iterations.   The percent change in activity between the standard reconstruction (OSEM, 400x400 matrix, 4 iterations, 24 subsets and 5 mm Gaussian filter) was computed for all reconstructions.  This work showed that repeatable quantification for PET amyloid imaging can only be achieved if the PET/CT scanner is calibrated properly, the attenuation map is accurate (e.g. no truncation of head holder) and the exact same reconstruction protocol is used if the SUVRs are to be compared across different longitudinal acquisition times for the same patient.

Keywords: PET Amyloid Imaging, Alzheimer's Disease, PET Quantification
Poster panel
(face) ID: 216

Poster Number:

Determining optimal timing allocation of an adaptive imaging system which employs data from both single and multi-pinhole apertures (#4234)

A. Lin1, M. Kupinski1, N. Ghanbari1, L. R. Furenlid1, X. Li1, C. Chaix1

1 University of Arizona, College of Optical Sciences, Tucson, Arizona, United States of America


As of recently, the idea of adaptive imaging has gained greater interest within the medical imaging community.  The overarching belief is that imaging quality and diagnostic capabilities can be improved on a case by case basis by modifying certain system parameters.  One such system is the AdaptiSPECT system, which employs a highly modular aperture design, capable of switching between single pinhole and multi-pinhole configurations within a single image acquisition.  The benefits and drawbacks of both pinhole configurations are well documented.  Under single pinhole operation, unambiguous spatial information is achieved at the cost of lower sensitivity, whereas with multi-pinhole apertures, the increase in sensitivity can be accompanied by image multiplexing which can lead to image artifacts.  By exploiting the modularity of AdaptiSPECT’s aperture, we propose an imaging scheme which collects image data under both pinhole configurations during one imaging session.  This allows us to capitalize on the benefits of both configurations with fewer disadvantages.  Since the total time of an imaging session is fixed in our studies, we are essentially trying to determine how system performance is affected with different time allocations between these two pinhole configurations.  From our simulation studies, which use a MOBY mouse phantom, we demonstrate a method to test how different observer models utilize the combined image data of both single and multi-pinhole configurations, and optimize the timing sequence based on these results.  Additionally, we illustrate some of the inherent limitations of using simplistic object models for optimizing system performance based on task-based measures of image quality.  We believe that this work can extend to other adaptive systems and will serve as a precursor to additional studies in performing real time system optimization based on simple object features.

Keywords: Adaptive Imaging, Timing Allocation, Multiplexing, Small Animal SPECT
Poster panel
(face) ID: 219

Poster Number:

Evaluation of HYPR-OSEM Using Experimental Phantom and Clinical Patient Data (#1294)

J. - C. (. Cheng1, 2, J. Matthews2, R. Boellaard3, 4, J. Anton-Rodriguez2, 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 evaluations of our newly developed HYPR-OSEM algorithm using experimental phantom and clinical patient data. HYPR-OSEM is an iterative reconstruction method which incorporates HighlY constrained back-PRojection (HYPR) de-noising directly within the widely used OSEM algorithm. Our previous work demonstrated that HYPR-OSEM can achieve noise reduction without degrading accuracy in terms of resolution and contrast, and it can attain better precision than OSEM with similar accuracy and better accuracy than filtered OSEM with similar precision based on simulation results. In this work, further evaluations have been conducted using experimental phantom and clinical patient data acquired on the High Resolution Research Tomograph (HRRT). The regional contrast recovery coefficient (CRC) vs image voxel noise, coefficient of variation (COV) in CRC vs bias in CRC, and root-mean-squared-error (RMSE) in CRC for various sizes of hot and cold regions (based on 50 realizations of both high and low count experimental phantom data) were compared across all forms of HYPR-OSEM and OSEM with and without a post reconstruction filter. In contrast to our previous simulation results, higher noise reduction was achieved by HYPR-OSEM for the HRRT data. HYPR-AU-OSEM showed the lowest noise-induced bias at low count level, the lowest RMSE in CRC, and the most stable performance in COV or reproducibility of CRC (i.e. the least sensitive to the number of iterations). With regard to the patient study, one can observe that HYPR-AU-OSEM images are less noisy than OSEM images, while HYPR-AU-OSEM images show higher contrast in both hot and cold regions as compared to the filtered OSEM. Additionally, cold regions are also more visible in the HYPR-AU-OSEM images. Although there is no gold-standard for the case of the patient study, these observations agree with those observed previously from simulations and experimental phantom studies. Further optimizations are under investigation.

Keywords: PET reconstruction, De-noising, HYPR-OSEM
Poster panel
(face) ID: 222

Poster Number:

Fourier-Domain Analysis of the Iterative Landweber Algorithm (#1384)

G. L. Zeng1, 2

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


This paper proposes a novel method of using the frequency-domain transfer function to investigate the property of an iterative algorithm for minimizing a quadratic objective function. This paper focuses on a general two-dimensional (2D) tomography problem, which can be X-ray computed tomography (CT), positron emission tomography (PET), and single photon emission computed tomography (SPECT). Two questions regarding to the linear iterative Landweber algorithm are considered. The first question is whether stopping early is equivalent to getting a minimum-norm solution. The second question is whether the low frequency components always converge first. Our answer to these two questions is No.

Keywords: iterative reconstruction, tomography, Fouier analysis
Poster panel
(face) ID: 225

Poster Number:

Machine Learning: Any Image Reconstruction Algorithm Can Learn by Itself (#1386)

G. L. Zeng1, 2

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


In the traditional machine learning sense, a neural network structure with single or multiple layers is required and the weights at the neurons are learned from training sets. The main application of machine learning is pattern classification. The author’s main research area is image reconstruction. Any image reconstruction or image processing algorithm has some parameters to be adjusted according to the tasks. For example in Bayesian algorithms, the weighting factor for the Bayesian influence and some thresholds for non-linear constraints are tedious and time-consuming to determine. In this preliminary research, we argue that any image reconstruction algorithm or any image processing algorithm can learn to find its optimal parameters by using the training sets. The neural network structure is not necessary.

Keywords: Machine learning, Image Reconstruction, Non-linear algorithm, Low-dose CT
Poster panel
(face) ID: 228

Poster Number:

A Solution for Scaling Problem in Joint Estimation of Activity and Attenuation (#1388)

T. Kobayashi1, K. Kitamura1

1 Shimadzu Corporation, Technology Research Laboratory, Kyoto, Japan


In time-of-flight PET reconstruction, joint estimation (JE) algorithms of activity and attenuation, such as maximum-likelihood attenuation correction factors (ML-ACF) and maximum-likelihood activity and attenuation reconstruction (ML-AA), cannot provide quantitative attenuation information due to the scaling problem. Then, some post-processing is essentially required to make the estimated attenuation information quantitative and to get quantitative activity images.

In this paper, we propose a method to estimate a quantitative attenuation image from an attenuation factor scaled by an unknown factor. The proposed method is based on the mathematical relationship between the scaled (i.e., non-quantitative) attenuation factors and the true attenuation image, and corrects the non-uniform offset in the attenuation image caused by the scaling uncertainty.

The proposed method assumes that the supports of activity and attenuation distributions are identical to ensure the attenuation factor sinogram estimated by ML-ACF not to be truncated, and the attenuation coefficient of a portion of the subject is known. Under these assumptions, we constructed the processing flow from the data acquisition to the quantitative activity reconstruction.

To validate the proposed method, we evaluated through 3-D brain PET simulations the quantitative accuracy of activity corrected for attenuation with using the offset-corrected attenuation image. The attenuation coefficient of the brain tissue is approximated well by that of the water, then transmission-less brain imaging would be the most suitable application of the proposed method. Under wide range conditions of counting statistics, the relative error between the true and estimated activity values was less than 5% at over 90% of the voxels within the brain regions. These results show that, under the above-mentioned assumptions, the scaling problem in the JE approach was solved with practical accuracy.

Keywords: TOF-PET, Attenuation correction, Joint estimation of activity and attenuation, Scaling problem
Poster panel
(face) ID: 231

Poster Number:

A Myth of Iterative Image Reconstruction Algorithms (#1586)

G. L. Zeng1, 2

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


The FBP (filtered backprojection) algorithm reduces image noise by smoothing the image. The iterative algorithm reduces image noise by noise weighting and regularization. A myth is that the iterative algorithm is able to reduce the noise without sacrificing image resolution. This paper uses counter examples to show that this myth is not true. The truth is that the iterative algorithm suppresses image noise by sacrificing image resolution as well; its performance may be better than the windowed FBP. The myth is caused by the comparison method that compares the iterative algorithm with the conventional FBP algorithm instead of the windowed FBP algorithm.

Keywords: Iterative algorithm, Controversy, Noise, Resolution
Poster panel
(face) ID: 234

Poster Number:

A Reconstruction Method based on a Data Analysis Scheme for SPECT Imaging in Parkinson’s Disease (#1804)

L. Koutsantonis1, C. Lemesios1, C. N. Papanicolas1

1 The Cyprus Institute, CASTORC, Nicosia, Cyprus


Dopamine Transporter (DaT) imaging provides a valuable clinical tool for the diagnosis and differentiation of Parkinsonism using 123I agents and Single Photon Emission Computed Tomography (SPECT). In this work, we implement a novel method for tomographic image reconstruction in the field of SPECT - brain imaging. The novel reconstruction method employs the Athens Model Independent Analysis Scheme (AMIAS), a data analysis framework of general applicability in inverse scattering problems. In this framework, the physical characteristics of the imaged distribution are parametrized and formulated as mathematical "objects" representing the uptake of the radioactive substance in the region of interest. Projection data were obtained from simulation studies and used to evaluate the efficacy of the method. The resulting AMIAS tomographic images are compared to those produced by the Maximum Likelihood Expectation Maximization (MLEM) and Filtered Back Projection (FBP) techniques.

Keywords: SPECT, Parkinsons Disease, AMIAS, Image Reconstruction, DATscan, Data Analysis, Tomography
Poster panel
(face) ID: 237

Poster Number:

Preliminary Patient Study of TV-Constrained Image Reconstruction from Low-Statistics List-Mode TOF-PET Data (#2195)

Z. Zhang1, S. Rose1, J. Ye2, A. E. Perkins2, C. - M. Kao1, E. Y. Sidky1, C. - H. Tung2, X. Pan1

1 The University of Chicago, Radiology, Chicago, Illinois, United States of America
2 Philips Healthcare, Cleveland, Ohio, United States of America


Advanced clinical positron-emission tomography (PET) scanners exploit time-of-flight (TOF) information of annihilation photons for improving PET imaging performance, and image reconstruction from TOF-PET data remains an active research topic. In the study, we investigate optimization-based image reconstruction from list-mode TOF-PET data collected with a clinical PET system employing the digital silicon photomultiplier (SiPM) technology. The reconstruction problem is formulated as an image-TV-constrained data-fidelity-minimization problem, and a primal-dual algorithm is tailored for image reconstruction through solving the optimization problem. In an attempt to evaluate the reconstruction performance, we have carried out studies on images reconstructed from data of physical phantoms and patient subjects. A focus of the work centers on investigating image reconstruction from data containing different amounts of photon accounts. Results of the study reveal that the proposed algorithm can yield image reconstruction with suppressed noise level and increased axial coverage over that obtained in conventional reconstructions.

