IEEE 2017 NSS/MIC/RTSD ControlCenter

Online Program Overview Session: M-03

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

MIC Poster Session I

Shortcut: M-03
Date: Wednesday, October 25, 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: 1

Poster Number:

Measurements with a PET Coincidence Setup Based on the PETA5 ASIC and FBK RGB-HD SiPMs (#1040)

D. Schug1, P. Gebhardt1, B. Weissler1, N. Groß-Weege1, T. Dey1, V. Schulz1, 2

1 RWTH Aachen University, Department of Physics of Molecular Imaging Systems, Institute for Experimental Molecular Imaging, Aachen, Northrhine-Westphalia, Germany
2 Philips, Research Europe, Aachen, Northrhine-Westphalia, Germany


We present results obtained with a positron emission tomography (PET) coincidence setup provided by the group of Peter Fischer from the University of Heidelberg. The setup employs two facing detector blocks which are build from an LYSO matrix of 30 x 30 x 10 mm3 in size with a crystal pitch of 2.5 mm read out in a one-to-one readout scheme by FBK RGBHD SiPMs with a SPAD size of 25 μm. Each SiPM channel is digitized by a single input channel of a PETA5 (Position-Energy-Timing ASIC). Four PETA5 chips, each with 36 inputchannels, are mounted on one of the ceramic detector PCBs which incorporate a liquid cooling maze. We used multiple 22Na point sources for calibration and evaluation purposes. In the calibration and result extraction, all available channels were taken into account simultaneously. The time-to-digital-converter (TDC) bins were calibrated and the noise pedestal for each channel was determined. The energy was calibrated and corrected for saturation using the 511 keV and the 1275 keV peak, identified in the noise-pedestal-corrected ADC-value spectrum. Time offsets of all readout crystals were calibrated using coincident and qualified singles. The distribution of the TDC bin sizes was evaluated to have a mean of 50.2 ps and a standard deviation of approximately 20.2 ps. We determined the energy resolution of the setup for qualified coincidences as 13.0% and the coincidence resolution time as approximately 310 ps. Optimizations of the calibration and evaluation are still ongoing and the final status will be presented.

Keywords: SiPM, PET, ASIC, time-of-flight, FBK RGB-HD, PETA5
Poster panel
(face) ID: 4

Poster Number:

A systematic study on the strip-line readout method for SiPM-based TOF PET (#1262)

H. Kim1, Y. Hua3, X. Lyu2, F. Xu2, Y. Wang2, J. Lyu3, C. Tian3, N. Eclov1, C. - T. Chen1, Q. Xie3, C. - M. Kao1

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


Signal multiplexing for SiPM-based TOF PET to achieve high multiplexing ratio and keep the fast timing performance is a challenging issue. We have developed a multiplexing method using a strip-line; multiple SiPMs are connected to a single micro-strip transmission line, and the position of hit SiPM is determined by the signal arrival time difference on the strip-line. The strip-line method is scalable; high multiplexing ratio can be easily achieved by increasing the number of SiPMs connected on a strip-line. To test the scalable feature of the strip-line readout systematically, we have implemented several strip-lines with different trace lengths and varied the number of SiPMs (Hamamatsu S13361-3050NE) on the strip-lines up to 32. Experimental tests have been conducted to assess the influences of the strip-lines on detector performance: position, energy, and coincidence time resolution. In the paper, preliminary results from the study are presented, and our ongoing efforts on strip-line readout for TOF PET are discussed.

Keywords: signal readout, strip-line, TOF PET
Poster panel
(face) ID: 7

Poster Number:

Development of a Multi-Pixel Photon Counter Module with an Unified Interface (#1345)

K. Shimizu1, S. Nakamura1, H. Tozuka1, K. Yamamoto1

1 Hamamatsu Photonics K.K., Hamamatsu, Shizuoka, Japan


A new PET detector module which has a unified interface is proposed. The detector module consists of a scintillator array, a multi-pixel photon counter (MPPC) array and a front-end circuit. While a general PET detector module has three kinds of interface; power, control and data, the proposed detector has only one Ethernet cable. Power of the detector module is fed by a Power over Ethernet (PoE) method, which is configured to pass electric power along with data on twisted pair Ethernet cable. The clock used in the module is recovered from the serial data and the recovered clock is used to acquire and send the data back. The synchronization can be done using IEEE1588 method with software correction. Data are transmitted through User Datagram Protocol (UDP) to an uplink. The uplink can send the request with current timestamp to a module, the module responds in a fixed time delay, and the uplink detects the response and compare to the current timestamp. The difference detected by the uplink would be the fixed delay plus cable delay. Using proposed modules, it is expected flexible, scalable and upgradable systems can be easily constructed. Using proposed modules, it is expected flexible, scalable and upgradable systems can be easily constructed.

Keywords: MPPC, PET, Detector Module
Poster panel
(face) ID: 10

Poster Number:

Preliminary Evaluation of a Depth-Encoding PET Detector Inserting Horizontal-Striped Glass Between Crystal Layers (#1665)

J. Yang1, K. B. Kim2, Y. choi2, J. Kang1

1 Chonnam National University, Department of Biomedical Engineering, Yeosu, Republic of Korea
2 Sogang University, Department of Electronic Engineering, Seoul, Republic of Korea


In our previous work, a depth-encoding PET detector inserting glass plate between crystal layers was designed and examined. The aim of this study was to propose and examine the newly designed DOI detector replacing the light guide from the plate shape to the horizontal-striped shape. Even though the detector components is similar, the detector concept spreading scintillation light is changed from dispersion in all (X-Y) directions to only X-direction using horizontal stripes glass.

DOI detector consisted of 4×4 LYSO arrays with an individual size of 3.17×3.17×10 mm3. All crystals were polished and separated by 0.2 mm thick white paint. The horizontal-striped glass with 1×4 array of 12×3 mm2 surface was inserted between top- and bottom-crystal. The bottom surface of bottom-layer was optically coupled to 4×4 GAPD array with a pixel size of 2.85×2.85 mm2. The 16 signals from GAPD array were multiplexed by the modified RCD networks and they were fed into the custom-made charge sensitive preamplifiers using high speed operational amplifiers. The 4 output signals were digitized and recorded by high speed oscilloscope based DAQ unit.

Experimental results showed that all crystal pixels could be identified along with the X, Y, and depth directions when the crystal blocks were stacked with the horizontal-striped glass of 3 mm thickness. The energy spectra that correspond to individual crystal position of the flood histogram could be acquired and the mean energy resolutions of the top-layer and the bottom-layer were 48% and 45%, respectively.

This study demonstrated that it is feasible to extract the 3D γ-ray interaction position from the 2D flood histogram obtained with the proposed DOI-PET detector. Further experimental study will be followed to characterize the DOI-PET detector performance including the light loss, count rate loss, energy resolution and time resolution as a function of horizontal-striped glass plate thickness.

Keywords: Depth of interaction (DOI), depth-encoding PET detector, glass plate, light sharing process
Poster panel
(face) ID: 13

Poster Number:

Development of a dynamic micro RI imaging system for single cells (#1950)

G. Hirumi1, 2, F. Nishikido1, H. Tashima1, H. Wakisaka1, T. Higuchi3, H. Haneishi2, T. Yamaya1, 2

1 National Institute of Radiological Sciences, Chiba, Japan
2 Chiba University, Chiba, Japan
3 University of Wurzburg, Wurzburg, Germany


Studies on cell regulation are attracting worldwide attention in order to realize regenerative medicine. Therefore, a nuclear medicine imaging method, which can use tracers having substantially the same composition as a target biomolecule, is required. Autoradiography is a high-resolution nuclear medicine imaging method. However this method does not have a dynamical imaging capability due to the principle on which it is based. In this research, we develop a nuclear medicine imaging system for dynamic cell observation.

Specifically, in order to prevent broadening of the scintillation position because of scintillation light spreading in the scintillator, β-rays are detected by a thin scintillator plate. A scientific CMOS camera with low readout noise and high resolution was used to detect scintillation light. The scintillator plate was a CsI crystal (150 μm thick) connected to an optical fiber (6 μm diameter) array. The scintillation light generated from the scintillator plate was extracted through optical fibers. The imaging lenses adopted in this research were a high-speed imaging and a conversion lens. The F-number of the former was set to 0.9 for a short exposure time, with a large amount of light passing through. Imaging resolution was adjusted to 6.5 × 6.5 mm2 / pixel and exposure time was 10 s.

In order to evaluate the imaging performance of the proposed system, we measured samples containing a radiopharmaceutical with a prototype system. In this measurement, two different activity 18F-solutions were used as the imaging target. They were fixed with a uniform thickness of 0.4 mm.

In the imaging results, we measured different image values depending on their activities. On the other hand, in the region of no 18F-solutions, the image value was almost constant. The output had sufficient linearity for images with activity exceeding 0.04 Bq / pixel. Imaging results showed that our system has sufficient sensitivity for imaging uptake of fluorodeoxyglucose by single cells.

Keywords: CMOS, beta-ray, cell imaging, Scintillation detector
Poster panel
(face) ID: 16

Poster Number:

Time and energy performance of a temporal imaging LYSO and CeBr3 detectors (#2089)

C. Tata Zafiarifety1, 2, A. Iltis1, H. Snoussi2, G. T. Zeufack1, M. Z. Hmissi2, L. Rodrigues1, E. Lamprou3, A. Aguilar3, V. Ilisie3, A. J. González3, J. M. Benlloch3

1 Damavan Imaging, Rosieres Pres Troyes, France
2 University of Technology of Troyes, UMR CNRS 6281, Troyes, France
3 Universitat Politècnica de València, Instituto de Instrumentación para Imagen Molecular (I3M), Centro Mixto CSIC, Valencia, Spain


The use of scintillators such as the CeBr3 or LaBr3 in gamma ray detectors for positron electron tomography (PET) are promising due to their inherent properties enabling to measure the annihilation photons time of flight (TOF). In particular, monolithic scintillators exhibit a number of interesting properties such as high gamma photon capture efficiency, high energy resolution and relatively simple detector assembly. Moreover, the photon depth of interaction (DOI) can also be estimated with simpler methods, when compared to the wider used crystal arrays.

It has been proven that such detectors can have better TOF capability when the 511 keV photon DOI is also taken into account. This can be done by correcting the delay time introduced by the scintillation photons travelling through crystal at the c/n speed, where n is the refractive index. We are using a new precise method called temporal imaging to accurately localize the point of interaction for only photoelectric events.

The tests in this work have been carried out for two pairs of monolithic blocks of LYSO and CeBr3. They had dimensions of 32x32x20 mm3 and 32x32x25 mm3, respectively. Measurements were performed using a small size (1 mm) 22Na source and about 25 uCi activity. LYSO crystals were covered with ESR Vikuity whereas the CeBr3 were Teflon wrapped. Both were coupled to a Phillips DPC3200 digital SiPM.

Before DOI corrections to the original data, we determined a coincidence resolving time (CRT) for the LYSO blocks of 465±20 ps. This improves to as more as 232±6 ps after the DOI correction. These results were obtained with an energy resolution of nearing 12%. Results using the CeBr3 blocks are still under analysis.

Keywords: PET, TOF, LYSO, CeBr3, DOI, monolithic, SiPM, CRT
Poster panel
(face) ID: 19

Poster Number:

A High Resolution PET Readout Method Using Time over Threshold and Center-of-Gravity Algorithm (#2208)

D. Oh1, S. Lee1, Y. Choi1, K. Park1, J. Jung1, K. B. Kim1

1 Sogang University, Department of Electronic Engineering, Molecular Imaging Research & Education (MiRe) Laboratory, Seoul, Republic of Korea


The development of high resolution positron emission tomography (PET) is challenging due to the readout complexity occurred by increasing the PET detector channels, ADCs and TDCs. The purpose of this study was to develop PET readout circuit combining time over threshold (TOT) method with center-of-gravity (COG) algorithm to reduce the complexity of readout electronics and to obtain high resolution imaging.

A PET detector was composed of a 6 × 6 array of 2 × 2 × 20 mm3 LYSO coupled to a 4 × 4 array of 3 × 3 mm2 MPPC. COG-TOT circuit consisted of OP-AMP, comparator and sum-circuit based TOT method and data acquisition (DAQ) system was composed of 16 counters and TDC implemented using FPGA. The output signals of silicon photomultiplier (SiPM) were amplified and fed into comparators. Individual signal of amplified SiPM output exceeding threshold voltage was converted into a digital pulse output, which was calculated by FPGA with internal counter operated by 350 MHz clock. The acquired counter values were used to estimate the position of gamma interacted pixel by using COG algorithm. The flood histogram was obtained using 22Na point source to evaluate the capability of discrimination of 36 crystal pixels. An acquired flood image showed that all 36 crystals from the detector could be clearly resolved. The energy resolution of a single pixel was 7.3% FWHM.

The results of this study indicate that COG-TOT method would be useful for achieving high resolution less than the pixel size of photosensor with reduced hardware complexity.

Keywords: Time over threshold (TOT), Center-of-gravity (COG), High resolution
Poster panel
(face) ID: 22

Poster Number:

Studies on Sub-millimeter Ce:LYSO, Ce:GAGG and a New Ce:GFAG Block Detector for PET Using Digital Silicon Photomultiplier (#2339)

M. N. Ullah1, E. Pratiwi1, J. H. Park1, S. Yamamoto2, K. Kamada3, 4, A. Yoshikawa4, 5, J. - Y. Yeom1, 6

1 Korea University, Department of Bio-Convergence Engineering, seoul, Republic of Korea
2 Nagoya University Graduate School of Medicine, Department of Medical Technology, Nagoya, Japan
3 Tohoku University, New Industry Creation Hatchery Center (NICHe), Sendai, Japan
4 C&A corporation, Sendai, Japan
5 Tohoku University, Institute for Material Research, Sendai, Japan
6 Korea University, School of Biomedical Engineering, Seoul, Republic of Korea


Detector spatial, timing and energy resolution affect the image quality of a time-of-flight positron emission tomography (ToF-PET) system. These are in turn dependent on various factors such as the choice of scintillator, photo detectors, reflector material, surface treatment (rough or polished) of scintillators, etc. In this study, we investigated the performances of sub-millimeter Ce:LYSO scintillator arrays (polished and rough surfaces with BaSO4 or ESR reflector), a gadolinium aluminum gallium garnet (Ce:GAGG, rough surface with BaSO4 reflector) and a new gadolinium fine aluminum gallate (Ce:GFAG, rough surface with BaSO4 reflector). The outer dimension of all block detectors were ~ 12 mm x 12 mm with 12 x 12 matrix and individual crystal element size was 0.9 x 0.9 x 6mm3. These detectors were optically coupled to a digital silicon photomultiplier (dSiPM, DPC-3200-22-44) with a 1 mm thick light guide. Experiments were conducted at sensor temperature of ~15oC with “integration interval” of 85 ns (For LYSO) and 165 ns (For GAGG and GFAG) and the “neighbor logic” was set as ‘on’. The 2-dimensional position histogram for 22Na gamma photons showed that all pixels were clearly resolved for all block detectors (peak-to-valley ratio ranging from 5.3 to 11.1).  The average energy resolution for LYSO (ESR, rough), LYSO (ESR, polished), LYSO (BaSO4, rough), LYSO (BaSO4, polished), GAGG and GFAG arrays were measured to be 17.3%, 13.7%, 25.3%, 15.8%, 13.6%, and 16.9 % FWHM respectively. The coincidence resolving times (with 3 x 3 x 5 mm3 LYSO crystal as reference) for the LYSO (ESR, rough), LYSO (ESR, polished), LYSO (BaSO4, rough), LYSO (BaSO4, polished), GAGG and GFAG arrays were 199 ps, 218 ps, 190 ps, 223 ps, 334 ps and 285 ps respectively. In conclusion, the new GFAG scintillator may be a promising candidate for future high resolution ToF PET systems, considering the tradeoffs between performances and its potential to be grown at a lower cost.