Keywords: Total-Variation, Optimization-based Algorithm, reconstruction, TOF-PET, List-mode data
Poster panel
(face) ID: 240

Poster Number:

From virtual pixel grids to overlapped PSF for PET systems with monolithic crystals (#2518)

L. Moliner1, C. Correcher2, J. Alamo3, J. Alvarez1, V. Gimenez1, A. J. González1, V. Ilisie1, S. Sánchez1, M. J. Rodríguez-Álvarez1

1 CSIC-UPV, I3M, Valencia, Spain
2 Bruker Española SA, I+D, Valencia, Spain
3 Oncovision, I+D, Valencia, Spain


In this work we take advantage of the monolithic continuous crystals in PET photon detection. The called Impact Bin Method (IBM) proposes to use the smaller bin size given by the electronics to position the photon impact and then, build a Point Spread Function (PSF) around this point. We use a rectangular PSF to facilitate the estimation of the voxels emission probabilities. These probabilities are calculated as the volumes of intersection between the tube of response generated by the PSF and the voxels. The electronic bin number is 300x300 for the (40x40) mm monolithic detectors considered. With IMB we avoid the regular virtual grid, so the PSF from different hits can overlap between them. The results show an improvement of 10% in spatial resolution and a 7% in the recovery coefficients without compromise the image uniformity.

Keywords: Detector pixel, monolithic crystal, PET, PSF
Poster panel
(face) ID: 243

Poster Number:

Superiorized polyenergetic reconstruction algorithm for reduction of metal artifacts in CT images (#2638)

T. Humphries1, A. Gibali2

1 University of Washington Bothell, School of STEM, Bothell, Washington, United States of America
2 ORT Braude College, Department of Mathematics, Karmiel, Israel


Artifacts caused by metal objects such as dental fillings, hip implants, and coronary stents are a significant source of error in many CT scans. These artifacts are caused by numerous factors, including beam hardening, noise, photon starvation, partial volume and exponential edge gradient effects, and scatter. We propose an iterative algorithm for CT image reconstruction which reduces these artifacts. The algorithm does so by (1) accurately modeling polyenergetic X-ray data, (2) statistically weighting the X-ray data to reduce the effect of noisy measurements, and (3) incorporating total variation (TV) as a secondary objective. Our numerical experiments indicate that all three of these features of the algorithm play an important role in reducing metal artifacts. The recently proposed superiorization methodology provides a solid mathematical foundation for our approach.

Keywords: computed tomography, superiorization, metal artifacts
Poster panel
(face) ID: 246

Poster Number:

Iterative Image Reconstruction via Sparse Representation for Sparse-view Computed Tomography (CT): Simulation Study (#2825)

G. Kim1, K. Kim1, H. Cho1, C. Park1, H. Lim1, S. Park1, D. Lee1, H. Lee1, J. Park1, S. Kang1

1 Yonsei University, Department of Radiation Convergence Engineering, Wonju, Gangwon-do, Republic of Korea


Conventional computed tomography (CT) based on filtered backprojection (FBP) reconstruction requires a large number of projections to reconstruct images of high quality, which results in large radiation doses and a long scan time. To overcome these difficulties, total-variation (TV) minimization methods based on compressed-sensing (CS) theory have often been adopted for sparse-view CT reconstruction. However, these methods still remain limited due to the edge smoothing through the TV minimization process and worse reconstruction quality with the number of projections typically less than 100. In this work, as an alternative, an iterative reconstruction method via sparse representation in terms of an overcomplete dictionary was investigated for sparse-view CT reconstruction. Here the sparse representation is an effective constraint for solving the optimization problem in an iterative CT reconstruction. The proposed method was compared with the FBP-based algorithm for sparse-view CT reconstruction. We implemented both algorithms and performed a systematic simulation using a three-dimensional (3D) XCAT head phantom. The reconstruction characteristics were evaluated and compared. The proposed algorithm yielded much better reconstruction quality in sparse-view CT reconstruction even with less than 50 projections compared to the FBP-based algorithm, preserving edge sharpening and much less streak artifacts. More details of the simulation results will be described in the paper.

Keywords: Sparse-view, Computed tomography, Sparse representation, Overcomplete dictionary
Poster panel
(face) ID: 249

Poster Number:

Performance Comparison of Parallelized Image Reconstruction in Heterogeneous Cluster Computing Environment (#2880)

S. Bae1, H. Lee1, 2, J. K. Bae3, K. - M. Kim3, K. Lee4

1 Advanced Radiological Sciences Laboratory, Seoul, Republic of Korea
2 Korea University, Research Institute of Global Health Tech, Seoul, Republic of Korea
3 Korea Institute of Radiological and Medical Sciences, Seoul, Republic of Korea
4 Korea University, Department of Bio-convergence Engineering, Seoul, Republic of Korea


Image reconstruction is indispensable for obtaining nuclear medicine diagnostic device images; however, time consumption is an issue in the acquisition of high-quality images. High-performance computing (HPC) and parallel processing techniques have been adopted as an effective solution to this problem.

A heterogeneous cluster computing environment is composed of computing machines with differing performances. Clusters are widely used in science, engineering, and other areas in which HPC is required.

In this study, we developed a method for parallelized image reconstruction in a heterogeneous cluster computing environment, as a part of the single photon emission computed tomography (SPECT) study in the Advanced Reconstruction for Radiation Imaging (ARRA) software. 

The proposed image reconstruction is designed to run in various computing environments. In addition, we developed three parallel processing methods that can be applied according to the computing environment, and efficiently solved the image reconstruction time consumption problem. Hybrid remapper (HR), a dynamic scheduling method, was converted to be suitable for iterative image reconstruction. As a result, the proposed image reconstruction is automatically optimized for the computing environment’s performance.

Keywords: Image reconstruction, parallel processing, variable-pinhole collimator
Poster panel
(face) ID: 252

Poster Number:

Optimization-based Reconstruction with Artifacts Reduction for Neurological Imaging (#3088)

D. Xia1, Y. - B. Chang2, A. H. Siddiqui3, Z. Zhang1, J. Manak2, E. Y. Sidky1, X. Pan1

1 The University of Chicago, Radiology, Chicago, Illinois, United States of America
2 Toshiba Medical Research Institute USA, Vernon Hills, Illinois, United States of America
3 University at Buffalo Neurosurgery, Orchard Park, New York, United States of America


In neurological interventional and surgical procedures, measured data are generally truncated due to the limited detector size and especially in the presence of some interventional devices outside of the imaging FOV. In this work, we investigate optimization-based image reconstruction with an aim to reduce truncation artifacts observed in clinical reconstructions. The reconstruction problem is formulated as a constrained optimization program in which a dataderivative fidelity term was introduced for effective suppression of the truncation artifacts. The investigation is enabled subsequently by tailoring the Chambolle-Pock (CP) algorithm to solve the optimization program containing the data-derivative fidelity term. The results of the study suggest that the optimization-based reconstruction, when applied to data collected from both the physical phantom and the patient subject, yields images with significant reduced artifacts in contrast to the clinical FDK reconstruction.

Keywords: Cone-beam CT, artifact reduction, optimization-based reconstruction, CP algorithm
Poster panel
(face) ID: 255

Poster Number:

SIRF: Synergistic Image Reconstruction Framework (#3216)

E. Ovtchinnikov1, D. Atkinson2, C. Kolbitsch6, B. A. Thomas3, O. Bertolli3, C. O. da Costa-Luis4, N. Efthimiou3, 5, R. Fowler1, E. Pasca1, P. Wadhwa8, 9, E. Emond3, J. Matthews7, A. J. Reader4, C. Tsoumpas8, C. Prieto4, M. Turner1, K. Thielemans3

1 RAL STFC, Visual Analytics and Imaging Systems, Harwell Campus, United Kingdom of Great Britain and Northern Ireland
2 UCL, Centre for Medical Imaging, London, United Kingdom of Great Britain and Northern Ireland
3 UCL, Institute of Nuclear Medicine, London, United Kingdom of Great Britain and Northern Ireland
4 King's College London, Department of Biomedical Engineering, London, United Kingdom of Great Britain and Northern Ireland
5 University of Hull, School of Biological, Biomedical and Environmental Sciences, Hull, United Kingdom of Great Britain and Northern Ireland
6 Physikalisch-Technische Bundesanstalt, Braunschweig and Berlin, Germany
7 University of Manchester, Division of Informatics, Imaging and Data Sciences, Manchester, United Kingdom of Great Britain and Northern Ireland
8 University of Leeds, Division of Biomedical Imaging, Leeds, United Kingdom of Great Britain and Northern Ireland
9 Imanova Ltd, London, United Kingdom of Great Britain and Northern Ireland

On behalf of CCP PETMR


The combination of positron emission tomography (PET) with magnetic resonance (MR) imaging opens the way to more accurate diagnosis and improved patient management. At present, the data acquired by PET and MR scanners are essentially processed separately, and the search for ways to improve accuracy of the tomographic reconstruction via synergy of the two imaging techniques is an active area of research. The aim of the collaborative computational project on PET and MR (CCP-PETMR), supported by the UK engineering and physical sciences research council (EPSRC), is to accelerate research in synergistic PET-MR image reconstruction by providing an open access software platform for efficient implementation and validation of novel reconstruction algorithms. We present the first release of the Synergistic Image Reconstruction Framework (SIRF) software suite from the CCPPETMR. SIRF provides user-friendly Python and MATLAB interfaces to advanced PET and MR reconstruction packages written in C++ (currently this uses the STIR, Software for Tomographic Image Reconstruction, for PET and Gadgetron for MR, but SIRF will be able to link to other reconstruction engines in the future as appropriate) and is capable of reconstructing images from real scanner data. Both of the available integrated clinical PET-MR systems (Siemens and GE) are being targeted, and a suitable data format exchange is being negotiated with the manufacturers.

Keywords: Positron Emission Tomography, Magnetic Resonance Imaging, Research Software Engineering, Scientific Programming
Poster panel
(face) ID: 258

Poster Number:

Reconstruction of Time-Of-Flight Projection Data with the STIR reconstruction framework (#3672)

N. Efthimiou1, E. C. Emond2, C. Cawthorne1, C. Tsoumpas3, K. Thielemans2

1 University of Hull, School of Life Sciences, Faculty of Health Sciences, Hull, United Kingdom of Great Britain and Northern Ireland
2 University College London, Institute of Nuclear Medicine, London, United Kingdom of Great Britain and Northern Ireland
3 University of Leeds, Division of Biomedical Imaging, Leeds, United Kingdom of Great Britain and Northern Ireland


This manuscript provides an update on the integration of Time-Of-Flight (TOF) reconstruction in the STIR image reconstruction toolkit. In this update, we provide support for reconstruction of TOF projection data. This needed modifications to many data classes, projectors and list-mode handling. The software is written to be able to handle different timing resolutions and TOF bins. In this work, we provide initial validation results. Using simple phantoms analytically projected, we calculate the contrast recovery ratio over a wide range of iterations. The results show the benefits of TOF reconstruction and are in good agreement with literature. The Contrast Recovery Ratio (CRC) at early iterations improves with better timing resolutions and smaller width of the TOF bins.