Keywords: scintillation detector, digital silicon photomultiplier, time-of-flight positron emission tomography, timing resolution
Poster panel
(face) ID: 25

Poster Number:

TOF PET detectors employing stripline based readout and MVT digitizers (#2421)

X. Lyu1, Y. Hua2, F. Xu1, C. - M. Kao3, C. Tian2, J. Lyu2, 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 previously proposed and developed two novel technologies for PET detectors, including a multi-voltage threshold (MVT) for digital data-acquisition (DAQ) of PET signals and a stripline (SL) based multiplexing readout that is easy to implement and can support time-of-flight (TOF) detection. The MVT digitizer can be implemented by using FPGAs to obtain many DAQ channels on a small electronics board. The combined use of these two technologies therefore have the potential to enable the development of TOF PET detectors with greatly reduced electronic complexity, power consumption and cost while at the same time support upgradability. In this work, by using Hamamatsu and SensL SiPM arrays we successfully demonstrate the possibility to use one SL, and hence two DAQ channels, to read up to 36 LYSO/SiPM pixels by use of  MVT digitizers. We show that snake-shaped SL paths can be used to strategically increase the separation in the differential-time measurement between adjacent pixels to allow accurate pixel discrimination without degrading the timing properties. Therefore, the readout can be easily applied to work with pixels smaller than those considered in the current work (3.2 mm and 4.2 mm pitches).  When doubling the number of pixels on an SL from 18 to 36, there are only mild degradations in the energy resolution and timing characteristics. We have not measured the TOF resolution but based on our previous results we anticipate achieving ~300 ps. Therefore, the use of SL-based readout and MVT digitizers is very promising for developing affordable high-performance PET detectors. At the meeting, we will report  the full experimental results.

Keywords: strip line, MVT, TOF PET detector, multiplexing readout
Poster panel
(face) ID: 28

Poster Number:

Performance evaluation of a new fast time response PMT coupled with fast LGSO (#2550)

S. Lee1, M. S. Lee1, J. S. Lee1

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


In this study, we measured performances of a newly developed Hamamatsu Photonics R13478 photomultiplier tube (PMT) and compared with an existing R9800 PMT series. R13478 contains an accelerating electrode placed between the focusing electrode and the first dynode for improvement of time resolution by reduction of transit time spread which results from the spatial difference of photoelectron emissions. Also, it is reported that dark current of R13478 is considerably lower than R9800.

For the demonstration, time resolution dependencies on supply voltage, time pickoff method and crystal position were investigated with 2.9 x 2.9 x 20 mm3 fast LGSO:Ce (0.025 mol%) crystal coupled. Two PMTs faced each other for coincidence resolving time measurement. For data acquisition, we used DT5742 waveform digitizer (CAEN) in the resolution of 12 bit and the sampling rate of 5 GHz.

As a result, R13478 showed better time performance than R9800 in diverse aspects. Rise time of R13478 was 1.54 ns, 100 ps shorter than R9800 due to reduced transit time. R13478 achieved 170 ps of optimized time resolution at recommended supply voltage, while it was 187 ps for R9800. The time resolution of R13478 was significantly better when extremely high voltage was supplied or low time pickoff threshold level was applied, which are the conditions vulnerable to noise. Time resolution was comparable with R9800 in the center and the edge positions when 1500 V was supplied in common. However, a significant difference in time resolution was measured in the laid position (R13478: 144 ps, R9800: 154 ps). This indicates that accelerating electrode reduced transit time spread because the laid position of coupled crystal widened the spatial distribution of electron emissions from the photocathode.

In conclusion, we proved excellent time performance of R13478 PMT by comparing with R9800 and we suggest it as a qualified photodetector for fast timing applications.

Keywords: Photomultiplier tube, Accelerating electrode, Time resolution, Transit time spread
Poster panel
(face) ID: 31

Poster Number:

A Study on the SiPM Based Dual-Ended Readout TOF-DOI PET Module (#2706)

Y. B. Han1, H. G. Kang1, S. H. Song1, K. M. Kim2, G. B. Ko3, J. S. Lee3, S. J. Hong1, 4

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


We present the timing and depth of-interaction (DOI) resolution of a dual-ended readout silicon photo-multiflier (SiPM) positron-emission-tomography (PET) detector module. The dual-ended readout PET detector module consists of a 6 × 6 array of saw-cut LYSO crystal (2.0×2.0×20 mm3) wrapped with ESR reflector, and two SiPMs(Hamamatsu, S13361-3050NE-04, Japan). In order to investigate the effect of crystal surface treatment on the timing and DOI resolution of the dual-ended readout PET detector, saw-cut and polished LYSO crystals (2.9×2.9×20 mm3) were used. We irradiated 511 keV annihilation gamma rays (beam divergence angle=2.75°) using a 22Na point source to the LYSO scintillating crystal (2.9×2.9×20 mm3). The energy, timing and DOI resolution were measured with a 4 mm DOI step with different crystal surface. The coincidence timing resolution and DOI resolution were 472 ± 9 ps, 2.08 ± 0.2 mm for the saw-cut LYSO crystal, and 469 ± 21 ps, 4.64 ± 1.84 mm for the polished LYSO crystal, respectively. The coincidence timing resolution of 6 × 6 array of saw-cut LYSO (2.0×2.0×20 mm3) module was 544 ± 65 ps. In future, the timing and DOI resolution of the PET detector module will be optimized by using the GATEv6.2 optical simulation.

Keywords: SiPM, LYSO, TOF, DOI, PET, Dual readout
Poster panel
(face) ID: 34

Poster Number:

Investigation of an optical amplification technique to improve sensitivity of an optical property modulation-based radiation detection method for PET (#3438)

L. Tao1, H. M. Daghighian1, C. S. Levin1

1 Stanford University, Radiology, Stanford, California, United States of America


We are exploring the modulation of optical properties as an alternate to scintillation detection with a goal to drastically improve coincidence time resolution (CTR) for positron emission tomography (PET). The optical amplification effect achieved by employing high quality factor (high-Q) optical resonant cavities (known as an etalon) has shown the ability to enable ultra-high sensitivity measurement for refractive index modulation. In this work, we investigate the use of the optical amplification effect for improving the detection sensitivity for optical modulation-based radiation detection method for PET, with the ultimate goal to achieve individual annihilation photon detection with <10 ps CTR. We studied the detection sensitivity of etalon cavities made of 50%, 90% and 95% reflection cavity mirrors, and compared the sensitivity to our previous spatial interference setup. We also studied the influence of probe laser light concentration on the detection sensitivity by using a focused probe laser beam aligned to intercept the interaction region of ionizing radiation photons with the detector crystal. A detection sensitivity boost of a factor of ~70 has been observed using an etalon cavity made of 95% reflection mirrors with a <1mm diameter focused probe laser beam, compared to our previous spatial interference setup. We also found that the rise time of the modulation signal is not affected by using high-Q cavities.

Keywords: ToF-PET, time resolution, ionizing radiation detection, optical modulation based detection, optical amplification, etalon, BSO
Poster panel
(face) ID: 37

Poster Number:

Timing Estimation Algorithm Incorporating spatial position for Monolithic PET Detector (#3643)

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

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


In pixelated PET detector, extraction of timing information is very simple and does not need any complex algorithm. However, in monolithic PET detector, the conventional timing extraction method is not optimal. Due to optical photon transport within monolithic crystal, the timing information of each SiPM signal has a correlation with gamma interaction position. Based on this correlation, timing estimation algorithm incorporating spatial position for the monolithic PET detector is studied to improve timing resolution.

The whole physical process in monolithic PET detector is analyze carefully and the timing properties of detector signals is modeled. Based on Monte Carlo simulation, the correlation between gamma interaction position and timing information is fully validated. Based on a 21×21×21 mm3 LYSO crystal and SiPM (SensL MicroFJ-30035), two different monolithic detectors (Single Ended and Dual Ended) are proposed to be evaluated. Based on Maximum Likelihood Estimation (MLE) and Artificial Neural Network (ANN), two different timing estimation algorithm were established.

The simulation results shows that the timing resolution of conventional, MLE and ANN timing estimation algorithm are 148.0ps, 129.4ps, 112.1ps in Single Ended detector and 121.1ps, 112.7ps, 97.3ps in Dual Ended detector. The timing resolution of Dual Ended detector is better than Single Ended detector with any timing estimation algorithms. Compared with conventional timing estimation algorithm, two timing estimation algorithms incorporating spatial position can achieve better timing resolution.

In conclusion, incorporating spatial position of gamma interaction into timing estimation algorithm can obviously improve timing resolution. The ANN has the best timing performance in both Single Ended detector and Dual Ended detector.

Keywords: PET, Monolithic detector, Maximum Likelihood Estimation, Artificial Neural Network, Timing estimation
Poster panel
(face) ID: 40

Poster Number:

Time over Threshold Data Acquisition System for PET (#3711)

V. Sánchez-Tembleque1, L. M. Fraile1, J. M. Udías1

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


Time over Threshold (ToT), paired with the increased number of ps-resolution timing IC chips becoming available, hold the potential to replace conventional DAQ electronics in nuclear radiation detection with cheaper, less power-hungry and simpler to design solutions. It remained to be shown whether relatively simple ToT implementations can provide energy measurements with resolution comparable to the state of the art ADCs. In this work we tune a ToT algorithm in two different scenarios: a LYSO-GSO array with 7x7 crystals of 1.55x1.55x7 and 1.55x1.55x8 mm3, in a block detector configuration coupled to a position sensitive PMT (PSPMT), similar to the one employed in the Argus and SuperArgus pre-clinical PET scanners, and to a monolithic ultra fast LaBr3(Ce)  inorganic scintillator, with a truncated cone shape (1"x1.5"x1"), coupled to ultra fast photomultipliers (PMTs), a kind of detector used in state-of-art Nuclear Physics experiments. We digitize energy (one channel) and position (four-channel) signals at 12 bits and 5 GS/s and then process them in silico with a RC-CR shaping stage (of about 100 ns constant-time) followed by a ToT algorithm, with time resolution in the 100 ps range. Energy resolution and flood field images obtained with Anger-like procedures obtained with the ToT algorithm are compared to the results of the conventional ADC based DAQ. We obtain ToT results for both scenarios with comparable performance in terms of energy resolution and fwhm and peak to valley ratios than the conventional setup.

Keywords: Positron Emission Tomography, Time over Threshold, Analog to Digital Converter
Poster panel
(face) ID: 43

Poster Number:

Simulation study of DOI measurement based on light-sharing-window method (#4045)

M. Yang1, S. Xie1, F. Weng2, Z. Zhao2, J. Xu2, Q. Peng1, Q. Huang3

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


The purposes of this study is to investigate the Light-sharing-window (LSW) method for DOI measurement by optic simulation. When a gamma interaction happens in a crystal, most of the scintillating photons will be detected by the photo sensors directly coupled with it. However, some scintillating photons will pass through light-sharing windows, enter the adjacent crystals and be detected by the photo sensors coupled to them. DOI can be derived from the percentages of the photons detected by the photo sensors.

Different light-sharing window designs (single window vs. dual windows; rectangular vs. triangle window) were simulated and compared. The effects of surface finish were assessed. The simulation results show: (1) the relations between DOI and energy differences between photo sensors are monotonic. Thus, they can be used to calculate the DOI; (2) When using single window, the DOI vs. Diff. Energy curves have two segments which are not desired for DOI measurement; (3) Crystals with rough surfaces are more desired since the qualities of their DOI vs. Diff. Energy curves are better. (4) The dual-window method has the best DOI vs. Diff. Energy curves in term of linearity.

We conclude that dual-window LSW method is a very promising method for accurate DOI measurements. The LSW method didn’t catch enough attentions before for two practical reasons. (1) it is very challenging to accurately fabricate the designed light-sharing windows on discrete scintillators. (2) LSW method requires to read out the light output from every single crystals accurately. In our previous studies, we have developed an anti-reflection coating technology which can “print” any designed light-sharing windows on a crystal. We have also developed an advanced read out electronics (called Pico-PET electronics), which is able to read out energy signals from hundreds of individual photo sensors accurately.  Next we will fabricate dual-window LSW detectors to validate this simulation study.

Keywords: light window, DOI measurement
Poster panel
(face) ID: 46

Poster Number:

Design of a Modular SPECT Camera with Improved Spatial Resolution near Edges (#4201)

X. Li1, G. Zubal2, M. A. King3, L. R. Furenlid1

1 University of Arizona, College of Optical Science, Tucson, Arizona, United States of America
2 Yale University, School of Medicine, New Haven, Connecticut, United States of America
3 University of Massachusetts, UMass Medical School, Worcester, Massachusetts, United States of America


We are designing a modular SPECT camera for specialized imaging applications. We proposed to use NaI(Tl) as our scintillation crystal whose shape is hexagonal with ~100 mm side lengths and >5 mm thickness. Our target average spatial resolution is ~1.5 mm (FWHM) which corresponds to a RMSE of 0.64 mm . We propose putting SiPMs on the edges to help improve the spatial resolution. Our preliminary simulation result shows that with 6 mm crystal thickness, 6 mm light guide, and 19 close-packed photomultiplier tubes (1.5’ diameter), the average spatial resolution across the detector plane is 4.38 mm (RMSE), due to the very high bias and variance in regions close to the edges; By attaching 4 SiPMs on each edge, the average spatial resolution is improved to 1.54 mm (RMSE). We are still optimizing the detector design by tuning the parameters such as thickness of crystal, thickness of light guide, positions of SiPMs on the edges and number of PMTs to achieve the target spatial resolution.

This work was partially supported by NIH/NIBIB grant 1R01EB022521.

Keywords: SPECT camera, spatial resolution
Poster panel
(face) ID: 49

Poster Number:

A fiber-based multimodal system with Laparo-PET_NIRF_visible for sentinel lymph node and tumor detection for laparoscopic surgery (#1277)

S. H. Song1, H. G. Kang1, Y. B. Han1, H. - Y. Lee2, K. M. Kim3, D. H. Jeong4, S. G. Lee4, S. J. Hong1, 5

1 Eulji University, Department of Senior Healthcare, Daejeon, Republic of Korea
2 Seoul National University, Department of Nuclear Medicine,College of Medicine, Seoul, Republic of Korea
3 Korea Institute of Radiological & Medical Science (KIRAMS), Division of Medical Radiation Equipment, Seoul, Republic of Korea
4 Seoul National University, Department of Chemistry Education, Seoul, Republic of Korea
5 Eulji Unversity, Department of Radiological Science, Kyeonggi, Republic of Korea


Robot-assisted and laparoscopic surgery using a near –infrared fluorescence (NIRF) tracer has been used for precise detection of sentinel lymph node (SLN) and specific tumors.  However, the limitation of penetration depth is a major problem caused by absorption and scattering in tissue. If this limitation can be overcome by additional modality, the surgery can become quite efficient with a reduced operation time.

We propose a fiber-based multimodal laparoscope coupled with gradient-index (GRIN) lens. The Laparo-PET detector consists of an internal and external module for detection of coincidence events. The internal and external modules are placed inside and outside the body, respectively. Segmentation of crystal map and gain normalization of coincidence events were conducted using a Matlab software. Light transfer of NIRF, visible light, GRIN lens was optimized with a ZEMAX software.

Using this method, we obtained not only a Na-22 point source with 3.2 cps/kBq and 4 mm spatial resolution but also a phantom contained ICG mixture.

Keywords: Laparo-PET, NIRF, robotic-assisted surgery
Poster panel
(face) ID: 52

Poster Number:

A Closed Polyhedron-Geometry Organ-PET for Small Animal Organs and Tissues (#1510)

L. Sensoy1, S. R. Cherry1

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


We designed and simulated a novel, compact, high resolution PET imaging device designed specifically for imaging of excised murine tissue and organs. The prototype scanner is a fully closed, 4π geometry scanner comprised of two monolithic hollow spherical scintillators, which are optically coupled to each other once the excised mouse organ is placed into its cavity. GLuGAG ceramic scintillator was proposed for this study for its high light-yield and easier fabrication for this particular design. By covering more than 95% of the solid angle, the scanner is able to achieve 2-4 times better geometric efficiency in coincidence detection than existing state-of-the-art small animal PET scanners while using significantly less scintillator material. The compact design reduces the deteriorating effects of noncollinearity of annihilation photons to a significant extent. We also aimed to demonstrate that only a single layer of silicon photomultiplier (SiPM) arrays coupled on the outside of this scintillator would be sufficient to extract the polar, azimuthal and radial locations of the interactions within the scintillator, as opposed to other scanner geometries where dual-ended readouts or various forms of multi-layered scintillators are conventionally used to extract depth-of-interaction (DOI) information. The results from GATE/Geant4-based Monte Carlo simulation of the proposed scanner demonstrated that it is possible to achieve accurate positioning of scintillation events within less than 1 mm for a scanner with a radial thickness smaller than 8 mm. The intrinsic coincidence detection sensitivity of the system can be improved by a factor of three if the radial thickness is extended to 20 mm at the expense of a reduction of spatial resolution. An optimization on the scintillator thickness can be achieved after an assessment of the statistical requirements for image reconstruction.

Keywords: small animal PET, high resolution PET, SAI PET
Poster panel
(face) ID: 55

Poster Number:

Development of a circular shape Si-PM-based detector ring for positron emission mammography (PEM) system (#1843)

K. Nakanishi1, S. Yamamoto1, H. Watabe2, K. Kato1

1 Nagoya University, Graduate School of Medicine, Nagoya, Aichi, Japan
2 Tohoku University, Cyclotron and Radioisotope Center, Sendai, Japan


In clinical situations, various breast positron emission mammography (PEM) systems have been used. However, clinical PEM systems have polygonal detector ring because the size of positon sensitive photomultipliers (PSPMTs) are relative large for small detector ring. Polygonal detector ring sometimes causes image artifact, so complicated reconstruction algorithm is need to reduce artifact. Consequently, we developed a circular detector ring for PEM to obtain images without artifact using a simple reconstruction algorithm. We used Lu1.8Gd0.2SiO5 (LGSO) scintillator block which was made of 1.5 x 1.9 x 15 mm pixels that were arranged in an 8 x 24 matrix. As photodetectors, we used silicon photomultiplier (Si-PM) arrays whose channel size was 3 x 3 mm. The detector unit was composed of four scintillator blocks, 16 Si-PM arrays and a light guide. The developed detector unit has angled configuration since the light guide is bending with four angles of 5.625゚. With these configuration, we can arrange 64 scintillator blocks in a detector ring, i.e. a detector ring can be made with nearly circular shape (regular 64-sided polygon) using 16 detector units. The inner diameter of the developed detector ring is 26 cm. This size is similar to the size of brain PET systems, so our PEM detector ring can measure not only breast but also brain. Measured radial, tangential and axial spatial resolution of the detector ring reconstructed by the filtered back-projection (FBP) algorithm were 2.1 mm FWHM, 2.0 mm FWHM and 1.7 mm FWHM at center of field of view (FOV), respectively. The sensitivity was 2.0 % at center of the axial FOV. With the developed detector ring, we could obtain a high resolution image of the breast and the brain phantoms. We conclude that our developed Si-PM-based detector ring is promising for the development of a high resolution PEM system that can also be used for brain PET system.

Keywords: PET, breast, PEM, Si-PM, LGSO
Poster panel
(face) ID: 58

Poster Number:

A Generic, Scalable, and Cost-Effective Detector Front-End block for PET (#2429)

N. Zeraatkar1, S. Sajedi1, M. Taheri1, S. Kaviani1, H. Khanmohammadi1, S. Sarkar1, H. Sabet2, M. R. Ay1, 3

1 Tehran University of Medical Sciences, Cellular & Molecular Imaging Research Center, Tehran, Iran (Islamic Republic of)
2 Massachusetts General Hospital & Harvard Medical School, Department of Radiology, Boston, Massachusetts, United States of America
3 Tehran University of Medical Sciences, Department of Medical Physics & Biomedical Engineering, Tehran, Iran (Islamic Republic of)


With the advancement of Silicon Photomultiplier (SiPM) technology, these devices are becoming the mainstream for use in PET detectors especially where compatibility with magnetic field is critical. SiPMs are now available in different pixel size and array size, however compared with the PMTs there are now more SiPM signals per unit area which leads to larger number of readout electronic channels and more complexity. In this regard, we are developing a generic PET detector block that is scalable and features reduced readout channel a modular design with applications in small animal, brain and clinical PET imaging. The block detector can readily accept a 12x12 array of SensL SiPMs with both standard and fast output signals with an overall detector dimension of 50x50 mm2. The block is divided into components: 1) the detector head that includes scintillator array coupled with SiPM array, as well as channel reduction and multiplexing board, 2) Digital Front-End (DFE) board that connects to the detector head with FCC cable. The DFE board calculates position, energy, and time of photon incidence and passes the information to a coincidence detection unit. On the detector head PCB board, Scrambled Crosswire Readout (SCR) method was implemented to reduce 144 fast outputs to 9 tile signals and 144 standard outputs to 16 energy channels. The tile signals are also used to generate time pickoff information for timing resolution. We implemented time-to-digital converter (TDC) in Xilinx’s SPARTAN6 field programmable gate array (FPGA) by using a 64-tap delay line. In this work, we used a 24×24 LYSO:Ce array with 2×2×10 mm3 pixels for our ongoing small animal PET project. In the 2D position histograms, all the pixels are clearly resolved using a simple centroid positioning algorithm. The measured energy resolution of the detector block after calibration for all crystal pixels is ~17.5% FWHM. The time resolution was measured at 1.2 ns.

Keywords: Positron Emission Tomography, SiPM, Front-End Electronics, Coincidence Detection
Poster panel
(face) ID: 61

Poster Number:

Evaluation of materials for shared-volume PET/MRI inserts (#2531)

L. Yin1, N. Groß-Weege1, D. Schug1, V. Schulz1, 2

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


We evaluated materials for a shared-volume PET/MRI insert. The aim of this shared-volume insert is to increase the sensitivity of both systems by passing the scintillators through the RF screen into the volume between RF screen and RF resonator. This repositioning allows for an increase of the scintillator height and the screen diameter and can be used to create inserts for smaller MRI bores. The setup defines requirements for the materials which are used to close the apertures created by the new integration structure: The main demands for the material are to be RF-shielding, non-magnetic and optical transparent. Here, we will assess the first two features plus the feasibility of the scintillators’ integration in the MRI screen volume. We use a Philips Achieva 3.0 T MRI to measure the created field inhomogeneity and an Agilent E5071C Network Analyzer for the determination of shielding effectiveness. The influence of the material on the B0 field is evaluated by identifying the depth at which the interference decays to 0.6 ppm. Utilizing the layer thickness and gap, the depth of the interference can be estimated and compared. The shielding effectiveness is obtained using a reference measurement of the experimental setup and they are compared to previously made assumptions. Here, the shielding effectiveness for both 3 T and 7 T is shown, as the targeted hybrid system will operate at 7 T. We observe, that the scintillators do not have RF-attenuation properties and that the field disturbance is low, increasing with the crystal height. No differentiation between LSO and LYSO could be made subject to the B0 field homogeneity. For the conductive materials, a maximal interference depth of 21 mm was measured. Surprisingly, the nickel composites proved compatible with the magnetic field and showed the best balance between shielding effectiveness and interference depth (2.2 mm for the laminated composite mesh HS 9000 with 34.9 dB).

Keywords: PET/MRI, shared-volume, shielding effectiveness, B0 homogeneity
Poster panel
(face) ID: 64

Poster Number:

Development of an optical lens based alpha-particle imaging system using position sensitive photomultiplier tube (#2685)

K. Ando1, M. Oka1, S. Yamamoto1

1 Nagoya University, Graduate School of Medicine, Nagoya, Japan


We developed an optical lens based alpha-particle imaging system using position sensitive photomultiplier tube (PSPMT). The alpha-particle imaging system consists of an optical lens, an extension tube and a 1 inch square high quantum efficiency (HQE) type PSPMT. After a ZnS(Ag) is attached to subject, the scintillation image of ZnS(Ag) is focused on the photocathode of the PSPMT by the use of the optical lens. With this configuration we could image the alpha particle distribution with energy information without contacting to the subject. The spatial resolution and energy resolution were ~0.8 mm FWHM and 50% FWHM at 5 mm from the optical lens, respectively. We could successfully image the alpha particle distribution in uranium ore. The developed alpha-particle imaging system will be a new tool for imaging alpha emitters with energy information without contacting the subject.

Keywords: alpha particle imaging, optical lens, PSPMT, without contacting
Poster panel
(face) ID: 67

Poster Number:

Experimental Results of Variable Pinhole Collimator for SPECT Imaging (#2854)

H. Cha1, Y. - J. Jung1, S. Bae2, K. Lee1, H. Lee3

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


A variable pinhole (VP) collimator is a novel collimator that can reconfigure its performance parameters. The acceptance angle and pinhole diameter can be optimized to the region of interest (ROI) for each rotation angle by overlapping several sheets of thin collimator layers. In this study, we drove each component of the VP single-photon emission computed tomography (SPECT) system and evaluated the initial performance of the system. We also designed the driving part of the VP collimator, gamma-ray detector, and SPECT head. A driving test was performed for each component and the integrated system. Planar images of the integrated VP SPECT system were obtained for the initial performance test. The SPECT head was consisted of a VP collimator, gamma-ray detector, and its driving modules. The dimension of the detector head was 28 × 28 × 30 cm3. The detector was composed of a 45.7 × 45.7 mm2, discrete CsI(Tl) scintillator array, four 52 × 52 mm2 tiled multi-pixel photon counter arrays, and front-end electronics. A field-programmable gate array-based data acquisition system was also used. The initial results were obtained with planar images to evaluate the performance of the VP collimator. However, the operation of the VP collimator, detector, and SPECT head can also be applied to tomographic imaging to improve the performance of the SPECT system. For that, the automatically driving VP SPECT system will be built and the performance of the system will be evaluated. In addition, we will conduct a quantitative evaluation of the system through the phantom experiments.

Keywords: single-photon emission computed tomography, Variable pinhole collimator
Poster panel
(face) ID: 70

Poster Number:

Impact of the ring resolution on the performance of the dual ring high resolution silicon PET (#2857)

A. Studen5, 1, V. Cindro1, N. H. Clinthorne3, H. Kagan4, E. Kalšek4, 1, C. Lacasta2, G. Llosa2, M. Mikuž5, J. F. Oliver1, 2, D. Žontar1

1 Jožef Stefan Institute, Ljubljana, Slovenia
2 IFIC/CSIC-UVEG, Valencia, Spain
3 University of Michigan, Ann Arbor, Michigan, United States of America
4 Ohio State University, Columbus, Ohio, United States of America
5 University of Ljubljana, Ljubljana, Slovenia


Purpose: The dual ring high resolution silicon PET prototype showed that high resolution detectors within the PET ring improve signal-to-noise ratio squared per detected event (SNR2E) by a factor of ten. The achievement was obtained at the expense of the poor resolution of the external ring BGO block detectors. A study was performed to evaluate SNR2E in a system with high resolution external ring detectors.

Methods: A new ring module was built from a LYSO scintilator measuring 5 x 5 x 1.5 cm2 segmented to 1.32 x 1.32 mm2 crystals combined with a Hamamatsu H9500 16 x 16 PSPMT detector. Two such modules replaced a pair of BGO blocks (5 x 5 x 3 cm3 size, 6 x 13 mm2 segmentation) on the opposite side of the 100~cm diameter external ring. Combined with silicon detectors in the internal 20~cm diameter ring, the system was used to study noise properties of extended F-18 FDG sources. A preliminary LYSO module prototype was built from the same scinitilator, a 15 mm thick light-guide and a set of four R5900-00-M16 PMTs.

Results: The preliminary LYSO module showed a spatial resolution of 4 mm FWHM. Combined PET setup allowed for separation of point sources 6~mm apart not separable by BGO modules. Final LYSO modules showed a clear crystal map where all but the crystals at the rim could be separated, giving resolution identical to crystal separation of 1.5 mm FWHM. Combined external ring PET with LYSO modules showed resolution of 3.2 mm FWHM for LYSO module data, and 6.3 mm FWHM for BGO module data. Resolution of events detected in the silicon detectors remained 1 mm FWHM. Imaging a uniform disk of 50~mm diameter the SNR2, internal ring data SNR2 was a factor of two better compared to external ring LYSO data.

Conclusions: There is an advantage in using proximity focusing high resolution detectors within the PET ring independent of the resolutions of the external ring.

Keywords: PET, high-resolution, LYSO, silicon detectors, PSPMT
Poster panel
(face) ID: 73

Poster Number:

Transformable Small-Animal PET Module Design Utilizing Mixed Resolution Detectors (#3187)

N. Eclov1, H. T. Chen1, H. Kim1, C. - T. Chen1, C. - M. Kao1

1 University of Chicago, Department of Radiology, Chicago, Illinois, United States of America


We are developing a transformable small-animal PET system utilizing mixed resolution detectors for improved spatial and contrast resolution and noise performance compared to systems with just clinical resolution or just high resolution detectors. Initial reconstruction studies of single-slice mixed resolution data were performed on a digital high resolution phantom with 1 mm pixels and hot regions a minimum of 4 mm in diameter. Reconstructions were performed using MLEM with a resolution model, with 2 mm and 6 mm high and clinical resolution detector elements, respectively. All reconstructions used 500k total counts, with mixed resolution cases using different percentages of high and clinical resolution data (50%-50%, 40%-60%, and 30%-70% high and clinical resolution, respectively). Though contrast was reduced compared to the high resolution data alone, the mixed resolution cases studied still resolved all of the 4 mm regions, while the clinical resolution data alone did not. The 50%-50%, 40%-60%, and 30%-70% data cases also showed reduced noise compared to the high resolution data alone, and approached the lower noise level of the clinical resolution data alone as a larger percentage of clinical resolution data was used in the reconstruction. Contrast-to-noise ratio (CNR) for the mixed resolution cases exceeded CNR for the high and clinical resolution. A transformable gantry concept was also developed to enable a mixed resolution detector system with adjustable detector positioning and field-of-view. Such a gantry has been built and awaits the installation of detector modules and further testing. These studies demonstrate the feasibility of a mixed resolution PET system and will inform further studies investigating the ideal design parameters of such a system as well as the capabilities of such a system enabled by a transformable gantry.