Keywords: Time-of-Flight, STIR, PET image reconstruction
Poster panel
(face) ID: 261

Poster Number:

Image Reconstruction Method for Multiple Isotope PET (#3851)

T. Fukuchi1, T. Okauchi1, M. Shigeta1, H. Haba2, Y. Watanabe3, S. Enomoto1

1 RIKEN, Center for Life Science Technologies, Kobe, Hyogo, Japan
2 RIKEN, Nishina Center for Accelerator-Based Science, Wako, Saitama, Japan
3 Nagoya University, Graduate School of Medicine, Nagoya, Aichi, Japan


PET is a useful tool for noninvasive imaging of the tracer distribution in living organisms. However, conventional PET is restricted to use for single tracer, because it is based on the detection of positron annihilation photons, which always have an energy of 511 keV regardless of the variety of the radionuclide. If PET could image multiple tracers, it would be a much more useful tool. Then, in order to realize the multiple tracer imaging by PET, we developed a multi-isotope PET (MI-PET) system based on a small-animal PET. Developed MI-PET system is composed of a PET scanner, which consists of pixelized GSO detectors and 8 additional BGO detectors. This MI-PET system can distinguish between different tracers using the additional γ-ray detectors to observe prompt γ-ray, which is successively emitted after the positron emission. However, because of limited detection efficiency for the prompt γ-ray, a superposed image of the first and second tracers and an image of only the second tracer were reconstructed. Therefore, we developed a specific image reconstruction method to obtain an isolated image of the first tracer. The image reconstruction method developed is based on event subtraction at the production of projection matrices for double- and triple-coincidence with the sensitivity correction. To compensate insufficient number of triple-coincidence events in 22Na blank scan, a smoothing method is applied to produce the global feature of sensitivity for the prompt γ-ray. In addition to this global feature, a local feature produced by double-coincidence events is inserted as sensitivity of PET detectors. Using sensitivity parameters thus produced, subtraction of a projection matrix of triple-coincidence events from double-coincidence events was executed to produce the projection matrix for isolated image of the first tracer. Finally, we succeeded in reconstructing isolated image of the first tracer from dual-tracer (18F-FDG and 22Na) experiment for a mouse.

Keywords: PET, Multi-Isotope, Image reconstruction
Poster panel
(face) ID: 264

Poster Number:

Dynamic cerebral perfusion computed tomography reconstruction using joint tensor and sparsity constraints (#3920)

D. Zeng1, 2, J. Lin1, 2, J. He1, 2, S. Li1, 2, Y. Liao1, 2, Y. Wang1, 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


Dynamic cerebral perfusion computed tomography (DCPCT) has the ability to evaluate the hemodynamic information throughout the brain.  However, due to multiple three-dimensional acquisitions protocol, DCPCT scanning imposes high radiation doses on the patients and the associated radiation dose has received growing concerns.  To address this issue, in this study, we propose a new robust DCPCT images reconstruction algorithm to improve low-dose DCPCT images and perfusion maps quality through tensor and sparse decomposition model.  For simplicity, the proposed model is termed as tensor-based robust principle component analysis (T-RPCA).  Specifically, the proposed model views the DCPCT sequential images as a mixture of tensor matrix and sparse matrix to describe the maximum temporal coherence of spatial structure among phases in a tensor framework.  Moreover, the tensor matrix corresponds to the “background” with spatial-temporal correlations, e.g. static anatomical contribution, which is stationary over time or similar in structure and the sparse matrix represents the time-varying component with spatial-temporal continuity, e.g. dynamic perfusion information, which is approximately sparse over time.  Efficient algorithms using alternating direction method of multipliers (ADMM) are developed to solve the proposed model.  Extensive experiments with a digital brain perfusion phantom, and preclinical monkey data clearly demonstrate that the presented T-RPCA model can achieve more gains than the existing popular algorithms in terms of both quantitative and visual quality evaluations.

Keywords: Dynamic cerebral perfusion, CT reconstruction, joint tensor and sparsity constraints, robust principle component analysis
Poster panel
(face) ID: 267

Poster Number:

Automatic Vertebrae Segmentation in Fluoroscopic Images for Electrophysiology (#1041)

S. Reiml1, T. Kurzendorfer1, D. Toth2, P. Mountney2, S. Steidl1, A. Brost2, A. Maier1

1 Friedrich-Alexander University Erlangen-Nürnberg, Erlangen, Germany
2 Siemens Healthineers, Forchheim, Germany


Worldwide, heart failure affected about 26 million people in 2014. The treatment is most often done minimally invasive. As fluoroscopic images provide little functional or anatomical information of the heart, most often pre-operative magnetic resonance imaging data is used, to provide additional information. The 3-D data needs to be fused with the fluoroscopic images. Therefore, an accurate 2-D/3-D registration is necessary. This can be achieved by registering the bones in fluoroscopic images with the bones in magnetic resonance imaging data. However, a special bone magnetic resonance scan is needed. In this paper, we propose a learning based approach in combination with an active contour for the segmentation of vertebrae in fluoroscopic images. After preprocessing the fluoroscopic images, a random forest classifier is trained to extract the vertebrae. The resulting probability map is used to initialize an active contour approach for the segmentation of the vertebrae. The proposed method was evaluated on 12 fluoroscopic data sets and compared to a gold standard annotation obtained from a clinical expert. This comparison yielded a Dice coefficient of 80% and a vertebrae detection rate of 86 %.

Keywords: segmentation, bones, vertebrae, fluoroscopic images, feature extraction, classification, machine learning, active contour
Poster panel
(face) ID: 270

Poster Number:

Robust Seed Mask Generation for Interactive Image Segmentation (#1197)

M. P. Amrehn1, 2, S. Steidl1, M. Kowarschik2, A. K. Maier1

1 University of Erlangen-Nuremberg, Department of computer science, Erlangen, Bavaria, Germany
2 Siemens Healthcare GmbH, Advanced Therapies Innovation, Erlangen, Bavaria, Germany


In interactive medical image segmentation, anatomical structures are extracted from reconstructed volumetric images. The first iterations of user interaction traditionally consist of drawing pictorial hints as an initial estimate of the object to extract. Only after this time consuming first phase, the efficient selective refinement of current segmentation results begins. Erroneously labeled seeds, especially near the border of the object, are challenging to detect and replace for a human and may substantially impact the overall segmentation quality.

We propose an automatic seeding pipeline as well as a configuration based on saliency recognition, in order to skip the time-consuming initial interaction phase during segmentation. A median Dice score of 68.22% is reached before the first user interaction on the test data set with an error rate in seeding of only 0.088 %.

Keywords: Interactive Image Segmentation, Seeding, HCI, Usability, Interaction, Segmentation, Medical Imaging
Poster panel
(face) ID: 273

Poster Number:

A fractional active contour model for medical image segmentation (#1795)

B. Chen1, 2, S. Huang1, J. Liang2, W. Chen1, 3, H. Lin1, B. Pan1, 3

1 Shenzhen University, College of Mathematics and Statistics, Shenzhen, Guangdong, China
2 State University of New York, Department of Radiology, Stony Brook, New York, United States of America
3 Shenzhen University, Shenzhen Key Laboratory of Media Security, Shenzhen, Guangdong, China


Considering both traditional local and global information for medical image segmentation remains a challenging task. Some hybrid methods have shown promise in relieving the challenge. In this paper, a new hybrid method is presented, which incorporates image gradient, local information and global information into a framework. The energy or level-set function in the framework integrates fractional order differentiation, fractional order gradient magnitude, and difference image information into the well-known local Chan-Vese model, which has been shown to be effective and efficient in modeling the local information. The presented new model can also enhance low frequency information, which is clinically desired. Experiments on synthetic images as well as real images were performed to demonstrate the segmentation accuracy and computational efficiency of the presented hybrid method. The dice similarity coefficient merit was employed as the comparative quantitative measures and showed a noticeable gain over a current hybrid method.

Keywords: image segmentation, active contour model, fractional order differentiation, level set
Poster panel
(face) ID: 276

Poster Number:

Dental root canal segmentation from super-resolved 3D cone beam computed tomography data (#1873)

R. Sfeir1, 2, J. Michetti1, 3, B. Chebaro2, F. Diemer3, 4, A. Basarab1, D. Kouamé1

1 University of Toulouse, Paul Sabatier, IRIT, CNRS, UMR 5505, Toulouse, France
2 Lebanese University, LaRIFA, Hadath, Lebanon
3 University of Toulouse, Paul Sabatier, Faculté de Chirurgie Dentaire, Toulouse, France
4 University of Toulouse, Clément Ader Institute, Toulouse, France


The treatment success of dental pulp cavity mainly depends on a good knowledge of the root canal anatomy. Micro computed tomography (μCT) is the standard technique for imaging the root canal anatomy, providing high spatial resolution and high contrast 3D images. μCT has however several issues limiting its clinical use: limited field of view only allowing extracted tooth imaging, high irradiation doses, limited access and long acquisition time (a few hours per tooth). Cone Beam Computed Tomography (CBCT) is an interesting alternative to μCT, able to image the dento-maxillo-facial structures using a conic X-ray beam moving around the patient head. In contrast to μCT, CBCT has a larger field of view, a reduced acquisition time and lower irradiation doses, making it suitable for clinical applications. Despite its numerous advantages, CBCT suffers from its insufficient spatial resolution, that prevents it, for the moment, from exploring the root canal system quantitatively. The objective of this paper is to evaluate the potential of super-resolution (SR) image processing to enhance the resolution of CBCT images and to further improve the root canal segmentation. We model this loss of spatial resolution by an image-degradation process consisting of both blurring and a down-sampling operators. Given the piecewise constant nature of CBCT slices, the reconstruction method uses total variation regularization to overcome the ill-posedness of SR. Finally, an automated segmentation procedure has been applied to μCT, native CBCT and super-resolved CBCT volumes in order to extract the root canal structure. Seven intact extracted teeth have been used to evaluate the potential of SR CBCT in detecting the root canal anatomy. Given the high-resolution and contrast of μCT data, the resulting segmented root canals were considered as the ground truth. For all the considered teeth, the SR CBCT volumes provided a smaller error compared to the native CBCT data.

Keywords: super-resolution, inverse problems, cone beam computed tomography, micro computed tomography, endodontics, image segmentation
Poster panel
(face) ID: 279

Poster Number:

Mismatch Correction for Free-breathing PET and Deep-inspiration Breath-holding CT in PET/CT Imaging (#2070)

H. Shi1, T. Feng2, Y. Dong3, H. Li2, J. Wang2

1 Zhongshan Hospital, Fudan University, Department of Nuclear medicine, Shanghai, China
2 UIH America, Inc., Houston, Texas, United States of America
3 Shanghai United Imaging Healthcare Co., Ltd, Shanghai, China


CT images acquired during deep-inspiration breath-holding (DIBH) have higher image quality than free-breathing (FB) CT images as they are free from respiratory motion artifacts. However, there exists significant mismatch between DIBH CT images and FB PET images since the displacement of organs in the thorax and abdomen at DIBH is significantly larger than that during regular FB. Our goal is to correct for the large mismatch between FB PET and DIBH CT in PET/CT imaging.