Keywords: Small-Animal PET, Positron Emission Tomography, Transformable PET, Mixed Resolution Detectors
Poster panel
(face) ID: 76

Poster Number:

A Very Compact Preclinical PET Scanner and Initial Results (#3451)

T. Xu1, 2, Q. Wei3, G. Gong1, 2, H. Liu1, 2, S. Wang1, 2, Y. Liu1, 2, T. Ma1, 2

1 Tsinghua University, Department of Engineering Physics, Beijing, China
2 Ministry of Education, Key Laboratory of Particle & Radiation Imaging (Tsinghua University), Beijing, China
3 University of Science and Technology Beijing, Department of Automatic and Electrical Engineering, Beijing, China


A very compact preclinical PET scanner based on DOI PET detector module has been developed. Eight singles detection units (SDU), composed of 2×3 detector modules and front-end electronics, form a PET ring which has a 175 mm bore size with a 70 mm transaxial field of view (FOV) and a 101 mm axial FOV. The whole scanner consists of only the PET ring and an open source white rabbit switch, which plays the role of clock synchronization and data acquisition system. This compact design makes it not only capable for dedicated PET imaging but also flexible to integrate with other imaging modalities, such as SPECT, CT and MRI. Initial performance evaluation of this PET scanner was performed. Crystals of the staggered 15×15+16×16 DOI detector block can be clearly identified. An average system energy resolution of 14.3% and 1.4 ns system timing resolution have been achieved. For image resolution study, a home made hot rod phantom was scanned. An OS-EM reconstruction algorithm was performed and 1.2 mm hot rods can be resolved. Parameters of this PET scanner are under optimization. In future, full performance of this PET system will be evaluated based on National Electrical Manufacturers Association (NEMA) NU-4, including spatial resolution, sensitivity, scatter and count-rate performance and image quality.

Keywords: preclinical imaging, PET, DOI, SiPM
Poster panel
(face) ID: 79

Poster Number:

Initial test and results of an AwakeAnimalPET system (#3541)

S. Lee1, A. Weisenberger1

1 Thomas Jefferson National Accelerator Facility, Newport News, Virginia, United States of America


We have developed AwakeSPECT, a small animal SPECT/CT system for imaging the mouse brain using an optical motion tracking system. The system successfully has been used in mouse brain studies without the need of anesthesia. As an extension of this technology we have built a prototype AwakeAnimalPET system by using the motion tracking system and modular PET detectors. The system consists of a commercial motion tracking system that utilizes near infra-red (NIR) strobes retro-reflective markers, and a modular PET system with 8 detector modules. The motion tracking system measures six degrees of freedom (x, y, z, roll, pitch, and yaw) of the moving target at 60 frames per second. The PET system uses a 32 channel flash ADC system developed at Thomas Jefferson National Accelerator Facility. Each PET detector module is built with a Hamamatsu H8500 position sensitive photomultiplier tubes (PSPMT) coupled to a LYSO scintillator array. Geometrical calibration was performed using Na-22 point source phantom using the same method used for AwakeSPECT. Initial proof-of-principle tests were performed using a foam mouse phantom with three IR reflectors and an embedded Na-22 point source. During the PET scan, the foam mouse was moved and rotated by hand motion. Image reconstruction with motion compensation resulted in good point image reconstruction.

Keywords: awake animal PET, motion tracking, motion correction
Poster panel
(face) ID: 82

Poster Number:

Comparison of Planar, PET and Well-counter Measurements of Total Tumor Radioactivity in a Mouse Xenograft Model (#3921)

M. V. Green1, 2, J. Seidel1, 2, M. R. Williams1, 2, K. J. Wong1, A. T. Ton1, F. Basuli3, P. L. Choyke1, E. M. Jagoda1

1 Molecular Imaging Program, Center for Cancer Research, NCI, Bethesda, Maryland, United States of America
2 Clinical Monitoring Research Program, Leidos Biomedical Research, Inc., Frederick, Maryland, United States of America
3 Imaging Probe Development Center, NHLBI, NIH, Bethesda, Maryland, United States of America


Introduction: Quantitative small animal radionuclide imaging studies are often carried out with the intention of estimating the total radioactivity content of various tissues such as the radioactivity content of mouse xenograft tumors exposed to putative diagnostic or therapeutic agents. We show that for at least one specific application, positron projection imaging (PPI) and PET yield comparable estimates of absolute total tumor activity and that both of these estimates are highly correlated with direct well-counting of these same tumors. These findings further suggest that in this particular application, PPI is a far more efficient data acquisition and processing methodology than PET.

Methods: Forty-one athymic mice were implanted with PC3 human prostate cancer cells transfected with PSMA (PC3 PSMA (+)) and one additional animal (for a total of 42) with a control blank vector (PC3 PSMA (-)). All animals were injected with [18F] DCFPyl, a ligand for the PC3 PSMA (+) receptor, and imaged for total tumor radioactivity with PET and PPI. The tumors were then removed, assayed by well counting for total radioactivity and the values between these methods intercompared.

Results: PET, PPI and well-counter estimates of total tumor radioactivity were highly correlated (R2 > 0.98) with regression line slopes near unity (0.95 < slope < 1.02) and intercepts near zero (-0.001 MBq < intercept < 0.004 MBq).

Conclusion: Total mouse xenograft tumor radioactivity can be measured with PET or PPI with an accuracy comparable to well counting if certain experimental and pharmacokinetic conditions are met. In this particular application, the PPI process is significantly more efficient than PET in making these measurements.

Keywords: Quantitative PET, quantitative planar imaging, cancer drug bio-distribution studies, mouse tumor xenograft models
Poster panel
(face) ID: 85

Poster Number:

Pupil-PET: a novel PET camera with adjustable FOV, AFOV and sensitivity (#4059)

Z. Zhao2, S. Xie1, F. Weng2, D. Shi1, F. Guo1, 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


The conventional PET cameras, consisting of rings of gamma detectors which may block the paths of radiation beams, are not ideal for in-beam monitoring. In this paper, we presented a novel PET design with adjustable field-of-view (FOV), axial field-of-view (AFOV) and sensitivity, like the pupil of the human eye. Thus, we named it Pupil-PET. The sensitivity of a Pupil-PET is no fixed and depends on the system configuration.

We designed a 16-module prototypic pupil-PET system which has two planes with Archimedean spiral tracks on both of them. Four step motors are used to adjust the aperture and the AFOV automatically with a precision of 1 mm per step. The system can automatically transform into a two-ring system with a 100 mm aperture and a 72 mm AFOV, a two-ring system with an aperture ranged from 100 mm to 200 mm and an AFOV ranged from 72 mm to 100 mm, or a two-ring system with a 200 mm aperture and a 38 mm AFOV. The sensitivity of the small-aperture two-ring configuration is about 8 times higher than that in a large-aperture one-ring configuration. Sixteen detector modules have been fabricated and tested for the prototypic Pupil-PET system. The detector module consists of a 10 x 10 LYSO crystal array (3 mm x 3 mm x 20 mm), a 10 x 10 SiPM array (3 mm), and a high performance readout board (call Pico-PET electronics) which are able to read out all 100 SiPMs in parallel.

In conclusion, our Pupil-PET has two unique features: (1) It is able to transform to adapt objects with different sizes. The large-aperture one-ring configuration can be adapted to achieve a large FOV for objects large in size, such as big animal and whole-body human. Small-aperture two-ring configuration can be adapted to achieve an 8 times higher sensitivity for objects small in size, such as small animal, human brain or breast. (2) The gaps between the detectors and between the rings in a Pupil-PET system provide omnidirectional paths for radiation beams.

Keywords: PET, particle therapies, In-beam PET
Poster panel
(face) ID: 88

Poster Number:

Mind-Tracker PET: A wearable PET camera for brain imaging (#4130)

J. Xu1, Z. Zhao2, S. Xie1, D. Shi1, Q. Huang2, Q. Peng3

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


PET is a powerful tool in both the neurologic studies and brain-related clinic applications. However, the conventional method of imaging subject in supine is not always desirable for brain imaging.

In this study, we have designed and fabricated a wearable PET camera named Mind-Tracker PET for brain imaging. The aperture and the AFOV of the system are 200 mm and 32 mm, respectively. The Mind-Tracker PET consists of 16 detector modules, 16 readout electronics boards (named Pico-PET electronics), and a 3D printed gantry. The detector module consists of a 10 x 10 LYSO crystal array (single crystal size: 3 mm x 3 mm x 20 mm, array size: 32 mm x 32 mm x 20 mm), and a 10 x 10 SiPM array (SensL, J-series SiPM, 3 mm, array size: 32 mm x 32 mm). The discrete crystals and the SiPMs are one-to-one coupled. All 16 crystal arrays and SiPM arrays are characterized by the Pico-PET electronics. The results show excellent performances in term of gain, dark current, break-down voltage and energy resolution.   

The total weight of the scanner is 3,325 grams, which is lighter than the commonly-used motorcycle helmet. The weight of the battery and the host PC are about 2,000 grams. They are assembled and put in a backpack. A healthy grownup can wear the system for brain imaging for hours. This Mind-Tracker PET system makes it possible to monitor the neurologic activities in the brain when the subject in more natural states, such as sitting and walking. We are currently calibrating the Mind-Tracker PET system and preparing for the phantom study.

Keywords: Brain imaging, wearable, PET
Poster panel
(face) ID: 91

Poster Number:

A depth-encoding PET detector using light sharing and single-ended readout with SiPM (#1396)

Z. Kuang1, Q. Yang1, 2, X. Wang1, X. Fu1, N. Ren1, Y. Zheng1, X. Zhang1, Z. Hu1, Y. Yang1

1 Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China
2 University of Grenoble Alpes 1, Medical Physic-UFR PHITEM, Saint-Martin-d'Hères, France


Detectors with depth encoding capability and good timing resolution are important to develop a high performance whole body and total body PET scanner. In this work, a low cost depth-encoding PET detectors using light sharing between two discrete crystals and single-ended readout with SiPMs is introduced. Unpolished LYSO crystals of 3×3×20 mm3 were read out by Hamamatsu 3×3 mm2 SiPMs with one by one coupling. The ratio of the energy of one SiPM to the total energy of two SiPMs is used to measure depth of interaction (DOI). The components between the crystals which include optical glue, air and ESR reflector are optimized based on the DOI responses measured in singles mode. The performance of three optimized detector was measured. For the best detector, a DOI resolution of 2.49 mm and a timing resolution of 427 ps were obtained for events with E>350 keV. The energy resolution of the detector is 10.9%. The detector showed great promising to build a high performance whole body and total body PET scanner. In the next step, scintillator block with the optimized components in between crystals obtained from this work will be built, SiPM row readout circuit will be designed and the performance of the detector block will be measured.

Keywords: depth-encoding, light sharing, single-ended readout, whole body PET scanner, SiPM
Poster panel
(face) ID: 94

Poster Number:

Evaluation of Sparse-Ring Digital Photon Counting Time-of-Flight PET: How Many SiPM Detector Elements Do We Really Need? (#1509)

J. Zhang1, B. Zhang2, Z. Hu2, M. V. Knopp1

1 The Ohio State University, Radiology, Wright Center of Innovation in Biomedical Imaging, Columbus, Ohio, United States of America
2 Philips Healthcare, Cleveland, Ohio, United States of America


The evolution of PET detectors moving from photomultiplier tube to solid-state SiPM is a technology leap. A major cost component of PET system is the detector element which is the combination of crystals and detectors. The study intends to leverage the improved detection capabilities of SiPM detectors in order to challenge current conventions of detector pattern designs by investigating the minimum requirement necessary to pursue sparse-ring PET configurations for oncologic FDG PET/CT. Using an investigational solid-state SiPM PET system, listmode data of oncology FDG PET studies were utilized for simulation of the detector sparsity. Comparative assessment were performed using A) the default full ring detector signals, the sparse-ring PET simulation using only B) 50%, C) 67% and D) 75% of the detector elements in tangential direction along the rings, as well as E) 50% and F) 60% of detector elements in axial direction. Uniformity and NEMA phantom PET with hot spheres were also used for the simulation. Blinded image review and SUV assessment were performed. Depend on the detector patterns, up to 50% of LYSO-to-SiPM detector elements can potentially be saved or the axial FOV can be extended by up to 50% without necessitating more detectors while achieving comparable PET image quality.

Keywords: solid state, SiPM, digital PET, sparse-ring, digital photon counting
Poster panel
(face) ID: 97

Poster Number:

A Room PET Scanner for Natural Environment Neuroscience Research (#1715)

M. F. Smith1, W. Xi2, J. E. McKisson2, J. McKisson2, S. Lee2, B. Kross2, A. G. Weisenberger2

1 University of Maryland School of Medicine, Department of Diagnostic Radiology & Nuclear Medicine, Baltimore, Maryland, United States of America
2 Thomas Jefferson National Accelerator Facility, Radiation Detector and Imaging Group, Newport News, Virginia, United States of America


A room size PET scanner for neuroscience research is proposed in which subjects can move about freely in a natural environment. This scanner, termed RoomPET, would open new frontiers in the study of brain function for normal subjects and those with neurological diseases, disorders and brain injury. A RoomPET scanner with walls entirely instrumented with static PET detectors would be prohibitively expensive. The key to a feasible RoomPET scanner is PET detectors with a limited vertical dimension that translate up and down to keep the brain in the field of view. Head position would be tracked by an array of video cameras, providing real-time data to guide detector motion. Geometric detector sensitivity and resolution for a square room are modeled using analytic formulae. For a head in the center of the room and centered within the vertical field of view of the detectors, geometric sensitivities range between 4 and 18% for room widths between 2 and 3 m and detectors with vertical dimensions of 15 to 40 cm. Spatial resolution is dominated by annihilation photon non-colinearity due to the large detector separation. This non-colinearity contribution is 4.4 and 6.6 mm full width at half maximum (FWHM) at the center of 2x2 and 3x3 m2 rooms, respectively, for a line of response orthogonal to the opposite walls. With pixelated scintillation crystal elements 4 mm on a side the system resolution for F-18 labeled radiotracers is estimated at 4.5 and 6.7 mm FWHM, respectively, for room widths of 2 and 3 m. Non-colinearity can be modeled in image reconstruction, though its blurring effect will not be fully compensated. A small RoomPET system could have geometric sensitivity of 5% with spatial resolution of 5 mm. Such a system would advance the understanding of brain function in healthy humans and in disease and brain injury conditions.

Keywords: PET, brain, head tracking, neuroscience
Poster panel
(face) ID: 100

Poster Number:

A Statistical Model for Positron Emission Tomography Scintillation Detectors with Double-Sided Readout (#1979)

C. - E. Chang1, C. J. Kenney1

1 SLAC National Accelerator Laboratory, Menlo Park, California, United States of America


We describe a statistical model for scintillation detectors. The studied geometry is a tall, cuboidal scintillator, sandwiched by a top and a bottom sensor. We consider single-sided and double-sided readout under head-on as well as side irradiation. We fit our model to published results for 3mm x 3mm x 20mm LYSO:Ce crystals. Fitted parameters are used to predict the performance of double-sided readout under head-on irradiation, which gives a better coincidence time resolution than single-sided configurations. When paired with low-mass top sensors to limit signal loss, and thin flex cables for a high packing density, double-sided readout under head-on irradiation may be more favorable than other configurations.

Keywords: TOF PET, scintillation detectors, double-sided readout
Poster panel
(face) ID: 103

Poster Number:

Scalable PCB cooling method for SiPM-ASIC PET detector modules (#2136)

B. J. Lee1, 2, C. - M. Chang1, 3, I. Sacco1, C. S. Levin1

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


We are developing an MR-compatible positron emission tomography (PET) detector system comprising silicon photomultipliers (SiPMs), and application specific integrated circuits (ASICs) mounted in the same printed circuit board (PCB). Since the heat generated from the ASICs can affect the performance of temperature-dependent SiPMs, a scalable PCB cooling method for stable and uniform SiPM performance was designed, developed and investigated.

To cool the SiPM surface within the PCB, a thermal layer is added below the SiPM layer, which is then connected to the bottom layer through thermal vias for efficient heat conduction. Then, either a “cooling plate” or “square tubes” is attached to the bottom layers of multiple detector modules. A proof-of-concept thermal experiment was performed to test the feasibility of the PCB cooling method for SiPM-ASIC PET detector modules, then a 3D transient heat transfer simulations was performed to analyze the temperature profile on the SiPM surface of the more realistic PCB. The thermal experiments were performed using a prototype PCB with a 6 W heater strip mimicking the ASICs and 15 °C water flowing through the cooling structure; the plate which has more contact to the PCB resulted in 18.0 °C average temperature and 0.8 °C maximum temperature difference across the 5 cm width SiPM layer, while square tubes showed slightly worse results of 18.7 °C and 1.3 °C, respectively. In realistic thermal simulations, the square tubes with thermal vias showed deviation from the experimental results with 21 °C average temperature and 6.5 °C maximum temperature difference. In addition, the thermal gradient was significantly worse without the thermal vias.

Considering the compact and simple geometry of the square tubes compared to the plate configuration, the former with thermal vias are an attractive choice to cool the SiPM-ASIC PET detector modules; ASICs can further optimize the minor non-uniform SiPM performance using digital calibrations.

Keywords: SiPM, temperature, cooling, thermal regulation, PET, ASIC, PCB
Poster panel
(face) ID: 106

Poster Number:

Learning Incident Gamma Positions for Monolithic Scintillator PET Detectors (#2419)

X. Hong1, S. Xie2, F. Weng1, Z. Zhao1, Y. Zan1, Q. Peng3, J. Xu2, Q. Huang1

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


Artificial neural network (ANN) has been developed as a fast and accurate method to estimate incident gamma photon positions in the monolithic scinitillator PET dectector. As an enhanced ANN, deep neural network (DNN) has gained much attention in recent years, which immensely improves the performance of ANN in many tasks. We trained a hybrid DNN with both convolution neural network (CNN) and multilayer perceptron (MLP) architectures for incident gamma photon position estimation on a monolithic crystal of 60 x 60 x 25 mm coupled with an array of 10 x 10 SiPMs with simulated optical photon maps via a Monte Carlo study. The average FWHM of the estimation error over X - Y plane is around 1.25 mm. Though more efforts are required, the current work shows the potential of DNN in estimating accurate incident gamma photon positions in thick monolithic crystals when high sensitivity is desired.

Keywords: PET, deep learning, monolithic crystal
Poster panel
(face) ID: 109

Poster Number:

Study on a prototype oval body PET insert for a 3T MRI system (#2766)

M. S. H. Akram1, T. Obata2, C. S. Levin3, F. Nishikido1, T. Yamaya1

1 National Institute of Radiological Sciences, Division of Radiation Measurement and Dose Assessment, Chiba shi, Chiba, Japan
2 National Institute of Radiological Sciences, Department of Molecular Imaging and Theranostics, Chiba shi, Chiba, Japan
3 Stanford University, Department of Radiology, Stanford, California, United States of America


We have developed an oval shape RF-penetrable PET insert in a motivation for PET/MR body imaging with an existing 3T clinical MRI system. In case of RF-penetrable PET insert, we use MRI built-in body RF coil (at least) as RF transmitter. For the RF field to penetrate through the small gaps in between the shielded PET modules, we keep them electrically floating with respect to the MRI ground. In this study, the MRI built-in body RF coil was used both as transmitter and receiver. We implemented 12 RF-shielded PET modules on a 440 mm major-axis and 350 mm minor-axis oval frame in which 8 PET modules contained PET detectors and electronics. The system was positioned on the MRI patient bed. Experimental results showed about 6 percentage-point RF field homogeneity reduction, about 50% SNR reduction for spin-echo and gradient-echo sequences compare to MRI-only data for the 150 mm diameter X 120 mm length cylindrical imaging region. Almost two times RF pulse amplitudes were required to achieve these responses. Also, 1.25 times increase in maximum local SAR was seen for the central 130 mm diameter transverse plane. The oval insert showed reasonable MRI performances under all studies. Implementing a separate RF receiver coil inside the PET insert might improve the SNR values. The system successfully cleared a major step towards developing large scale oval PET insert for body imaging in an existing MRI system.

Keywords: oval PET insert, RF-penetrable PET insert, PET/MRI body imaging
Poster panel
(face) ID: 112

Poster Number:

Development of Pulse Restoration and RF Filtering Method for Tandem PET-MRI (#2910)

J. Y. Lee1, Y. Choi1, K. B. Kim1, F. Chen2, J. Yan2, Y. Wang2

1 Sogang University, Electronic Engneering, Molecular Imaging Research & Education (MiRe), Seoul, Republic of Korea
2 Minfound Medical Systems, Zhejiang, China


We are developing whole body PET using charge transmission method (CTM) for Tandem PET-MRI. Since CTM could induce PET signal distortions and RF coil transmitter of 1.5-T MRI could generate RF pulse noise of about 63.8 MHz, PET system suffered from pulse distortions and noise from MRI signal. The purpose of this study was to develop a simple circuit performing pulse restoration and RF filtering to improve the performance of PET. Additionally, the effect of PET detector module to MRI was assessed which is necessary for the development of tandem PET-MRI. The pulse restoration circuit was consisted of C-R shaping circuit and summing amplifier. The C-R shaping circuit was designed to shorten the fall time and the rise time of PET signal. Summing amplifier was designed to restore amplitude and photopeak position. The RF noise was selectively filtered out using R-C low-pass filter circuit. The performance of the PET detector using the circuits developed in this study was evaluated outside and inside MRI using various MR sequences (spin echo, turbo spin echo and gradient echo). MRI performance, such as the uniformity and SNR of MR image, were also measured. The amplitude of PET signal increased from 350 mV to 500 mV. The rise time and the fall time shortened from 80 ns to 50 ns and from 25 ns to 18 ns respectively. The performance of time resolution improved from 5.7 ns to 1.8 ns. The noise contaminated in the PET signal was filtered about -15 dB. The effects of the PET detector module on MRI performance were negligibly small (uniformity: 0.5% decrease and SNR of MR image: 5% decrease). This study demonstrated that the method developed in this study considerably improved the performance of PET without adverse effect on MRI. Future study will be performed to apply the method to develop a commercial Tandem PET-MRI.

Keywords: Positron emission tomography, Tandem PET-MRI, Pulse restoration, RF filter, Interference of MRI, PET performance, Signal shaping and mixing, MR compatibility
Poster panel
(face) ID: 115

Poster Number:

Development of innovative PET module with Depth of Interaction and Timing capabilities (#3205)

G. Stringhini1, 2, M. Pizzichemi1, A. Polesel1, 2, M. Paganoni2, S. Tavernier3, E. Auffray1

1 CERN, Geneva, Switzerland
2 University of Milano Bicocca, Milano, Italy
3 Vrije Universiteit Brussel, Brussel, Belgium


PET scanners have to guarantee high diagnostic performances in term of spatial resolution and sensitivity in order to be able to early detect cancer masses. On the other hand, the overall complexity of the system has to be reasonably low to minimize the maintenance costs. In this research work, we present an innovative PET module with timing and Depth Of Interaction (DOI) capability that provide high spatial resolution in a four to one coupling configuration between the scintillators and the detector using a single side readout. The scintillators are 64 LYSO crystals of the dimension of 1.53x1.53x15 mm3 arranged in a 8x8 matrix. A layer of Enhanced Specular Reflector (ESR) is used as a separation foil and the lateral surfaces of the crystal are optically treated to be unpolished. The detector is a 4x4 SiPMs array from Hamamatsu with an active area of 3x3 mm2. On top of the crystals a light guide of glass (1mm thickness) and an ESR layer are placed to recirculate the scintillators. An experimental setup was built at CERN to test and verify the performances of the module. The results show that we can achieve the separation of the 64 crystals, an energy resolution of 12% FWHM, a DOI resolution of 3mm FWHM and a Coincidence Time Resolution (CTR) less then 300 ps FHWM. Combining the DOI and the timing information it is possible to improve the time performances of the system. Simulation results prove that the FWHM of detection time distribution can be decreased by correction for the DOI. These simulations are going to be validated in the months to come. Preliminary measures with two module in coincidence with a rotating source in the centre of the field of view show a spatial resolution of 1.5 mm of FWHM in all directions.

Keywords: PET, Depth of interaction, Timing, Scintillators Module
Poster panel
(face) ID: 118

Poster Number:

A Novel TOF-PET Detector Based on SiPM and Cherenkov Radiation (#3388)

X. Liu1, 2, T. Xu1, 2, Y. Xia3, 4, T. Ma1, 2, Y. Liu1, 2

1 Tsinghua University, Department of Engineering Physics, Beijing, Beijing, China
2 Key Laboratory of Particle & Radiation Imaging, Ministry of Education (Tsinghua University), Beijing, Beijing, China
3 Science and Technology on Reliability and Environmental Engineering Laboratory, Beijing, Beijing, China
4 Institute of Spacecraft Environment Engineering, Beijing, Beijing, China


Time of flight (TOF) technology has been widely used in positron emission tomography (PET) systems. Cherenkov TOF-PET has become a promising TOF-PET choice due to its excellent timing performance.

In this work, we designed a Cherenkov TOF-PET detector based on PbF2 crystal and SensL dual-output SiPM. An experimental platform was set up to evaluate the timing performance of the detector. The acquired experimental data were pre-processed by amplitude correction and noise correction and then used to evaluate detector timing performance. The optimum coincidence resolving time (CRT) was obtained using lead edge discrimination (LED).

Factors that could influence the CRT, including ambient temperature, SiPM over voltage, crystal size and crystal surface treatment are investigated. The best CRT of 320 ± 7 ps was achieved by using black painted PbF2 crystals at -25 ℃ and 6 V over voltage.

We expect that a better timing property of Cherenkov TOF-PET could be achieved by using SiPM with better single photon timing resolution (SPTR) and higher sensitivity near UV region.

Keywords: Cherenkov radiation, TOF-PET, SiPM, lead flouride
Poster panel
(face) ID: 121

Poster Number:

Comparison between one-to-one and a multiplexed readout scheme with SiPM for ToF PET (#3708)

C. L. Kim1, D. McDaniel1

1 GE Healthcare, Waukesha, Wisconsin, United States of America


Silicon Photomultipliers are known for its high PDE and excellent timing resolution for TOF PET. While it required a multiplexed readout scheme or an ASIC due to its small size, it made possible to have one-to-one coupling between a crystal and a SiPM, which was not practical with PMTs. Several studies have demonstrated under a 200ps coincidence resolving time in FWHM with a single 20mm long LYSO (crystal size used in a whole-body PET scanner) coupled to a SiPM and a single channel fast electronics. An ideal electronics/ASIC would be such that it shows the same or similar timing performance with a block detector consisted of an array of crystals and SiPMs. As a first step to understand what it takes to achieve this, a 2x3 LYSO crystal array was optically coupled to a 2x3 array of SiPMs. The crystal dimensions were 3.95 x 5.3 x 25mm3, and one layer of ESR film was inserted in between crystals to have one-to-one coupling. First, light output and coincidence resolving time of the 2x3 array was measured with a multiplexed timing scheme. Second, each one-to-one detector was individually measured by disabling the rest of channels. For the former, inter-crystal Compton scatter events were separately analyzed for its timing resolution. On the contrary to the expectation, one-to-one scheme showed 25% less light collection and slightly worse timing resolution of 265ps compared to 259ps in the multiplexed scheme, even though the latter scheme had six times more dark currents. Inter-crystal Compton scatter events showed only a slightly worse timing resolution of 257ps compared to the 252ps of the map peak events. In the final paper, we will present the result based on six 2x3 crystal arrays. Furthermore, the effect of inter-crystal Compton scatter will be presented as a function of gamma ray incident angle to the crystal array.    

Keywords: PET SiPM ASIC multiplexing
Poster panel
(face) ID: 124

Poster Number:

Timing and Readout System for Digital Positron Emission Tomography based on WR devices (#3951)

H. Li1, 2, G. Gong1, 2, S. Huang3, X. Liu1, 3

1 Ministry of Education, Key Laboratory of Particle & Radiation Imaging, beijing, China
2 TSINGHUA, Department of Engineering Physics, beijing, China
3 Beijing NOVEL MEDICAL Equipment Ltd., beijing, China


White Rabbit is a prompt new technology provides sub-nanosecond accuracy and picoseconds precision of synchronization as well as Ethernet package transmission. In this paper, we present our work of adapting WR technology for a pre-clinical digitized Positron Emission Tomographic scanner. The PET scanner ring consists of 8 detector modules, each containing 6 LYSO arrays coupled to an 8x8 SiPM matrix. The signals from the SiPM are processed by a dedicated 64-Channel Readout ASIC, EXYT, and then digitized and time-tagged with a commercial ADC and FPGA based TDC respectively. Each detector modules contains a WR slave node to synchronize the time reference of ADC and TDC. Time-tag measured data are transmitted via the WR link and collected by a standard WR switch, which is modified to carry firmware based timestamp coincidence logic to reduce the data throughput. The compressed data are then repacked and sent out to a reconstruction PC. The new timing and readout scheme greatly simplifies the PET design with open hardware and brings a fully modulized structure to reality.

Keywords: Timing, PET, Synchronization
Poster panel
(face) ID: 127

Poster Number:

Sparse digital Silicon Photomultiplier (dSiPM) readout of a scintillator array (#4026)

W. C. J. Hunter1, R. S. Miyaoka1, D. Sowards-Emmerd2, C. - H. Tung2, A. E. Perkins2, M. Rielly2

1 University of Washington, Radiology, Seattle, Washington, United States of America
2 Philips Healthcare, Advanced Molecular Imaging, Cleveland, Ohio, United States of America


We examine methods of reducing the number of digital Silicon Photomultipliers (dSiPMs) used to readout a discrete scintillator array and the impact it has on gamma-ray-imaging performance. A reduced number of channels can lower sensor cost and possibly decrease readout complexity and latency. Current state-of-the-art is a one-to-one coupling of crystals and SiPM pixels. We aim to decrease the sensor area (and number of sensors) by a factor of 2 to 4 while reading out the same crystal array.