The list-mode PET data were divided into four equal-count frames using amplitude based data-driven respiratory gating technique. Frame 1 and Frame 4 corresponded to end-expiration and end-inspiration respectively, and they were reconstructed using OSEM image reconstruction algorithm. To avoid attenuation-activity mismatch artifacts, the attenuation map used in attenuation correction was modified by filling the lung region with attenuation coefficient of soft tissue. 3D motion vector field from Frame 4 to Frame 1 was calculated from the reconstructed PET images using B-spline image registration algorithm. Assuming the respiratory motion during FB is consistent during FB and DIBH, the motion vector field from DIBH to end-inspiration phase is modeled by scaling the calculated respiratory motion from Frame 4 to Frame 1. The scalar was determined by maximizing the mutual information between transformed DIBH attenuation map and PET image at Frame 4. DIBH attenuation map was transformed using the scaled motion vector field from Frame 4 to Frame 1 to obtain attenuation map at Frame 4. Attenuation maps for other frames were obtained by transforming attenuation map at Frame 4 using corresponding motion vector fields.

The method was applied to clinical patient data acquired using DIBH CT and FB PET protocol. While DIBH attenuation map creates significant mismatch artifacts in PET images, the proposed method was able to generate attenuation maps matching FB PET frames and reduce mismatch artifacts.

Keywords: activity-attenuation mismatch, PET/CT, breath-holding, free-breathing
Poster panel
(face) ID: 282

Poster Number:

Automated Fovea Detection Based on Unsupervised Retinal Vessel Segmentation Method (#2184)

M. Tavakoli1, P. Kelley1, M. Nazar2, F. Kalantari3

1 Indiana University-Purdue University, Physics, Indianapolis, Indiana, United States of America
2 Shahid Beheshti Medical Sciences, Biomedical Sciences, Tehran, Iran (Islamic Republic of)
3 University of Texas Southwestern Medical Center, Radiation Oncology, Dallas, Texas, United States of America


The Computer Assisted Diagnosis systems could save workloads and give objective diagnostic to ophthalmologists. At the first level of automated screening of systems, feature extraction is a fundamental step with the fovea localization as one of these features of interest. The fovea is a small fossa on the fundus, which is represented by a deep-red or red-brown color in color retinal images. By observing retinal images, it appears that the main vessels diverge from the optic nerve head (ONH) and follow a specific course that can be geometrically modeled as a parabola, with a common vertex inside the ONH and the fovea located along the apex of this parabola curve. Therefore, based on this assumption, the main retinal blood vessels are segmented and fitted to a parabolic model. With respect to the core vascular structure, we can thus detect fovea in the fundus images. For the vessel segmentation, our algorithm addresses the image locally where homogeneity of features is more likely to occur. The algorithm is composed of 4 steps: image partitioning (windows), local Radon transform, vessel validation, and parabolic fitting. In order to extract blood vessels, sub-vessels should be extracted in local windows. The high contrast between blood vessels and image background in the images cause the vessels to be associated with peaks in the Radon space. The largest vessels, using a high threshold of the Radon transform, determines the main course or overall configuration of the blood vessels which when fitted to a parabola, leads to the future localization of the fovea. In effect, with an accurate fit, the fovea normally lies along the slope joining the vertex and the focus. The darkest region along this line is the indicative of the fovea. 

The results show that, among 20 images of the first public database (DRIVE) we detected fovea in 85% of them. For the second public database (MESSIDOR) among 200 images we detect fovea correctly in 88% on them.

Keywords: retinal vessel, Fovea, Radon transform, parabola, top-hat
Poster panel
(face) ID: 285

Poster Number:

Connectional fingerprint of mild cognitive impairment based on FDG-PET and PiB-PET (#2751)

S. J. Son2, 3, M. Kim2, 3, S. H. Lee2, 3, H. Park1, 3

1 Sungkyunkwan, School of Electronic and Electrical Engineering, SUWON-SI, GYEONGGI-DO, Republic of Korea
2 Sungkyunkwan, Department of Electronic, Electrical and Computer Engineering, SUWON-SI, GYEONGGI-DO, Republic of Korea
3 Institute for Basic Science, Center for Neuroscience Imaging-Research (CNIR), SUWON-SI, GYEONGGI-DO, Republic of Korea


Alzheimer’s disease (AD) is one of the important aging-associated diseases. Physical and cognitive functions are gradually lost, ultimately leading to death. Most AD patients transition from a mild cognitive impairment (MCI) state before the final diagnosis of AD. Detecting changes from normal state to MCI state is crucial to apply optimal treatment options depending on the progression of AD. Connectional fingerprint can provide important information to characterize connectivity patterns related to MCI and AD. Here, we computed connectional fingerprint of normal control (NC) and MCI patients using fludeoxyglucose (FDG)- and Pittsburgh compound B (PiB)-positron emission tomography (PET) to identify unique pattern of information to describe the MCI group. A connectivity analysis was implemented using whole brain networks. Group-wise differences between NC and MCI patients were assessed using betweenness centrality (BC). Left Lingual and right putamen showed significant decreases in BC values from NC to MCI in FDG-PET. Fusiform also showed a significant decrease in PiB-PET. Finally, we defined the unique patterns of connectional fingerprint using ten brain regions which showed significant differences in BC between NC and MCI. The regions were superior frontal L, anterior cingulum L, caudate L, lingual L, middle occipital R, inferior occipital R, putamen R, hippocampus R, fusiform R, and paracentral lobule R. Our study reported significant brain regions that showed functional degeneration using BC for NC and MCI groups. A connectional fingerprint in terms of radar plot was given for NC and MCI groups to effectively visualize the fingerprint information.

Keywords: Connectional fingerprint, Connectivity, FDG-PET, PiB-PET, Alzheimer's disease
Poster panel
(face) ID: 288

Poster Number:

Deep Learning-based Inpainting for Virtual DSA (#2890)

M. Unberath1, J. Hajek1, T. Geimer1, F. Schebesch1, M. Amrehn1, A. Maier1

1 Friedrich-Alexander University Erlangen-Nürnberg, Pattern Recognition Lab, Erlangen, Bavaria, Germany


Digital subtraction angiography (DSA) requires subsequent acquisition of fluoroscopy images with and without contrast injection. These are subtracted to yield images of the contrast agent only that are, ideally, not truncated. Relying on consecutively acquired angiograms, however, makes the method susceptible to inter scan motion and, hence, inapplicable to coronary angiography.
For cardiac imaging, virtual single-frame DSA pipelines received considerable attention. These methods estimate non-contrast background images from contrast enhanced projections via vessel segmentation and image inpainting.
In this work, we devise a convolutional neural network (CNN) in the well-known U-net architecture to regress image intensity values in coronary angiograms. The network is trained on simulated DSA acquisitions only, but is applied to both numerical phantom as well as clinical data.
We found promising results on both numerical phantom and real data when comparing our learning-based method to competing analytic approaches.
The performance encourages further research on deep learning-based virtual DSA that should investigate possibilities for omitting vessel segmentation.

Keywords: Regression, DSA, Material decomposition
Poster panel
(face) ID: 291

Poster Number:

Noise reduction and crosstalk correction in joint activity and attenuation reconstruction using deep learning (#2995)

D. Hwang1, K. Y. Kim1, S. K. Kang1, J. S. Lee1

1 Seoul National University, Biomedical Sciences, Seoul, Republic of Korea


Although joint estimation of activity and attenuation using Maximum likelihood reconstruction of activity and attenuation (MLAA) is a promising method for PET attenuation correction, it suffers several problems including the crosstalk artifacts, slow convergence speed, and noisy attenuation maps. In this work, we developed a convolutional neural network (CNN) to overcome these limitations of MLAA. We have applied this new approach to the most challenging PET cases for the joint estimation ([18F] FP-CIT PET scans with highly specific binding to striatum of the brain). Eight patient scans were acquired using a Biograph mCT 40 scanner (effective timing resolution = 580 ps, Siemens Healthcare). Sinograms of prompt counts and correction factors were generated using e7 toolkit. We reconstructed all data sets using the OSEM and MLAA with the TOF information (8 iterations and 21 subsets). Five patient data were used as traing set, and the others as test set. For training and testing, activity and attenuation maps derived from MLAA were used as input X, and μ-map converted from CT as label data Y. Data were augmented by rotating and reslicing. The cost function was the L2-norm between MLAA attenuation map and CT-derived -map. The cost function is minimized with adaptive moment estimation (Adam). To improve the reliability of the network, we employed the dropout with probability 0.7, batch size 10, and epoch 20, respectively. The CNN generates much less noisy and more uniform images compared to original MLAA attenuation maps. Also, the air cavities and bones are better resolved in the proposed CNN output. The CNN was also useful for mitigating the crosstalk problem in MLAA reconstruction. We can expect the better quality of reconstructed images when using the proposed method instead of MLAA attenuation map.

Keywords: Deep learning, Joint estimation, Crosstalk, Maximum likelihood reconstruction of activity and attenuation (MLAA)
Poster panel
(face) ID: 294

Poster Number:

Histology Image Classification using Multistage Segmentaion with Shannon Entropy Features (#3689)

K. Masood1, J. Iqbal2

1 University of Jeddah, Jeddah, Saudi Arabia
2 University of Windsor, Engineering, Windsor, Canada


We propose multi stage segmentation of glands in colon histology images for classification into benign or malignant samples. In literature, it is reported that most of the time of pathologists is spent on examining benign cases, hence if we develop a model which automatically screens normal samples from malignant samples, then pathologists may spend their valuable time in determining the grade and physical state of complex cancerous regions. Classification of the colon tissue samples is a multi stage process consisting of three fundamental algorithms. For coarse segmentation, meanshift algorithm is used. The algorithm is based on the Parzen-window estimation and well suited for dealing with artifacts and noise in the images as in our case there is some imaging noise incorporated in the data. The other advantage of the algorithm is that it is an iterative procedure and also it is unsupervised which make it very fast in comparison to other algorithms used for segmentation. Thus by using mean-shift algorithm, we reduce the computational complexity of our multi stage algorithm. The fine segmentation of the samples is performed using wavelet packet features with AdaBoost models as classifiers. AdaBoost algorithm provides an insight into the performance of different features as oppose to other ensemble methods like random forest that sample the feature space or SVM which projects the data into higher dimensional space. AdaBoost method selects the discriminative features from rest of the lot, hence boosting the classification performance. In feature extraction stage, Shannon entropy is used to differentiate between benign and malignant classes.It is verified in our results and as proposed that entropy values are highre in malignant samples and using a threshold with tuning parameters, almost 94% classification accuarcy is achieved.