In this study, we implement sparse sensor patterns using custom inhibit maps in a dSiPM array from Philips Healthcare (PDPC TEK). A tailored reflector on the readout face is also used.  Several sparse sensor patterns were considered. Algorithms for event processing and resulting performance tradeoffs are reported for 2 sparse sensor patterns (36 and 16 pixels) and compared with the performance of a fully-populated sensor array (64 pixels). Although every one of the sixty-four 4-by-4 mm^2 crystals are resolved, trade-offs in timing and energy resolution are observed.

Keywords: digital silicon photomultiplier, sparse-sensor readout, PET, scintillator array
Poster panel
(face) ID: 130

Poster Number:

A Novel Designed Small Angle CT System based on Overlay Rotation (#2437)

B. Wang1, 2, Y. Xiao1, 2, D. Yu1, 2, Z. Chen1, 2

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


A small angle CT system is limited to scan a big object, which is larger than the field of view. Half-Covered Helical CT has been proposed to solve the problem, but additional scan process is also needed. This paper describes a novel designed turntable to make use of the small angle system for large object testing, whose scan process is as same as the conventional CT system. In this newly designed overlay rotation CT scanning system, two turntables are used to make up a two-degree-of-freedom system. These two turntables rotate at constant rotation rates and projection data is collected at the same time. In this paper, new projection data rebinning method is adopted and the reconstructed algorithm is described. Numerical experiments are taken to test a simple prototype and verify the feasibility of this CT system. In conclusion, this novel designed overlay rotation CT system is effective to get enough projection data for reconstruction.

Keywords: overlay rotation, small angle CT system, two-degree-of-freedom
Poster panel
(face) ID: 133

Poster Number:

Comparison of image and data domain methods for three-material decomposition in dual-energy CT (#3814)

T. Humphries1, R. McGarity1, K. Uy1

1 University of Washington Bothell, School of STEM, Bothell, Washington, United States of America


Conventional material decomposition in dual-energy CT (DECT) involves recovering tIn this work we propose a novel three-material decomposition method in the data domain which is based on a conservation of volume constraint. We compare the approach with an image-based decomposition method using a recently proposed polyenergetic reconstruction algorithm to generate the reconstructed images. Our preliminary results indicate that the image-based decomposition using polyenergetic reconstruction provides better image quality than the data-domain approach.

Keywords: dual energy ct, material decomposition
Poster panel
(face) ID: 136

Poster Number:

More Accurate and Less Noisy Spectral Deconvolution Strategy using Photon Counting Detectors (#1673)

S. Wang1, 2, L. Zhang1, 2, X. Xu1, 2

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


Tomographic reconstruction based on photon counting detectors is vulnerable to the spectral distortion induced by the non-ideal detector response. To improve the quantitative property of the reconstructed images, correction or compensation for the response should be carried out. A novel dual-update spectral deconvolution strategy based on the EM algorithm is proposed in this article. The proposed method will update the deconvolution spectral counts as well as the detector response matrix one by one to eliminate the possible bias induced by the approximation of detector response matrix. Numerical simulation and experimental verification illustrated that the proposed method could give us more accurate and less noisy deconvolution results.

Keywords: photon-counting detector, computed tomography, spectral deconvolution
Poster panel
(face) ID: 139

Poster Number:

Response Function Estimation for the Photon Counting Detector using Multiple Balanced K-Edge Filters (#2274)

X. Xu1, 2, L. Zhang1, 2, S. Wang1, 2

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


The introduction of photon counting detectors into the computed tomography (CT) has a lot of benefits. It is vital to know the energy response functions for energy spectrum correction and system optimization. The goal of this paper is to estimate the energy response functions using conventional X-ray source and filters. In conventional dual-energy CT, the filters is the step wedges made of various materials. Recently, spectral CT using multiple balanced K-edge filters is proposed. The transmission matrix of different filters is analyzed by singular value decomposition (SVD). To get a stable solution, the rank of the transmission matrix should larger than the rank of the response matrix, which is equal to the energy bins. The simulation results and the experiment results showed that the rank of transmission matrix with multiple balanced K-edge filters is larger the rank of transmission matrix with step wedges filters. Using multiple balanced k-edge filters, we can estimate the response matrix of photon counting detectors easily and flexibly.

Keywords: response function, photon-counting detector, k-edge filters, step wedge filters
Poster panel
(face) ID: 142

Poster Number:

Proposal of The Fluoroscopes Using Gamma Rays Generated from Electron Positron Pair Annihilation with Low Exposed Dose (#3965)

H. Matsunaga1, Y. Emoto2, K. Fujiwara2, H. Ito2, H. Kawai2, S. Kimura2, A. Kobayashi2, T. Mizuno2, T. Nakamura2, T. Tanaka1, T. Yuzawa1

1 Chiba University, faculty of science, Chiba, Tokyo, Japan
2 Chiba University, Graduate School of Science, Chiba, Tokyo, Japan


X-ray Computed Tomography (CT) is the important equipment in medical care today. However, its exposed dose is very high. We propose two fluoroscopes using gamma ray generated from electron positron pair annihilation. Compared to X-ray CT, they have good ability to see through the organism. In addition, the exposed dose of them is less than 1/10 or 1/100 of that of X-ray CT. Positrons generated from 68Ge are irradiated to the inorganic scintillator. Annihilation points are measured using sheets of wavelength-shifting fiber in both fluoroscopes. The position resolution of annihilation is 0.2 mm. When a positron is annihilated, two gamma rays with energy of 511 keV are generated back to back. One gamma ray is irradiated opposite to a living body. By measuring a point where the gamma ray is irradiated, we can identify the expected point where gamma ray is irradiated to the living body. The position resolution of irradiation is 0.2mm. But absorbed or scattered gamma ray can't be measured. As a feature of one fluoroscope, it is possible to measure  transmittance of gamma ray, by measuring existence or absence of gamma ray at the expected arriving point. Therefore, most part of scattering events in the organism can be removed. As a result , the exposed dose of the fluoroscope is less than 1/10 of that of X-ray CT. As a feature of the other fluoroscope, it is possible to measure the point where a gamma ray is scattered by measuring the energy and the arriving point of the gamma ray. Therefore, most of the noises due to Compton scattering is removed. As a result, the exposed dose of the fluoroscope is less than 1/100 of that of X-ray CT.

Keywords: gamma ray, gamma ray CT
Poster panel
(face) ID: 145

Poster Number:

Evaluation of a new fast elastic motion compensation approach in PET/CT: comparison to optical flow based motion correction (#1330)

S. Pösse1, D. Mannweiler1, F. Büther1, 2, I. Hong3, J. Jones3, M. Schäfers1, 2, K. P. Schäfers1

1 University of Münster, European Institute for Molecular Imaging, Münster, North Rhine-Westphalia, Germany
2 University Hospital of Münster, Department of Nuclear Medicine, Münster, North Rhine-Westphalia, Germany
3 Siemens Medical Solutions USA, Inc., Molecular Imaging, Knoxville, Tennessee, United States of America


Physiological respiratory motion during PET data acquisition degrades PET image quality and accuracy. This leads to image blurring and motion artifacts which may influence clinical decision making. Therefore, motion-compensated image reconstruction has been proposed to overcome this problem. In this study we evaluated a new ultrafast elastic motion compensation approach (eMoco) on patient data and compared the results to an optimized motion correction image reconstruction method (MCIR). The eMoco method uses a deblurring kernel during the iterative image reconstruction which estimates the motion between the optimal gate and the rest of the data. For the MCIR method 10 respiratory gates were generated out of listmode PET data and motion vector fields were estimated using optical flow. To improve optical flow based vector fields, an optimization strategy was performed based on varying four internal parameters: alpha (regularization), tau (step size), number of iterations, and number of multi-levels. An optimal parameter set was found and used for all data evaluations. F-18-FDG PET/CT data of 6 patients were acquired on a Siemens Biograph mCT. Changes in standardized uptake values (SUVmax, SUVmean, SUVpeak) and their metabolic volumes were calculated for individual lung lesions. As a result, SUVmax increased with both motion correction approaches (12.6±9.40% for eMoco, 8.83±6.40% for MCIR) showing the clear impact of motion correction on image quantitation. Similar findings could be observed for SUVmean (11.9±9.90% - eMoco; 8.92±6.03% - MCIR) and SUVpeak (13.5±10.0% - eMoco; 6.78±7.28% - MCIR). As expected, the metabolic volumes decreased significantly (25.1±20.8% - eMoco; 13.2±10.6% - MCIR). As a conclusion, the proposed fast eMoco technique performed similar to the "gold standard" MCIR technique making this approach a valuable alternative for clinical PET motion correction strategies.

Keywords: PET/CT, motioncorrection, opticalflow
Poster panel
(face) ID: 148

Poster Number:

Estimation of Mid-Myocardial Surface in Myocardial Perfusion SPECT Using a Model-based Reconstruction Method (#1609)

Y. Dong1, E. C. Frey1, Y. Du1

1 The Johns hopkins university, Radiology, Baltimore, Maryland, United States of America


Wall motion is an important cardiac functional parameter that provides diagnostic information about cardiovascular disease. It can be evaluated by comparing mid-myocardial surface positions from gated myocardial perfusion SPECT (MPS) images. We have developed a model-based reconstruction method for MPS using activity and left ventricle (LV) geometry models. The activity model divides the myocardium into segments with uniform activity concentration as in standardized segmental analysis. The geometry model parameterizes the endo- and epi-cardial surfaces using rays originating from the LV axis. Along each ray, the two surfaces are defined by the endocardial radius and the wall thickness. The image generated from the two models is fitted to the reconstructed image with resolution and noise modeling of the image formation process. The resulting geometry parameters can be used to compute the mid-myocardium as the radius plus half the wall thickness. The goal of this study was to evaluate the model-based reconstruction method in estimating the mid-myocardial surface comparing to three existing methods: (1) assuming that the mid-myocardial position is the center of mass of the radial profile; (2) assuming that the voxel having the maximum intensity along the radial profile is the mid-myocardium; and (3) fitting a square wave model to the radial profile through the reconstructed image with 1D resolution modeling and using the resulting parameters to compute the mid-myocardial. The methods were applied to realistic, analytically-simulated MPS images. The LV geometry was based on the XCAT phantom. Two noise levels were used with 50 realizations for each. Results showed that the proposed method greatly improved the estimation accuracy of the mid-myocardial surface regarding the three methods. The RMSEs were 0.64 mm and 1.18 mm at the two noise levels for the proposed method, compared to 2.94mm and 5.82mm, 1.93mm and 4.16mm, and 1.24mm and 3.70mm for methods 1-3, respectively.

Keywords: myocardial perfusion, model-based reconstruction, wall motion
Poster panel
(face) ID: 151

Poster Number:

Calibration-free method for beam-hardening compensation: preliminary results (#1986)

C. Martinez1, 2, C. de Molina1, 2, M. Desco1, 2, M. Abella1, 2

1 Universidad Carlos III de Madrid, Departamento de Ingenieria e Ingenieria Aeroespacial, Leganés, Madrid, Spain
2 Instututo de Investigación Sanitaria Gregorio Marañón, Madrid, Madrid, Spain


Due to the polychromatic nature of the spectra in X-Ray CT, mean energy of the x-ray beam increases with the traversed material as lower photons are preferentially absorbed. This effect, known as beam hardening (BH), results in two artifacts on the reconstructed images: cupping on homogeneous regions and streaks on heterogeneous areas.

A simple correction method implemented in most CT scanners is linearization. This method is based on the characterization of the relation between the total attenuation measured at the detector and the thickness traversed, namely BH function, by means of a soft-tissue equivalent phantom. The Ideal function can be obtained as the line tangent to the BH function at zero. The relation between BH function and Ideal function is the linearization function that can be used to correct projection data of subsequent scans. This method assumes the object is homogeneous, generally soft tissue, correcting only the cupping artifact. We recently presented a new method extending this idea to 2D, adding bone to the original phantom to be able to correct both cupping and streaks. However, while soft-tissue equivalent materials proposed in the literature show good properties, it is difficult to find an optimum material with the same attenuation to X-rays as bone, furthermore when bone composition vary along the body and change with the age.

Here we propose method that avoids the calibration step, solving the problem of searching an equivalent bone material, by using the information of the acquired sample to create the real beam hardening function.

Preliminary results using simulations with the Digimouse phantom showed a noticeable reduction of both artifacts. Improvement based on the root mean square error with respect to the monochromatic case was of 98% in bone and 89% in soft tissue.

Keywords: artifact, beam hardening, CT, cupping
Poster panel
(face) ID: 154

Poster Number:

Continuous Bed Motion Acquisition Self Time Alignment Method for Patient-Based Quality Control (#2105)

V. Y. Panin1, M. Aykac1

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


The development of silicon photomultiplier (SiPM) based detectors resulted in significant improvement of time-of-flight (TOF) resolution in PET. Higher standards in validating TOF information are desirable, since TOF reconstruction is sensitive to the misplacement of events based on time measurements. On photomultiplier (PMT) based scanners, time alignment estimation was stable over long periods of time, so frequent time calibration procedures were not necessary. New high performance SiPM PET scanners will require close monitoring in performance, at least over the initial phase of clinical use. This may result in a renewed interest in a self-time alignment (TA) method, primarily as a quality control (QC) procedure.

In this work we use the idea behind the self-TA procedure, where unknown object activity is estimated from non-TOF data. Modeled TOF data are compared against measured data in order to find individual detectors’ time offsets (TOs). TOs are estimated simultaneously by matching the TOF center of mass between modeled and measured TOF data. This method can be considered as a refinement of the currently used Siemens TA procedure. The method is generalized for Continuous Bed motion (CBM) acquisition, which is becoming standard on Siemens scanners. An additional advantage is that the CBM daily Quality Control phantom can be axially shorter, since it is not required to cover all the field of view at once.

The algorithm was tested on a Siemens next generation SiPM PET/CT scanner prototype. An image quality (IQ) phantom and patient data were used for this investigation. The IQ phantom data reconstruction showed that undetected residual TOs can be corrected by use of the self-TA procedure.

Keywords: PET, time alignment
Poster panel
(face) ID: 157

Poster Number:

Quantification of Gadolinium Nanoparticles Concentration with SPECT and Spectral Photon Counting CT (#2395)

O. Kochebina1, A. Halty1, J. Taleb2, D. Kryza2, M. Janier2, C. Mory1, D. Bar-Ness1, P. Douek1, T. Baudier1, S. Rit1, D. Sarrut1

1 CREATIS - INSA Lyon, Villeurbanne, France
2 LAGEP-Université Lyon 1, Villeurbanne, France


Gadolinium nanoparticles (NP) combined with conventional radiotherapy could be used for oncology treatment. Indeed, NPs concentrated in a tumor could enhance its radiosensitization. The noninvasive quantification of NP concentration is a crucial task for radiotherapy treatment planning as this will define the delivered dose. Image based quantification could be achieved, for example, with spectral photon counting CT (SPCCT) for Gd detection or single photon emission CT (SPECT) for detection of NPs coupled with In-[111] tracer. This presentation is focused on Gd-NP quantification on phantoms with nanoSPECT/CT scanner. The results show that activities above 1 MBq could be detected and quantified. The discrepancy between quantification activity measurements and dose calibrator ground values is ~10-15% while the level of image noise is up to 4-6%. Such a large disagreement could be due to several correction factors that should be taken into account: attenuation, scatter, dead time, kinetic of the activity distribution, partial volume effect etc. The analysis of SPCCT images gives ~25% discrepancy between the observed Gd concentrations and the reference value, which are provided by the NP manufacturer. The noise level for SPCCT is much higher than for SPECT. This is partially due to different voxel size, which is almost 14 times smaller for SPCCT. However, we obtain the linear correspondence between results from SPCCT images and activity from SPECT images. The slope, SPCCT vs. SPECT, from log likelihood linear fit is 0.298±0.008.

Keywords: Image quantification, SPECT, SPCCT, Gadolinium nanoparticles
Poster panel
(face) ID: 160

Poster Number:

Attenuation Correction Methods for Dual Gated Myocardial Perfusion SPECT/CT (#2769)

Q. Zhang1, D. Zhang1, G. S. P. Mok1, 2

1 University of Macau, Biomedical Imaging Laboratory (BIG), Department of Electrical and Computer Engineering, Faculty of Science and Technology, Macau, China, Macao Special Administrative Region
2 University of Macau, Faculty of Health Sciences, Macau, China, Macao Special Administrative Region


Dual respiratory-cardiac gating (DG) method was shown to produce motion-freeze myocardial perfusion SPECT images. However, appropriate attenuation correction (AC) for DG SPECT is to be determined. This study aims to evaluate the performance of various attenuation maps for DG cardiac SPECT. We used the 4D Extended Cardiac Torso (XCAT) phantom to simulate a male patient with respiratory cycle of 5 s and z-motion amplitude of 2 cm. The respiratory cycle was divided into 1152 frames which were grouped to 6 respiratory and 8 cardiac phases, i.e., a total of 48 phases. The corresponding attenuation maps were grouped to represent different AC maps: DG CT (DCT), respiratory gated CT (RCT), average CT (ACT), interpolated CT (ICT), HCTs at end-inspiration (HCT-in), end-expiration (HCT-ex) and mid-respiration (HCT-mid) respectively. The ICT was obtained by interpolation based on the motion vector generated between HCT-in and HCT-ex from affine+b-spline image registration. We used an analytical projector to simulate a LEHR collimator with 120 projections over 180˚. Both noise-free and realistic noisy projections were generated and reconstructed by OS-EM method using different AC maps respectively. For each cardiac phase, reconstructed images from different respiratory phases were registered to end-expiration using affine+b-spline method and summed to get a registered cardiac image. Polar plots were generated for 8 registered cardiac images and relative difference (RD) was computed for each segment for the 17-segment analysis. For all cardiac phases, the average RDmax for AC with ACT, HCT-in, HCT-ex, HCT-mid, RCT and ICT comparing to DCT were 4.09%, 9.24%, 5.04%, 4.02%, 3.22% and 2.85% respectively. Since DCT and RCT are clinically challenging due to the high radiation and implementation complexity, ICT is recommended, followed by HCT-mid or HCT-ex for AC in DG cardiac SPECT, with good accuracy and relative low radiation dose.

Keywords: Dual gating, SPECT/CT, Attenuation map, Myocardial perfusion
Poster panel
(face) ID: 163

Poster Number:

SNR Analysis of Polychromatic Fan-Beam XFCT System Using a Photon-counting Detector Array (#2826)

S. Zhang1, L. Li1, Z. Chen1

1 Tsinghua University, Engineering Physics, Beijing, China


X-ray fluorescence computed tomography (XFCT), which is considered as a promising approach to determine the bio-distribution of high Z elements such as gadolinium (Gd) and gold (Au), was first performed using synchrotron radiation. However, due to the limit access of monochromatic synchrotron source, XFCT experiments have been carried out more on benchtop XFCT systems with polychromatic x-rays produced by x-ray tubes in recent years. The XRF signal stimulated by polychromatic X-ray source is strongly affected by Compton-scattered photons. In order to reduce the dose of scattered photons with energies near the energy of fluorescent photons, filters consisting of lead (Pb), copper (Cu), aluminum (Al), tungsten (W) or tin (Sn) were used in previous studies. However, using filters will reduce the intensity of incident beam, which causes larger statistical noises. In addition, the energy response of the detector will weaken the effect of filtering, which also causes the decrease of signal quality. In this paper, we investigated the signal-to-noise ratio (SNR) of projection data detected by a CZT photon counting detector array in a benchtop fan-beam XFCT system. The phantom is stimulated by a polychromatic X-ray source filtered by different filters consisting of Pb, W, Sn, Cu, and Al. Simulations show that due to the statistical noises and the energy response of CZT detector, optimizing incident X-ray spectrum using common filters doesn’t help improving signal-to-noise ratio of projection data at a fixed power of x-ray tube.

Keywords: X-ray fluorescence, XFCT, SNR, filter, photon-counting detector
Poster panel
(face) ID: 166

Poster Number:

Data driven time alignment for TOF-PET (#3091)

A. Rezaei1, G. Schramm1, J. Nuyts1

1 KU Leuven, Nuclear Medicine, Leuven, Belgium


The addition of time of flight (TOF) information in positron emission tomography (PET) reconstructions enables a more robust estimation of the tracer distribution uptake. Furthermore, it enables the estimation of the attenuation image from the PET data, if the spatial distribution of the activity is wider than the TOF resolution of the scanner. The use of TOF-PET data requires accurate alignment of the timing information, which is typically obtained by an additional acquisition of a known source of activity, from background LSO radiation or from the clinical data. In this work we propose a method to align the TOF data using the TOF-PET emission data, and we analyze the effects of an inaccurate TOF alignment on joint activity and attenuation reconstructions in a uniform phantom scan as well as a patient brain scan.

Keywords: time-of-flight, time alignment, PET
Poster panel
(face) ID: 169

Poster Number:

Effects of PET Attenuation Mismatches at Different TOF-PET Timing Resolution Levels (#3184)

W. Qi1, L. Yang1, C. Chan1, E. Asma1

1 Toshiba Medical Research Institute, USA., PET reconstruction, Vernon hills, Illinois, United States of America


We investigate the impact of mismatched attenuation maps on PET ROI quantitation as a function of the extent of the attenuation mismatch, system timing resolution and distance between the ROI and the region where the mismatch occurs. We simulated Toshiba's Celesteion$^{\mathrm{TM}}$ TOF-PET/CT scanner geometry at 200, 400 and 800ps timing resolutions with a 40cm diameter uniform cylindrical phantom as well as a nonTOF version of the scanner. The phantom had 20mm diameter spheres inserted at the center of the phantom and 19 cm off-center transaxially. We reconstructed images using OSEM with perfectly matched attenuation maps and with attenuation maps that were offset in-plane 0.5cm to 2cm in 0.5cm increments and compared quantitation in these mismatched cases to those with the perfectly matched case. The results show that quantitation is affected only 2\% for the centered sphere for the largest mismatch case at 200 ps timing resolution whereas the quantitative impact of the same error for the nonTOF system is 13\%. For the sphere at the location of the mismatch, even the 200ps system had a 68\% quantitation error

for the largest mismatch case while the nonTOF system had a 77\% error. We conclude that while systems with high timing resolution can largely make up for quantitation errors for lesions away from areas with large attenuation mismatch, all systems are significantly affected if the lesion occurs at or around the location of the mismatch.

Keywords: TOF-PET, Attenuation mismatch, Time resolution
Poster panel
(face) ID: 172

Poster Number:

PET/MRI Brain Attenuation Correction Using a Random Forest Regressor (#3251)

C. N. Ladefoged1, A. E. Hansen1, I. Law1, L. Højgaard1, F. L. Andersen1

1 Rigshospitalet, Department of Clinical Physiology, Nuclear Medicine and PET, Copenhagen, Denmark


AIM: A large number of methods have been proposed to account for the missing bone during MR attenuation correction (AC), including our UTE-based RESOLUTE method. Since the UTE images are often noisy in air/tissue-interfaces, we originally added region specific noise suppression to the method. While this approach have proven robust in adults and pediatrics, the method does require a template with drawn regional masks in order to be successful in new study cohorts, e.g. in imaging of animal models. In this study, propose a refinement to the original method that replaces the need for regional masks using a machine-learning approach, while preserving the patient specific information.

METHODS: We included 201 [18F]FDG patients imaged on a Siemens PET/MRI scanner, of which 50 was used for training of a Random Forest model. For feature selection, we extracted local mean and standard deviation information at varying neighborhoods around each voxel for each input image, as well as the subsequently calculated gradient and R2* maps. Furthermore, we calculated the spatial information and local binary patterns. The random forest regressor was trained using 500 trees, and evaluated on the remaining 151 patients. The classified bone in the predicted attenuation maps was replaced with the estimated bone density, similar to the technique used in RESOLUTE, and was evaluated against a co-registered CT reference.

RESULTS: Across all patients, the Jaccard index was higher in the proposed method (0.76) compared to UTE and RESOLUTE (0.56 and 0.68). The relative %-error in PET uptake was improved over Dixon and UTE using the proposed method both globally and regionally, with errors within 2.5% of PET with CT.

CONCLUSIONS: Compared to RESOLUTE, the proposed method removes the need for regional masks, without compromising with the PET accuracy, suggesting the updated version can replace the original RESOLUTE, making it readily applicable to larger patient cohorts.

Keywords: PET/MRI, Attenuation correction, machine learning, random forest, brain
Poster panel
(face) ID: 175

Poster Number:

Robust Estimation of Scatter and Primary Signals using Multi-View Information for Moving Blocker-Based Cone-Beam Computed Tomography (#3548)

C. Zhao1, L. Ouyang2, X. Chen2, J. Wang2, M. Jin1

1 University of Texas at Arlington, Physics, Arlngton, Texas, United States of America
2 University of Texas Southwestern Medical Center, Radiation Oncology, Dallas, Texas, United States of America


The moving blocker method has been demonstrated to be effective for scatter correction (SC) of cone-beam computed tomography (CBCT). However, a fixed thresholding was usually used to detect the blocker edge in the projection images, which only works in high contrast uniform regions. The edge detection errors, caused by low contrast and non- uniform intensity, can lead to significant artifacts in the reconstructed images. In this work, we propose to use the projection images at multiple adjacent views to estimate the scatter signal directly without identifying the blocked regions. Subsequently, the scatter signal can serve as an adaptive threshold to locate the unblocked regions for the estimation of the primary signal. The experimental results using an anthropomorphic pelvis phantom CBCT data show that the proposed method can robustly estimate the scatter and primary signals by avoiding blocker edge detection errors and lead to much improved reconstruction images. 

Keywords: Scatter correciton, Robust estimation, Cone-beam computed tomography (CBCT), Multiple projection views
Poster panel
(face) ID: 178

Poster Number:

Study of a Convolutional Autoencoder for Automatic Generation of MR‐based Attenuation Map in PET/MR (#3722)

K. S. Lee1, L. Tao1, C. S. Levin1, J. Best‐Devereux1

1 Stanford University, Radiology, Stanford, California, United States of America


Quantitative PET image reconstruction requires an accurate map of photon attenuation coefficients (μ-map) in order to correct the emission data. Current PET/MR imaging systems use methods based on MR image segmentation with subsequent assignment of empirical attenuation coefficients. Delineation of bone in MR images has been challenging, especially in the head and neck areas, due to the difficulty of separation of bone from air, which often causes segmentation and tissue classification inaccuracy. Machine learning and deep learning have been implemented in various areas, including medical imaging to solve problems that can be easily recognized by human perception, but hard to resolve from digitized media. In this study, we examine a novel deep learning based attenuation map generation method that can mitigate the error that causes segmentation or tissue classification discrepancies. We explore the use of a convolutional autoencoder network to directly convert the MR image to an attenuation map through trained data based on CT-based attenuation maps at 511 keV (CTAC) for PET attenuation correction for the head and neck region. We found that bone in the head, shoulder and neck areas started to show up in the network output images after 50 iterations. The training loss is systematically reduced to 70% at 10 iterations, and continues to go below 50% at 50 iterations. Ongoing work will focus on examining the impact of the attenuation map that is directly generated from the MR data on PET quantification, compared with CTAC as a gold standard with the goal of improving PET quantification.

Keywords: PET/MR, machine learning, convolutional autoencoder, MR-based attenuation map, attenuation correction, PET quantification
Poster panel
(face) ID: 181

Poster Number:

Mapping the Response of a Double-Sided Strip High Purity Germanium Detector (#3963)

R. S. Perea1, 2, L. R. Furenlid3, S. Shokouhi2, T. E. Peterson1, 2

1 Vanderbilt University, Physics and Astronomy, Nashville, Tennessee, United States of America
2 Vanderbilt University Medical Center, Institute of Imaging Sciences, Nashville, Tennessee, United States of America
3 University of Arizona Department of Medical Imaging, Center for Gamma-Ray Imaging, Tucson, Tennessee, United States of America


We are developing a dual-headed small-animal SPECT system using double-sided strip HPGe detectors. HPGe provides superior energy resolution (~1% FWHM at 140 keV), which allows for good scatter rejection and facilitates dual- or multi-isotope imaging. Reconstructed images in our first prototype small-animal SPECT system exhibited artifacts, which we attributed to mis-positioning of events near the strip edges. To study the detector response and acquire data to facilitate the development of methods to recover multi-strip events, we have scanned a detector with a focused beam (~25x25 µm2 @ 131 keV) at the Advanced Photon Source. These data showed dependencies with our position ratios (calculated using the fast signal from adjacent strips), the position of the beam on the strip, and the depth of interaction. Furthermore, the data also show position dependence of charge-shared events near and within the gap. We are using these data to develop improved position estimation schemes that remove spatial distortions and recover multi-strip events. We also plan to investigate other methods for position estimation, such as maximum likelihood, with the aim of improving spatial resolution. These developments are expected to result in higher contrast and sensitivity in our SPECT images.

Keywords: HPGe, SPECT
Poster panel
(face) ID: 184

Poster Number:

Detection of lung density variations with Principal Component Analysis in PET (#4072)

O. Bertolli1, V. Cuplov2, S. Arridge3, C. W. Stearns4, S. D. Wollenweber4, B. F. Hutton1, K. Thielemans1

1 University College London, Institute Of Nuclear Medicine, London, United Kingdom of Great Britain and Northern Ireland
2 IMIV, U1023 Inserm/CEA/Université Paris-Sud and ERL 9218 CNRS, Université Paris-Saclay, Orsay, France
3 University College London, Centre for Medical Image Computing, London, United Kingdom of Great Britain and Northern Ireland
4 GE Healthcare, Waukesha, Wisconsin, United States of America


Respiration during PET acquisition leads to chest volume changes and therefore density variations. These changes affect the attenuation and the detected counts, therefore introducing respiratory-induced effects that could impair PET quantitation. It is desirable to detect these effects so that appropriate correction can be applied without the need for the acquisition of Cine CT.

Principal Component Analysis (PCA) has been used for the detection of respiratory movement from PET data. In this study, we propose the use of PCA to detect respiratory-induced density variations in the upper part of the lungs, where motion is minimal.