Keywords: Mean-shif segmentationt, Wavelet Packets, Shannon Entropy, Colon Cancer
Poster panel
(face) ID: 297

Poster Number:

Advancements in data-driven respiratory motion extraction methods for clinical list-mode 18F-FDG PET datasets acquired from a commercial PET scanner (#4009)

T. - S. Lee1, J. Wang1, J. Xu1, P. Olivier2, A. E. Perkins2, C. - H. Tung2, B. M. W. Tsui1

1 Johns Hopkins University, Radiology, Baltimore, Maryland, United States of America
2 Philips HealthTech, Highland Heights, Ohio, United States of America


We have significantly improved four data-driven respiratory motion (RM) extraction methods for various activity distributions in clinical myocardial perfusion (MP) and 18F-FDG PET datasets. They are activity distributions: (1) with high myocardial uptake, (2) same as (1) but with portion of the heart outside the image, and with high image intensity (3) in the liver and (4) in the lung area without attenuation compensation. In Method #1, a 3D volume-of-interest (VOI) was placed over the heart region of the PET image obtained from the total acquisition time period. The surrogate RM signals were obtained from the centroids of the image intensity of the myocardial activity uptake within the same VOI of PET images obtained from rebinned list-mode data in short time intervals. The Fourier Transform (FT) of the time sequence of surrogate RM signals and smoothing reveal the RM peak and its average period, Pav. In Methods #2, #3, and #4, specially-shaped 3D VOIs were placed over the heart, the top of the liver, and the bottom of the lungs, respectively. Then, the same procedures used in Method #1 were employed except using the total counts within the corresponding VOI. The location, sizes and shapes of the VOIs were optimized for the highest signal-to-noise (S/N) in the RM peak extraction. The improved RM extraction methods were evaluated using 14 patient datasets. Method #1 was shown to work well for 79% of the datasets, and Pav showing high S/N and excellent agreement (Pearson correlation coefficient 0.997) with those obtained from an external RM monitoring belt system. Method #2 was applied successfully to 14%, and Methods #3 and #4 to the rest of datasets. Excellent agreements were also found in cross comparison between the methods.We conclude that the improved data-driven RM extraction methods which showed successful results in various PET image datasets will provide an important first step for the motion compensation application in commercial PET scanners.

Keywords: respiratory motion extraction, myocardial perfusion, PET
Poster panel
(face) ID: 300

Poster Number:

Maximum-Likelihood Event Parameter Estimation from Digital Waveform Capture (#4229)

N. C. Momsen1, G. Richards1, M. Kupinski1, H. H. Barrett1, L. R. Furenlid1

1 University of Arizona, Department of Medical Imaging, Tucson, United States of America


A preclinical Single Photon Emission Computed Tomography (SPECT) system design is presented, that will be used in rabbit myocardial perfusion and related cardiac studies. The system includes digital-waveform capture of all PMT signals, which allows for optimal maximum-likelihood estimation in the gamma-ray event parameter estimation and tomographic reconstruction processes. In a typical gamma ray camera, only the integrated signal of a gamma ray event is recorded. Here, because the entire waveform is recorded, it is possible to incorporate information from the waveform shape into the maximum-likelihood estimation. The optimal usage of these parameters in estimating event parameters (x, y, z, energy, and time) is explored. The detector, a full-size clinical SPECT camera with 61 PMTs, was retrofitted with an array of active buffers that tap into raw low-level signals before they reach the Anger-logic network. Methods for system calibration and integration are discussed, along with predictions and measurements of system performance.

Keywords: Maximum-Likelihood Estimation, SPECT, Waveform-Capture
Poster panel
(face) ID: 303

Poster Number:

Patient-Specific Hybrid Whole-body PET Parametric Imaging From Speed Modulated Continuous Bed Motion Dynamic Data (#1774)

F. A. Kotasidis1, V. Garibotto1, H. Zaidi1

1 Geneva University Hospital, Division of Nuclear Medicine & Molecular Imaging, Geneva, Genève, Switzerland


In our previous work a  hybrid whole-body dynamic protocol was proposed, enabling full compartmental modelling, as well as whole-body (WB) Patlak analysis, in continuous bed motion (CBM) acquisition mode. While such an approach improves upon previous WB dynamic protocols, kinetic parameters are estimated with reduced precision and accuracy due to limited counting statistics. In this work we propose a new patient/pathology specific whole-body dynamic imaging protocol with the aim to maximize the counting statistics in the regions of interest and achieve optimum kinetic parameter accuracy and precision based on a personalized dynamic data acquisition. This is achieved by modulating the CBM bed speed within each pass depending on the FOV of primary interest as opposed to utilizing a constant speed for the entire FOV. Therefore in cases where the pathology is known and localized to a certain degree, such as in response to therapy or based on information from other modalities or clinical findings, the protocol can be modulated in such a way as to maximize the counting statistics and kinetic parameter accuracy and precision in a disease and patient specific way. Using dynamic WB simulations as well as initial clinical dynamic WB scans, we demonstrate the kinetic parameter estimation benefits and the clinical feasibility of the proposed protocol.

Keywords: wholebody PET, kinetic modelling, parametric imaging, dynamic imaging, continuous bed motion
Poster panel
(face) ID: 306

Poster Number:

A Model to Explain 11C-PBR28 SUV Profile in Parkinson’s Disease and Unaffected Lrrk2 Mutation Carriers (#2272)

R. Mabrouk1

1 University of British Columbia, Brain Research Center, Vancouver, British Columbia, Canada


In vivo Positron Emission Tomography (TEP) imaging is widely used to study multiple aspects of brain functions in Parkinson’s disease (PD) including neuroinflammation link to the disease. Imaging studies with 11C-PBR28 are routinely employed to explore microglial cells activation via Tanslocator Protein (TSPO). In this paper, we used Principal component analysis technique to explore 11C-PBR28-SUV profile. Participants consisted of healthy control (HC), sporadic PD (sPD), PD patients with a Leucine-rich repeat kinase 2 (Lrrk2) genetic mutation (Lrrk2-PD) and unaffected Lrrk2 mutation carriers (UC). The genotype effect was found as a principal feature of group-depending 11C-PBR28 binding. This effect is observable in both absolute- and Subject residual Profile-SUV analyses. The PC1 of the absolute SUV analysis shows scores highly significant elevated in HAB-HC compared to MAB-HC (p = 0.001) and in HAB-sPD compared to MAB-sPD (p = 0.006). Scores were moderately significant elevated in HAB-UC compared to MAB-UC (p = 0.02). PC1 scores, however, were found obscuring the difference between HAB and MAB subjects in Lrrk2-PD (p = 0.9). In comparison, PC2 scores resulting from SRP-SUV analysis show a significant difference between HAB and MAB subjects in each group (in HC, p = 0.002; in sPD, p = 0.003; in Lrrk2-PD, p = 0.04 and in UC p = 0.004). Specifically, the brainstem regions (Pons, Midbrain and Medulla) had a higher magnitude of SUV compared to the cortical regions (Temporal, Frontal, Parietal and Occipital lobes).

Keywords: PET, PD, Lrrk2, PCA
Poster panel
(face) ID: 309

Poster Number:

Components Determination in Hypoxic Glioblastoma Measured with 18F-FMISO PET imaging (#3273)

R. - A. Abdo1, F. Lamare2, M. Allard2, P. Fernandez2, M. Bentourkia1

1 Université de Sherbrooke, Nuclear Medicine and Radiobiology, Sherbrooke, Québec, Canada
2 Université de Bordeaux2- EPHE, Service de Médecine Nucléaire, Bordeaux, Bordeaux cédex, France


Glioblastoma (GBM) tumors are characterized by high levels of hypoxia, which results in poor prognosis even after treatment by radiotherapy and chemotherapy. The most current technique used for imaging hypoxic GBM is 18F-Fluoromisonidazole (18F-FMISO) with PET imaging. The aim of this study is to decompose the dynamic 18F-FMISO in single tissue components and to study the spatial distribution of GBM in the tumor over time in comparison to the corresponding anatomic Magnetic resonance imaging (MRI) images. GBM patients (n=13) were imaged with-FMISO PET/CT. In addition, an MRI T1-weighted images were also taken for co-registration with PET images. Scanning protocol begins with 15 minutes dynamic images of the brain. Also, static images were taken after 2, 3 and 4 hours. We used the spectral analysis technique to decompose the whole 3D dynamic image into its components in order to isolate the hypoxic regions and determine the number of tissue components. The results show non-uniformity in tumor shape with time frames. The tumor becomes more uniform after 2 h, even though, a difference in shape has been found with later frames. Some tumor regions show accumulative response while others, even appearing with comparable uptake of 18F-FMISO, they were not observed as hypoxic based on the results of spectral analysis. Clearly, blood volume, hypoxic and perfused tissues and even necrotic regions can be isolated and studied separately. Therefore, the pattern of changes in tumor shape over time can be explained for a better tumor treatment.

Keywords: Glioblastoma; Hypoxia; 18F-FMISO; Cancer; Spectral analysis; PET imaging
Poster panel
(face) ID: 312

Poster Number:

Clustering Analysis for Neurotransmitter Response Profiles of Dynamic PET data (#3394)

R. Misiunaite1, 2, G. I. Angelis1, S. R. Meikle1

1 The University of Sydney, Brain and Mind Centre, Sydney, NSW, Australia
2 Technical University of Denmark, Biomedical Engineering, Kgs. Lyngby, Denmark


In this paper, we have investigated clustering as a potential aid in reducing false positive rates when modelling neurotransmitter activation using the lp-ntPET model. The aim was to investigate whether clustering time- activity curves (TACs) before kinetic modelling improves specificity for detecting activation states. We focused mainly on the k-means clustering algorithm, which was applied prior to kinetic modelling. We generated random sets of TACs corresponding to [11C]raclopride kinetics and we investigated the impact on the clustering approach of (a) the level of noise in the TACs, (b) the activation magnitude and (c) the ratio between the number of TACs with activation and without. For statistical measures of clustering performance we calculated the sensitivity and specificity for each case and compared results with those obtained from single voxel modelling without k-means clustering. Results show that applying k-means clustering before the kinetic modelling at noise levels from 1k counts to 1M counts assures better sensitivity. The single voxel modelling results in better specificity at higher noise levels. Additionally, the number of TACs has an effect on sensitivity and specificity together with the activation magnitude.

Keywords: clustering, neurotransmitter, dynamic PET data, time activity curves, k-means, activation, single voxel modeling
Poster panel
(face) ID: 315

Poster Number:

Development of an Iterative Reconstruction Method for Low Dose CBCT in Proton Therapy Patient Positioning (#1357)

T. Yamaguchi1

1 Sumitomo Heavy Industries, Ltd., Technology Research Center, Yokosuka-shi, Kanagawa, Japan


Cone Beam Computed Tomography (CBCT) is used to determine patient position before each irradiation, which results in increased radiation exposure for the patient. To reduce the patient exposure, reduction of either X-ray imaging time or X-ray tube current is necessary. However, this leads to increased noise and reduced contrast in images. To meet this requirement, we have developed an image reconstruction method employing an iterative algorithm that is robust to noise and that can achieve high image contrast. In the iterative image reconstruction process, a system matrix corresponding to the measured geometry is required. In gantry-mounted CBCT, the X-ray tube and the flat panel detector (FPD) deviate some from their ideal position, resulting in a different geometry for each measurement angle. However, full calculation of system matrices requires a large data capacity or significant computation time. To overcome these issues, our method calculates the system matrix for an ideal geometry in advance and performs positional deviation correction on the projection data, allowing us to obtain a reconstructed image with positional deviations taken into account. We applied both the Feldkamp method and the proposed method to digital phantom data with positional deviations and Poisson noise, and by comparing the reconstructed images, we were able to confirm improved noise reduction and image contrast. In addition, by calculating the ratio of the ROI value in the high absorption region with different size and that of BG in the module for evaluating resolution on each iterative image, the number of updates required to restorate the fine distribution was evaluated. As a result, it was found that about 200 updates is required to recover the distribution with 3lp/cm.

Keywords: Iterative reconstruction, Cone Beam Computed Tomography, CBCT, Low Dose CT, Gantry-mounted CBCT
Poster panel
(face) ID: 318

Poster Number:

Prompt gamma-ray imaging with a nuclear emulsion for in vivo dose verification in proton therapy (#1581)

T. Toshito1, 2, M. Kimura1, H. Ogino1, Y. Shibamoto2, M. Nakamura3, O. Sato3

1 Nagoya Proton Therapy Center, Department of Proton Therapy Physics, Nagoya, Japan
2 Nagoya City University, Graduate school of Medical Sciences, Nagoya, Japan
3 Nagoya University, Graduate School of Science, Nagoya, Japan


We propose a new concept of single photon emission computed tomography (SPECT) system using nuclear emulsion technique to detect prompt gamma-rays for in vivo three-dimensional dose verification in proton therapy. It is capable of being dead-time-free at clinical dose rates, neutron-background-free and of high three-dimensional resolution. Prompt gamma-rays above 1 MeV are converted to e+e- pairs, and electrons and positrons are three-dimensionally tracked in nuclear emulsion. Tracks are measured by using a high-speed emulsion read-out system. Direction and energy of primary gamma-rays can be reconstructed from initial directions of secondary tracks, ranges and angular deflections caused by multiple Coulomb scattering. A first prototype emulsion gamma camera was tested at Nagoya Proton Therapy Center. The gamma camera which measures 62.5 x 50 x 3.2 mm3 (0.28 radiation length) was put on the sidewall of the phantom. A water phantom was irradiated with 200 MeV proton beams using a spot scanning nozzle at clinical dose rates. Emulsion films were developed and read out at Nagoya University. The off-line analysis was conducted for data reduction and track reconstruction. The total number of read-out tracks per emulsion film was 7 x 107. Tracks in each film were connected between both sides of the base of each film (base track) in off-line data reduction process. The number of base tracks per emulsion film was 5 x 106. Connection of base tracks between films was performed to reconstruct secondary tracks. As a result, gamma-ray events which have e+/e- pair topology were localized. We examined a first prototype nuclear emulsion gamma-ray camera system for proof-of-principle. The detector system operated within expectations. We detected several-MeV gamma-ray events. Our developed emulsion gamma camera confirmed its potential as a detector for emulsion SPECT system for in vivo dose verification in proton beam radiotherapy.

Keywords: nuclear emulsion, gamma camera, particle therapy, proton therapy, SPECT
Poster panel
(face) ID: 321

Poster Number:

Hadrontherapy with radioactive ion beams: Performance evaluation using FLUKA (#2394)

R. S. Augusto1, 2, A. Mohammadi3, H. Tashima3, E. Yoshida3, A. Ferrari1, K. Parodi2, T. Yamaya3

1 CERN, EN-STI, Geneva, Switzerland
2 Ludwig-Maximilians-Universität, Experimental Medical Physics, Munich, Germany
3 National Institute for Radiological Sciences, Imaging Physics Team, Chiba, Japan


In this work the Monte Carlo particle transport code FLUKA was used to assess how Hadrontherapy treatments can benefit from the use of radioactive ions beams (RIB), estimating the gain in imaging potential with comparable dose delivery as stable beams. Experiments performed at HIMAC enabled the production of high purity RIBs of 11C and 15O, through a projectile-fragmentation method, out of beams of 12C and 16O. In this work both stable and RIBs were impinged in water and PMMA phantoms, with simultaneous in-beam PET acquisition. The data were compared with the FLUKA code, evaluating the different beams’ performance in the context of Hadrontherapy, namely: dose deposition, fragmentation, LET Dose maps and annihilation events at rest evolution in time. The results of this study indicate an improvement of imaging signal due to higher presence of short-lived β+ emitters in the vicinity of the Bragg Peak. This effect, combined with in-beam PET acquisition, allows for range verification with reduced irradiation time, while decreasing degradation of dose-PET correlation due to biological washout.

Keywords: Dosimetry, FLUKA, Hadrontherapy, HIMAC, In-beam PET Imaging
Poster panel
(face) ID: 324

Poster Number:

Potential of Radioactive Ion Beams for Therapy and PET imaging (#2536)

A. Mohammadi1, E. Yoshida1, H. Tashima1, F. Nishikido1, A. Kitagawa2, T. Yamaya1

1 National Institute of Radiological Sciences, Department of Radiation Measurement and Dose Assessment, Chiba, Chiba, Japan
2 National Institute of Radiological Sciences, Heavy-Ion Radiotherapy Promotion Unit, Chiba, Chiba, Japan


In advanced ion therapy, visualization of the range of incident ions in a patient body is important to exploit the advantages of this therapy. It is ideal to use radioactive beams for in-beam positron emission tomography (PET) imaging in particle therapy due to the high quality of PET images. We have already shown the feasibility of this idea through an in-beam PET study for a 11C ion beam using our OpenPET system. An 15O beam has potential for treating hypoxic tumors and its short half-life (122 s) makes it more convenient for PET imaging with a short PET scanning time compared to 11C with a half-life of 20 min. In this work, we focus on investigation of the potential difference between the Bragg peak position and the position of the maximum detected positron-emitting fragments by a PET system for radioactive beams of 11C and 15O. For this purpose, we calculated the depth doses in water and the yield of positron emitting-fragments in a PMMA phantom for both beams using the Monte Carlo method-based code, PHITS. The measured depth doses were reproduced well by the simulation for both beams. The simulated results for the positron-emitting fragments were in good agreement with the measured activity by our OpenPET. The differences between the Bragg peak position and the position of the maximum positron-emitting fragments in an irradiated PMMA phantom for both beams were around the spatial resolution of the OpenPET. We observed an increase of differences between the Bragg peak position and the position of the maximum positron-emitting fragments when the 15O beam was broadened, although the differences were almost zero for the ideal mono-energy beam. At the conference, in addition to the above, we will discuss a conversion algorithm from the PET peak position to the Bragg peak position for a radioactive ion pencil beam with a practical energy spread.

Keywords: heavy ion therapy, in-beam PET, PET imaging, radioactive ion beams, range verification
Poster panel
(face) ID: 327

Poster Number:

Development of simple proton CT system with novel MCS correction methods (#2805)

M. Takabe1, T. Masuda1, M. Arimoto1, J. Kataoka1, S. Tanaka2, T. Nishio3, T. Toshito4, M. Kimura4, T. Inaniwa5

1 Waseda University, Graduate School of Advanced Science and Engineering, Shinjuku-ku, Tokyo, Japan
2 The University of Tokyo, Department of Nuclear Engineering and Management, School of Engineering,, Bunkyo-ku, Tokyo, Japan
3 Tokyo Women’s Medical University, Department of Radiation Oncology, Shinjuku-ku, Tokyo, Japan
4 Nagoya Proton Therapy Center, Nagoya-shi, Aichi, Japan
5 National Institute of Radiological Sciences, Department of Accelerator and Medical Physics, Chiba-shi, Chiba, Japan


Proton therapy has been developed rapidly in recent years. Because of the high dose concentration of protons, the proton range in a patient’s body, which is characterized by the water equivalent length (WEL), should be accurately predicted for safe and effective proton therapy. A current treatment plan in proton therapy is based on X-ray computed tomography (CT) imaging, which causes an uncertainty of a few percent in the proton range due to the different physical mechanism between protons and X-rays. A more accurate prediction of the proton range requires a proton CT system. There is a well-known proton CT system, based on silicon strip detectors. However, this system is complex and expensive. We developed a simple proton CT system consisting of a CCD camera and an X-ray intensifying screen. Experiments were performed with 70-MeV and 200-MeV proton beams with two beam types, i.e., broad and narrow beams, mimicking the passive scattering and spot scanning systems used in proton therapy. To reduce blurring of images caused by multiple Coulomb scattering (MCS), we developed correction methods and successfully obtained clear images with little MCS effect. The WEL values between the acquired CT and simulated images are in good agreement within a 1-σ error for polymethyl methacrylate (PMMA), isopropyl alcohol, and Vaseline. 

Keywords: CCD camera, proton therapy, water equivalent length, proton CT
Poster panel
(face) ID: 330

Poster Number:

Comparative study of the performance of proton imaging set-ups in terms of spatial resolution (#3135)

N. Krah1, M. Fontana3, I. Rinaldi2

1 CNRS/UMR 5822-5220/CREATIS/Inserm U1044/INSA/Centre Léon Bérard/University 1, Lyon, France
2 CNRS/UMR 5822/University 1 Lyon, Villeurbanne, France
3 IPNL, Villeurbanne, France


Proton imaging (PI) has long been proposed as alternative or complementary imaging modality in ion beam therapy offering a direct probe of the relative stopping power. However, it is not yet integrated into the clinical work flow. PI suffers from limited spatial resolution due to multiple Coulomb scattering (MCS).

MCS causes protons to traverse matter on non straight stochastic trajectories leading to spatial blurring in PI. The most likely path (MLP) is a statistical estimate of proton trajectories based on information about position and angle of the protons before and after the imaged volume. We have extended and devised the mathematical framework of the MLP. To account for beam geometry we have introduced a cone beam system of reference. We have further included expressions to describe the uncertainty on the proton position and angle. We solved convolution integrals analytically. As a quantitative figure of merit for the intrinsic spatial resolution we use the average width of the uncertainty envelope around the MLP. Based on this, we compare typical types of PI set-ups (PIS).

The spatial resolution of a PIS with passive field and pixel detector is only 5 times lower compared to single tracking (ST) device, provided the protons' entry and exit angles are measured accurately. For an uncertainty of +/- 1 degree, ST is comparable to a set-up combining pencil beam scanning (PBS) and a pixel detector. PBS in combination with a range telescope is limited by beam spot size for thinner phantoms (<20 cm), but comparable to a passive field coupled to a pixel detector for thick (>20 cm) phantoms.

The derived expression for the MLP is compact and easy to implement numerically. Through numerical parameters in the expression, different kinds of PIS can conveniently be analyzed without the need to re-derive any formulae. The comparison study helps to select an appropriate PIS depending on the requirements on spatial and water equivalent thickness resolution.

Keywords: proton imaging, proton radiography, proton tomography, spatial resolution, multiple Coulomb scattering
Poster panel
(face) ID: 333

Poster Number:

Scintillating fiber optic dosimeter for brachytherapy: results for a breast phantom. (#3247)

L. Moutinho1, 2, M. Torres1, F. Castro1, 2, J. Melo2, A. Goncalves1, H. Freitas2, S. Pinto5, P. Silva2, L. Peralta3, 4, A. Pereira5, J. Santos5, M. Costa6, J. Veloso1, 2

1 Universidade de Aveiro, I3N - Department of Physics, Aveiro, Portugal
2 NU-RISE, Aveiro, Portugal
3 Universidade de Lisboa, Faculdade de Ciências, Lisboa, Portugal
4 Laboratório de Instrumentação e Fisica Experimental de Particulas, Lisboa, Portugal
5 IPOPFG - Porto, Porto, Portugal
6 APNOR - Associação de Politecnicos do Norte, Porto, Portugal


Brachytherapy is a radiotherapy modality where the radioactive material is placed close to the tumor, being a common treatment for skin, breast and prostate cancers. These treatments can be of low-dose-rate, using isotopes with a mean energy of 30 keV, or high-dose-rate, using isotopes such 192 Ir with a mean energy of 380 keV. Currently, these treatments are performed without proper dosimetry for quality control and quality assurance. We developed a dosimeter base on fiber optics and SiPM using small diameter probes that can be inserted into the patient’s body using standard brachytherapy needles. By performing real-time dosimetry in breast and prostate brachytherapy it will be possible to perform real-time dose correction when deviations from the treatment plan are observed. The developed dosimeter was evaluated in-vitro. The studies consisted in the characterization of the dosimeter with 500 μm diameter sensitive probes using an in-house made gelatin breast phantom with a volume of 566 cm³. A breast brachytherapy treatment was simulated considering a tumor volume of 27 cm³ and a prescribed absolute dose of 5 Gy. The dwell times estimated from the experimental measurements are in agreement to the prescribed dwell times, with a relative error below 3%. The measured signal-to-noise ratio (SNR) including the stem effect contribution is below 3%.

Keywords: dosimeter, scintillators, scintillating optical fiber, silicon photomultiplier., brachytherapy
Poster panel
(face) ID: 336

Poster Number:

Energy-Matching X-ray Fluorescence Computed Tomography: An Evaluation of the Effects of Incident Beam Energy on Imaging Performance of Nanoparticles (#3653)

J. George1, L. - J. Meng1

1 University of Illinois, Urbana Champaign, Department of Nuclear Engineering, Urbana, Illinois, United States of America


Interest has gathered around X-ray Fluorescence Computed Tomography (XFCT) due to its ability to image metals of interest without destroying the sample. Groups have explored the potential of XFCT to image the biodistribution and therapeutic efficaciousness of anti-cancer drugs such as platinum and gold nanoparticles. Yet, the relatively low sensitivity of XFCT has inhibited the modality from systemic in vivo use thus far. For instance, for GNPs, therapeutic concentrations within the tumor are on the order of 10 μg/ml (.001% wt)  , while recent benchtop XFCT studies only demonstrated the ability to detect concentrations around .25% gold by weight. Thus, the sensitivity must be significantly improved to make XFCT applicable in the pre-clinical/clinical setting.

This work investigates specifically how the incident X-ray beam spectra (filtered, monochromatic, etc) can affect the signal-to-noise ratio (SNR), and therefore the sensitivity, of the system. We propose, therefore, (1) a semi-theoretical model for estimating a SNR value for given incident X-ray spectra and element of interest i.e. yttrium and hafnium, (2) an experimental evaluation of these schemes on the XRF spectra, and lastly (3) an imaging study of a plastic mouse phantom with an yttrium or hafnium nanoparticle loaded channel to examine the incident beam effect on image quality through SNR and contrast-to-noise ratio (CNR).  In both the semi-theoretical and experimental cases, a 17.4 keV monochromatic beam will also be used as an incident beam. The energy is right above the K-edge of yttrium (17.07 keV), which we believe would provide a greater fluorescent yield. Any advantages to such “energy-matching” will also be examined.

Keywords: XFCT, X-ray, Fluorescence, Nanoparticle
Poster panel
(face) ID: 339

Poster Number:

Interaction Vertex Imaging with a Proton CT Scanner for In Vivo Dosimetry and Range Verification in Particle Therapy (#3845)

V. Bashkirov1, R. Schulte1, 5, R. Johnson2, M. Pankuch3, V. Rykalin4

1 Loma Linda Univerisity, Basic Sciences / Biomedical Engineering Sciences, Loma Linda, United States of America
2 University of California Santa Cruz, Santa Cruz Institute of Particle Physics, Santa Cruz, United States of America
3 Northwestern Medicine, Chicago Proton Center, Warrenville, Illinois, United States of America
4 ProtonVDA Inc, Batavia, Illinois, United States of America
5 University of California San Francisco, Radiation Oncology, San Francisco, United States of America


In this work, we report first results of the feasibility of therapeutic proton beam range verification with a proton computerized tomography (pCT) scanner registering particles scattered at large angles. The pCT scanner was placed at 90 degrees to the proton beam axis at the Northwestern Medicine Proton Center facility in the vicinity to phantoms irradiated with therapeutic proton beams. The pCT scanner was used to track deeply scattered particles and measure their energy. Experimental data were acquired with a cylindrical water phantom and pediatric head phantom and used to estimate proton range in the phantoms. The estimate of the proton range in the phantom was based on an analysis of scattered particle yield, particle energy loss in the phantom, and tracking information. An obvious correlation between the measured distributions along the beam axis of the charged particle yield Nsctr, particle energy loss E, the product E*log(Nsctr), and planned proton range in water was found. We also performed a Monte Carlo simulation study using a detailed Geant4 model of the experimental setup with the water phantom. The measured distributions were in reasonable agreement with the predictions from simulations.

Keywords: hadron therapy, protons, range verification, in vivo dosimetry, interaction vertex imaging, proton CT scanner
Poster panel
(face) ID: 342

Poster Number:

Data Driven Cone Beam CT Motion Management for Radiotherapy Application (#4092)

A. Akintonde1, 2, J. McClelland1, H. Grimes3, S. Moinuddin3, R. Sharma5, S. Rit4, K. Thielemans2

1 University College London, Centre of Medical Imaging Computing, London, United Kingdom of Great Britain and Northern Ireland
2 University College London, Institute of Nuclear Medicine, London, United Kingdom of Great Britain and Northern Ireland
3 University College London Hospital, Department of Radiotherapy, London, United Kingdom of Great Britain and Northern Ireland
4 Universite de Lyon, Creatis Medical Imaging Research Center, Lyon, France
5 University College London, University College London Cancer Institute, London, United Kingdom of Great Britain and Northern Ireland


The ability to identify respiratory motion is crucial during radiation therapy treatment. In our study we introduced a novel data driven method based on principal component analysis to extract a signal related to respiratory motion from cone beam CT projection data. Projection data acquired on cone beam CT devices normally has two motion component information within it, (1) respiratory induced motion and (2) detector rotational induced motion. Our novel approach of extracting respiratory induced motion signal from projection data was based on computing PCA for different section of the data set independently, and introducing a technique of combining the extracted signal from each section in a manner to represent the respiratory signal from the entire data set. We tested our method using simulation data set from XCAT software and a real patient data set. The respiratory signal extracted with the XCAT simulation yielded a correlation of 0.9341 when compared to the ground truth respiratory signal. The comparison of our method to other techniques also showed similar result. Initial results for the real patient data set are encouraging but show need for futher refinements.

Keywords: Radiotherapy, Cone Beam Computed Tomography, Principal Component Analysis, Respiratory Motion
Poster panel
(face) ID: 345

Poster Number:

Validation of a realistic simulation of the HRRT using SimSET (#2014)

E. Raptis1, L. Parkes2, J. Anton-Rodriguez1, S. F. Carter2, K. Herholz2, J. C. Matthews1

1 The University of Manchester, Division of Informatics, Imaging, and Data Sciences, MAHSC, Manchester, United Kingdom of Great Britain and Northern Ireland
2 The University of Manchester, Division of Neuroscience and Experimental Psychology, MAHSC, Manchester, United Kingdom of Great Britain and Northern Ireland


Monte-Carlo simulations of the High Resolution Research Tomograph (HRRT) (Siemens/CTI HRRT PET brain scanner) were developed and conducted using SimSET. We were interested in conducting simulations with comparable resolution to that of a real HRRT in order to investigate partial volume correction. A series of test objects were simulated including point sources, a 3-D Shepp-Logan digital phantom and a brain amyloid scan and compared to measured HRRT data. We found that we needed to include additional detection mispositioning in order to match the resolution of real HRRT data. With this additional mispositioning we got comparable spatially variant resolutions between the simulations and real data, although the simulated data was less noisy than the real data.

Keywords: Positron Emission Tomography, HRRT, Monte-Carlo simulations, spatial resolution
Poster panel
(face) ID: 348

Poster Number:

Preliminary Brain SPECT Multi-Pinhole Collimator Mechanical Design for DaTscan Imaging (#2109)

Y. He1, 2, A. Könik1, J. De Beenhouwer3, T. J. Fromme1, 2, C. Lindsay1, K. S. Kalluri1, X. Li4, N. C. Momsen4, L. R. Furenlid4, M. A. King1

1 University of Massachussets Medical School, Radiology, Worcester, Massachusetts, United States of America
2 Worcester Polytechnic Institute, Worcester, Massachusetts, United States of America
3 University of Antwerp, iMinds-Vision Lab, Antwerp, Belgium
4 University of Arizona, Radiology, Tucson, Arizona, United States of America


A mechanical design using SOLIDWORKS®, a 3D Computer-Aided Design (CAD) software, is reported for constructing a multi-pinhole (MPH) collimator that will be used with a fan beam collimator for high resolution and sensitivity imaging of the striatal region of patients suspected of having Parkinson’s disease. The design strategy was to alter an existing commercial single-pinhole collimator by inserting the MPH components while keeping the portion of the collimator that attaches to the head of the gamma-camera. The MPH array consists of 9 apertures focused at a common point at the center of the striatal region of the brain with their entrance and exit ports shaped to allow no overlapping of their projections on the detector of the gamma-camera. A shutter mechanism is included at the side of the aperture plate towards the detector, which will allow additional 8 apertures to image independently, thus increasing sensitivity, but with the projections multiplexed with those of the original pinholes. Testing of the material properties of the potential alloys used in the construction of the MPH collimator provided validated values which were used in SOLIDWORKS simulation of structural deformation during operation to ensure safety and imaging accuracy. The CAD model of the MPH collimator was imported into the GATE Monte Carlo package to simulate brain imaging and validate the model geometry.

Keywords: GATE, Multi-pinhole, SPECT, XCAT
Poster panel
(face) ID: 351

Poster Number:

Simulation-based Design Study of a PET Detector for High Performance Brain PET-MR (#2268)

M. Lenz1, 2, L. L. Caldeira1, 2, H. Xu1, N. J. Shah1, 3, U. K. Pietrzyk1, 2, C. W. Lerche1

1 Forschungszentrum Jülich GmbH, Medical Imaging Physics / Institute of Neuroscience and Medicine, Jülich, North Rhine-Westphalia, Germany
2 University of Wuppertal, School of Mathematics and Natural Sciences, Wuppertal, North Rhine-Westphalia, Germany
3 JARA, RWTH Aachen University, Department of Neurology / Faculty of Medicine, Aachen, North Rhine-Westphalia, Germany


In order to find a favourable design for a new UHF MR compatible high-performance brain PET insert, simulation studies are performed with the toolkit GATE. Four scanner configurations with comparable scintillator volume are compared, showing that the option with the smallest number of modules but largest module size yields the highest peak sensitivity of 13.7%. Moreover, different lightguide thicknesses are examined with respect to identifiability of individual scintillator pixels by simulating the propagation and detection of optical photons. The thinnest lightguide (1.00mm) is preferred, yielding the best point separability in floodmaps. Comparisons with measured data are work in progress.

Keywords: Positron Emission Tomography (PET), Sensitivity, Depth of Interaction (DOI), Simulation
Poster panel
(face) ID: 354

Poster Number:

A fast approach for whole-body voxel-based dosimetry: multiple voxel S-value approach (#2578)

M. S. Lee1, J. H. Kim2, J. S. Lee1, 3

1 Seoul National University, Interdisciplinary Program in Radiation Applied Life Science, Seoul, Republic of Korea
2 Korea Research Institute of Standards and Sciences, Daejeon, Republic of Korea
3 Seoul National University, Biomedical Sciences, Seoul, Republic of Korea


Patient-specific dosimetry is getting important due to growing interests in personalized medicine. Generally, voxel-based dosimetry with dose point kernel or voxel S-value (VSV) is widely used. However, these approaches cannot take medium heterogeneity into account. Direct Monte Carlo (DMC) is known as a state-of-art dosimetry, but it requires extensive computational cost and time. Here, we propose a new method, multiple VSV approach, to take medium heterogeneity into account with extensively reduced computational time

The proposed approach uses multiple VSVs instead of using only water VSV. Different numbers (N) of VSVs that cover the whole-body density ranges were pre-calculated with GATE simulation. N density-specific dose maps were acquired by VSV convolution over PET cumulated activity map, and CT-based segmentations were conducted for each density region. Final dose map was acquired by summing N segmented density-specific dose maps. Different Ns were tested: N=1 (single water VSV), 4, 6, 8, 10, and 20. Multiple VSV approach was implemented on a digital phantom and patient data (n=80) and compared with GATE DMC. Finally, patient dosimetry (n=8) were conducted with multiple VSV and compared with DMC and MIRD dosimetry. Errors at voxel and organ-level were reported for eight organs.

Multiple VSV approaches agreed with DMC well in phantom and patient studies and showed more accurate results than single VSV in lungs and bones. False estimations were observed at lung boundaries, while errors decreased as N increased. Multiple VSV (N=8) showed the best results with voxel-level errors of -5.36%. For patient dosimetry, our proposed method showed much improved results than single VSV and MIRD dosimetry. Organ-level errors were -4.64%, -13.02%, and 18.51% for multiple VSV, single VSV, and MIRD dosimetry, respectively.

In conclusion, we proposed a fast voxel-based dosimetry approach by using multiple VSVs and showed comparable results with DMC with ⅹ1/850 reduced time.

Keywords: Radiation dosimetry, GATE simulation, Voxel S-value, Heterogeneous medium
Poster panel
(face) ID: 357

Poster Number:

An Investigation of Quasi-Vertex Views in Brain SPECT Imaging-Initial Results (#2894)

J. C. Goding1, K. S. Kalluri2, A. Könik3, Y. He4, G. Zubal6, L. R. Furenlid5, M. A. King7

1 UMass Medical School, Dept. of Radiology, Nuc. Med. Div., Worcester, Massachusetts, United States of America
2 UMass Medical School, Dept. of Radiology, Nuc. Med. Div., Worcester, Massachusetts, United States of America
3 UMass Medical School, Dept. of Radiology, Nuc. Med. Div., Worcester, Massachusetts, United States of America
4 UMass Medical School, Dept. of Radiology, Nuc. Med. Div., Worcester, Massachusetts, United States of America
5 UArizona, College of Optical Sciences, Tucson, Arizona, United States of America
6 Yale School of Medicine, Clinical Neurosciences Imaging Center, New Haven, Connecticut, United States of America
7 UMass Medical School, Dept. of Radiology, Nuc. Med. Div., Worcester, Massachusetts, United States of America


A next-generation, adaptive, brain-imaging, single photon emission computed tomography system is currently under development at the Universities of Massachusetts and Arizona. The proposed multi-pinhole based modular gamma camera configuration enables the acquisition of “quasi-vertex” views i.e., close to the vertex views. This study is concerned with understanding how activity inferior to the brain will influence these views and ultimately, the reconstruction of the volume of interest.  Analytical models can provide some measure of the level of direct impact from radioactive emissions originating in an organ or tissue outside of the region of interest. Scattered gamma radiation will be detected as well but given the difficulty in modeling such phenomena, Monte Carlo simulations are used to quantify its effect. Using computer generated phantoms, the influence of activity from organs and tissues in the thorax, neck and lower head in comparison to brain structures in the quasi-vertex views will be detected, identified and quantified. The simulation consists of two components: the detectors and the sources. The detectors are simulated gamma cameras modules consisting of a tungsten pinhole collimator, an air gap, and a NaI(Tl) scintilator surrounded by lead shielding. Two detector positions were considered for the preliminary work: one at a 93.5° angle of elevation from the longitudinal axis (frontal view), the other at 0° (the vertex). The sources were three phantoms with the same amount of activity (I-123, primary gamma photons at 159 keV) set in only the striatum, salilvary glands and the liver respectively. An unexpected result was that activity from the liver, recorded by the frontal view detector was higher than that seen at the vertex. By changing the shielding to a higher density material, no such detections were observed. This result highlights the need for high fidelity detector and source models for realistic simulations.

Keywords: SPECT, Brain imaging, Quasi-vertex view, GATE simulation, XCAT phantom
Poster panel
(face) ID: 360

Poster Number:

Double Photon Emission Coincidence Tomography (DPECT): Hybrid Single and Coincidence Photon Imaging for Cascade Decay Isotopes (#3267)

L. Cheng1, 2, H. Liu1, 2, P. Fan1, 2, Y. Liu1, 2, T. Ma1, 2

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


The main problem of SPECT is the low sensitivity, which leads to higher noise. Different from most widely used radionuclide, Tc-99m, In-111   was used  in our study. Based on detection of cascade 171 and 245 keV photons, more accurate location information can be obtained from the coincidence cascade event. We performed 2D numerical simulation and compare the benefit of cascade events. Results have shown that noise will be reduced up to 10% without or with 10 mm spatial blurring. And If the detector sensitivity can be improved to ~1%, the noise reduction can be up to 50% and more. Furthermore, single angle reconstruction may be practical using DPECT, which may simplify clinical operation and is more user-friendly.  

Keywords: In111, cascade photons, coincidence
Poster panel
(face) ID: 363

Poster Number:

easyPET: An EDUGATE simulation toolkit based on the educational easyPET system (#3656)

P. M. M. Correia1, J. Menoita1, A. L. M. Silva1, N. O. Romanyshyn1, J. F. C. D. A. Veloso1

1 i3n- Departamento de Física da Universidade de Aveiro, - Departamento de Física da Universidade de Aveiro, Aveiro, Portugal


EasyPET is a new concept of a Positron Emission Tomography (PET) scanner using an innovative acquisition method based on two rotation axes for the movement of detector pairs. Due to its simplicity, its suitable for educate porpuses, to teach students about the PET tecnhology and its basic concepts, from the radiation detecting and analog pulse analysis to the concidence sorting and image reconstruction. The concept allows achieving high and uniform position resolution over the whole field of view (FoV), by eliminating parallax errors due to the depth of interaction (DoI), which are typical of ring based PET systems, so quality images are obtained even without state-of-the-art image reconstruction algorithms. The technology developed at the University of Aveiro is pattend-pending, is licended to CAEN S.p.A, and included in the educational catalog of the company. In this work, a simulation toolkit based in the easyPET available from CAEN is presented. It is able to simulate all the physical aspects of the product, such us the scaning range, variable Field-of-View (FOV), scintilator energy resolution, coincidence time and energy window, among others. A simple image reconstruction algorithm based on Filtered-back-projection (FBP) is distributed, and the output of the toolkit gives a Sinogram and a Raw backprojection image out-of-the-box for a quick analysis in classroom. 

Keywords: simulations, GATE, PET, easyPET
Poster panel
(face) ID: 366

Poster Number:

Quasi 2D computation of parallel hole collimator point spread function with septal penetration (#4023)

K. Kacperski1, D. Świtlik1

1 The Maria Skłodowska-Curie Memorial Cancer Centre and Institute of Oncology, Medical Physics, Warsaw, Poland


A simple method for computing a parallel hole collimator point response functions has been proposed. It only requires a 2D image of the collimator hole pattern as the input to calculate total photon paths through septa, and subsequently, the fluxes of geometric and septal penetrating photons. 2D collimator hole pattern is much  easier to prepare than a fully 3D model, and can be done using widely available software. The average fluxes of geometric and septal penetration photons are computed exactly. The scattered photons can be approximated using e.g. the buildup factors, known from radiation shielding. The method can be useful as a quick check in collimator design, before a fully 3D simulation is used. Point spread functions have been computed for several collimators and photon energies, and compared with those measured on a gamma camera. Less than 5 % accuracy has been obtained for the geometric part, and 10 – 20 % outside the geometric region.

Keywords: SPECT, collimator design, septal penetration
Poster panel
(face) ID: 370

Poster Number:

The simulation and reconstruction of mixed detector cylinder PET system (#4200)

H. Xu1, S. Wang1, Y. Shang1, W. Cao1, M. Gao1, C. - M. Kao3, Q. Xie1, 2, P. Xiao1, 2

1 Huazhong University of Science and Technology, Biomedical Engineering Department, Wuhan, Hubei, China
2 Huazhong University of Science and Technology, Wuhan National Laboratory for Optoelectronics, Wuhan, Hubei, China
3 Univesity of Chicago, Department of Radiology, Chicago, Illinois, United States of America


Spatial resolution of a PET system depends on the size of crystal to some extent. In general, the smaller the size of each crystal, the higher the spatial resolution of the PET system. The problem is that with the crystal size becomes smaller, more electronic channels will be needed and the cost of the PET system will increase. Besides, sometimes when imaging some organ like bladder, high spatial resolution is not that necessary. To fix these problems, here we put forward a new method that combines the high spatial resolution crystal with the normal spatial resolution crystal to construct a novel mixed detector module and then we used this module to construct a cylinder PET system. One part used small-size crystals and another part used normal-size crystals in this detector module. Inspired by this method, we can combine high resolution detector module with existing PET machine to construct a mixed detector PET system very quickly. The results demonstrated that these kind of mixed detector cylinder PET system had a good combination of this two kinds of crystals, and we can get different quality reconstruction image by adjusting the position of the target.

Keywords: PET, Mixed Detector, Spatial resolution, Image reconstruction