To assess the ability of PCA to detect density-only changes, we include both simulation and patient data. Simulated data were generated from an FDG-PET patient study, with Cine CT. Both PET and CT data were displacement gated (5 gates) with an external respiratory signal device (RPM). Motion fields from the inhalation state to the other 4 gates were obtained from the CT data using non-rigid registration, and the Jacobian determinants (that have been shown to represent the local volume changes) of the deformation fields were obtained. Projection data were created from the 4 PET gates generated by applying the Jacobian to the end-inhalation gate and attenuated with the deformed CT inhale gate and PCA was applied. PCA was also applied on dynamic sinograms of 16 PET/CT datasets of the upper lung, acquired with RPM.

The trend of the PCA weight factor in both simulation and patient data suggests that the signal is related to respiratory-induced changes. The variation of the total counts in the simulated and patient data demonstrate that PET data changes are dominated by attenuation variation. The PCA signal from patient data showed high correlation with the RPM (mean 0.76, max 0.94) even if no moving organs (other than lung) are in the field-of-view, demonstrating that PCA can be used to detect respiratory-induced density changes in the upper lung.

Keywords: Lung density variation, PET, PCA
Poster panel
(face) ID: 187

Poster Number:

A quick and easy method for patient-specific quantification of spatially varying noise in dedicated-cardiac SPECT images (#1726)

S. G. Cuddy-Walsh1, 2, R. G. Wells2, 1

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


Increased noise in single photon emission computed tomography (SPECT) images can decrease the detectability of heart disease in a patient. Image noise depends on the amount of attenuation introduced by the patient tissues and on the camera geometry. Camera geometry is more important for pinhole cameras as they have larger noise gradients in the field-of-view (FOV) than parallel hole cameras. We have developed a 1 minute method for estimating patient specific camera noise and validated it against a bootstrapping (BSV) method (5h, 392 BSV images). Our projected voxel (PV) method calculates the Poisson noise in the sum of detected counts determined via an attenuated forward projection of each voxel in a reconstructed image.  An empirical formula is used to calibrate the noise estimate with count level dependent coefficients. Calibration was performed for a uniform 99mTc-water sphere phantom at 8 count levels and applied to 31 patient images with mild or no visual disease (350MBq, 99mTc-tetrofosmin rest dose).  Discovery NM530c SPECT (GE Healthcare) acquired images were reconstructed with 60 iterations of MAP-EM and CT attenuation correction (GE Infinia HawkEye 4). Phantom and clinical cases both show a linear relationship between PV and BSV estimation methods with R2 = 0.70 and CV = 17%. BSV noise varied from 4–12% over the length of the heart for an individual patient and from 6–11% on average. The 3-fold change in noise over the length of the heart has the potential to affect the detectability of heart disease as it is of similar magnitude to the inter-patient variability measured with normal databases.  Use of patient-specific image noise maps may thus assist with physician interpretation.

Keywords: personalized medicine, personalized imaging, image noise, SPECT, cardiac imaging, pinhole, noise method, position-sensitive noise, noise gradient, variable disease detectability, dedicated cardiac SPECT
Poster panel
(face) ID: 190

Poster Number:

PET Image Reconstruction from Under-sampled Data (#1981)

Z. Hu1, J. Gao1, Y. Yang1, X. Liu1, H. Zheng1, D. Liang1

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


To decrease the cost of positron emission tomography (PET) scanning systems, image reconstruction algorithms for under-sampled data have been extensively studied. For example, total variation (TV) minimization regularization was nested in the maximum likelihood-expectation maximization (MLEM) method to suppress noise (MLEM-TV). However, the gradient operator of the TV cannot distinguish true structures from noise; that is, an image reconstructed by the MLEM-TV algorithm may lose some fine features. Recently, we developed a feature refinement (FR) approach for statistical interior computed tomography (CT) reconstruction. Its potential usefulness for feature preservation of interior tomography with truncated projection measurements has been demonstrated. Inspired by this study, the FR step is applied to PET imaging for recovery of fine features. In the implementation of the proposed method, filtered back-projection (FBP) reconstruction is first used as the initial guess to accelerate convergence. Moreover, the FR step, as an important processing operation, is performed after each MLEM-TV iteration to extract the desired structural information lost in the TV minimization. Here, a feature descriptor is specifically designed to distinguish structure from noise and artifacts. A modified steepest descent method is adopted to minimize the objective function. To study the MLEM-TV-FR algorithm and evaluate its performance, we simulated realistic brain PET imaging data using an emission activity phantom and compared the results to those of FBP, MLEM and MLEM-TV. The reconstruction results indicate that the proposed method is more effective than other methods at suppressing noise, reducing artifacts and preserving features in under-sampled data.

Keywords: positron emission tomography (PET), maximum likelihood-expectation maximization (MLEM), under-sample, total variation (TV), feature refinement
Poster panel
(face) ID: 193

Poster Number:

New brain phantoms suitable for brain scanners with hemisphere detector arrangement (#2845)

G. Akamatsu1, H. Tashima1, H. Wakizaka1, T. Maeda1, Y. Iwao1, E. Yoshida1, T. Yamashita2, T. Yamaya1

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


The brain PET imaging technique is very useful for clinical practice and molecular imaging research. Both high sensitivity and high resolution are required for accurate imaging. Therefore, we developed the brain-dedicated compact PET scanner which has a hemispherical helmet detector unit and an add-on detector unit located at the chin position (helmet-chin PET). We developed its prototype system using 4-layer depth-of-interaction detectors. To realize clinical applications of the PET, we need to evaluate the PET system regarding their absolute quantitation, image contrast, noise and uniformity using appropriate phantoms. However, existing standard phantoms are not applicable for the helmet-chin PET scanner because of its hemisphere geometry. Therefore, in this study, we developed two new phantoms, a small sphere contrast phantom and a 3-dimensional hemisphere brain phantom, which model a brain tumor and static regional cerebral blood flow, respectively. These phantoms have an adaptive structure for use with the helmet-chin PET scanner. We applied the developed contrast phantom to the helmet-chin PET prototype and a commercial whole-body PET/CT scanner to evaluate imaging performance of image contrast and noise. As a result, the helmet-chin PET prototype showed higher image contrast and lower image noise compared with the commercial whole-body PET/CT scanner. In conclusion, we were able to evaluate imaging performance of the helmet-chin PET prototype using the developed phantoms. These phantoms are considered to be useful for evaluating the imaging performance of brain-dedicated PET scanners, and for determining the appropriate scanning and reconstruction parameters.

Keywords: Positron emission tomography, brain PET, helmet PET, brain phantom
Poster panel
(face) ID: 196

Poster Number:

A Phantom Design and Demonstration of Contrast-Size Features in PET (#3638)

S. D. Wollenweber1

1 GE Healthcare, MICT Engineering, Waukesha, Wisconsin, United States of America


Lesion detection is an important task for radionuclide imaging. Emulation of the broad spectrum of oncologic imaging conditions is difficult. Fillable phantoms are often used in the assessment of system and algorithm design as they enable repeated measurement of objects of known size and contrast. This work aimed to design and demonstrate the use of 3D-printed cage-like spheres that, when placed into a background of random close-packed balls, produce PET image features with known size and contrast including various sizes and contrast levels. This phantom design is modeled after prior work where 3D-printed nylon dodecahedrons produced ~3:1 contrast features in phantoms filled with the non-porous acrylic balls. Such a phantom is useful in the assessment of system and algorithm designs as it allows quantitative comparison of contrast recovery as a function of feature size and contrast all from the same phantom measurement. Many such features can be placed into a given phantom, enabling measurement of precision and bias with spatial and statistical variation included. Using percent contrast measurement across the sizes and contrasts demonstrated the capability to quantitatively assess differences in system and algorithm design.

Keywords: PET, Image Quality, Phantoms
Poster panel
(face) ID: 199

Poster Number:

Performance Evaluation of Quantitative SPECT/CT: Applying NEMA NU2 PET Measurements to SPECT (#3952)

H. Ryu1, E. Eslick1, 2, K. P. Willowson3, S. R. Meikle1, D. L. Bailey1

1 The University of Sydney, Faculty of Health Sciences, Sydney, NSW, Australia
2 The University of Sydney, Sydney Vital Translational Cancer Research Centre, Sydney, NSW, Australia
3 The University of Sydney, Institute of Medical Physics, Sydney, NSW, Australia


As quantitative SPECT systems have recently become commercially available, we compared their performance with that of PET systems. SPECT and PET systems were tested using the methodology defined in NEMA-NU2 Performance Measurements of Positron Emission Tomographs (2012).

99mTc and 177Lu were used to test on the SPECT camera (Symbia Intevo, Siemens Healthcare) and 18F was used for the PET camera (Biograph mCT, Siemens Healthcare). The spatial resolution was measured in radial and tangential directions in six positions. Scatter fraction and count losses were measured with a line source inserted into a cylinder phantom. The sensitivity was measured using a uniform line source surrounded by aluminum sleeves of varying thickness at the centre of the FOV and at 10cm offset. Five aluminum sleeves with different diameters were used to measure the attenuated count rate with accumulated wall thicknesses to extrapolate to “zero attenuation”.

The spatial resolution for radial/tangential directions at 1 cm were: 5.29/4.30 mm, 13.1/14.1 mm and 21.6 mm/23.4 mm for 18F, 99mTc and 177Lu, respectively. The radial and tangential resolutions at 10 cm were: 5.25/5.84 mm, 12.4/10.8 mm and 20.5/17.6 mm for 18F, 99mTc and 177Lu, respectively. The event rates of 99mTc showed linear increase with activity. The count loss of 177Lu at the highest activity concentration (13 GBq) was 20%. The sensitivities at 0/10cm off axis were: 9.63×103/9.59×103 cps/MBq, 119.4/116.7 cps/MBq, and 47.6/47.5 cps/MBq for 18F, 99mTc and 177Lu, respectively. The scatter fractions of 18F, 99mTc and 177Lu were 0.32, 0.25 and 0.36, respectively. Imaging performance was evaluated using a NEMA IEC body phantom. The background variabilities were 1.92%, 1.38% and 1.83% for 18F, 99mTc and 177Lu, respectively. The ratio of the SUV values in hot spheres and the background on the quantitative image was within 4% error compared to the true ratio.

This method can be used for the evaluation of a quantitative SPECT system.

Keywords: NEMA NU2, Quantitative measurements, SPECT/CT, PET
Poster panel
(face) ID: 202

Poster Number:

Alternative Derivations of M-Line Methods in Cone-Beam CT (#1059)

D. Shi1

1 Shenyang Campo Medical Imaging Technology Co., Ltd., Benxi, Liaoning, China


In this work, we provide alternative derivations of backprojection-filtration (BPF) and filtered backprojection (FBP) reconstruction algorithms along M-lines in cone-beam CT. In the derivations of BPF algorithms, three different modified data functions are employed. One of the modified data functions is equivalent to the one in the original work except for a minus sign in our result. The other two modified data functions will result in an equivalent relationship without boundary terms compared to the original work. In our derivation of FBP algoirthms along M-lines, we linked the object function to its Fourier data instead of its Radon data as in the original derivation. By doing so, we found that the BPF and FBP algorithms share exact same intermediate format. It turns out our derivation of FBP method along M-lines is a little simpler.

Keywords: M-line, BPF, FBP, cone-beam CT, Reconstruction algorithm
Poster panel
(face) ID: 205

Poster Number:

Performance Improvements in HYPR-OSEM (#1339)

J. - C. (. Cheng1, 2, J. Matthews2, R. Boellaard3, 4, I. Janzen5, 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 methods which improve the performances in convergence and in contrast recovery coefficient (CRC) versus noise of the newly developed HYPR-OSEM algorithm. HYPR-OSEM is a reconstruction algorithm which incorporates HighlY constrained back-PRojection (HYPR) de-noising directly within the widely used OSEM algorithm. 3 forms of HYPR-OSEM have been proposed. Previously, we have demonstrated that all forms of HYPR-OSEM can improve SNR without degrading accuracy in terms of resolution and contrast, and they can achieve better precision than OSEM with similar accuracy and better accuracy than filtered OSEM with similar precision. However, slower convergence in CRC was observed from all forms of HYPR-OSEM. In this work, we investigated the effect of the filter kernel size used in the HYPR operator. Furthermore, we introduced the Iterative HYPR (IHYPR) operator as an effort to accelerate the convergence in CRC. 10 independent noisy realizations of a simulated contrast phantom with various sizes of hot and cold regions were used for the evaluations. CRC vs voxel noise was generated for each hot and cold region. It was observed that the noise reduction performance of HYPR-F(B)-OSEM is not very sensitive to the filter kernel size used in the HYPR operator, whereas better CRC vs noise trajectories can be achieved by wider kernels for HYPR-AU-OSEM. On the other hand, the CRC convergence for HYPR-AU-OSEM becomes much slower with a wider kernel. When the IHYPR operator was introduced into the AU method (i.e. IHYPR-AU-OSEM), similar CRC convergence speed with respect to OSEM was attained without excessively degrading the CRC vs noise trajectories. In summary, the AU method has been determined to be the more effective form of HYPR-OSEM in terms of accuracy and precision, and IHYPR-AU-OSEM can achieve better CRC vs noise trajectories with similar convergence speed as compared to OSEM with and without a post reconstruction filter.

Keywords: IHYPR, HYPR-OSEM, PET reconstruction
Poster panel
(face) ID: 208

Poster Number:

Randoms Smoothing Based on Rates of Object Singles for CBM PET Scans (#1905)

B. Feng1, V. Y. Panin1

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


For quantitative PET, estimation of mean randoms from the measurement of delays relies on accurate estimate of the singles rate. For a continuous bed motion scan, the detector singles rates change constantly over time during a CBM scan, due to different parts of the object being measured at different times, effects of detector dead time and radioactive decay, and shielding for radiations outside the gantry. The conventional formulation based on detector singles rates involves averaging a nonlinear function of singles rates over time, which greatly increases the number of unknowns and the difficulty of estimation. In this work, a new formulation of the CBM mean randoms was proposed, based on the object singles rates, which can be estimated by the single-bed random smoothing algorithms. Under approximations, a simplified approach is also derived, which requires no knowledge of the detector singles efficiencies, physical degrading factors, and acquisition parameters. The simplified approach first processes the 3D CBM delays as single-bed data and then applies plane-by-plane scaling to the mean random sonograms, to equalize the total counts of each sinogram plane to that of raw delays. Our method was evaluated by simulations and phantom measurements. Unbiased estimations were obtained with much smaller variances than the raw delays. In conclusion, we developed a CBM randoms smoothing method that utilizes the fully 3D delays and leads to estimates that approximate the maximum likelihood estimates.

Keywords: PET Randoms Smoothing, CBM
Poster panel
(face) ID: 211

Poster Number:

An Image Reconstruction Method with Locally Adaptive Gating Scheme for PET Data (#2167)

J. Wang1, Y. Dong2, H. Li1, T. Feng1

1 UIH America, Inc., Houston, Texas, United States of America
2 Shanghai United Imaging Healthcare Co., Ltd, Shanghai, China


Traditional gating scheme using a pre-determined gating number for the whole FOV is unable to achieve balance between temporal resolution and statistics for each local region. The shortcoming of traditional gating method becomes more problematic for PET system with extended FOV such as the EXPLORER system. Our goal is to develop a new image reconstruction method incorporating locally adaptive gating to optimize image quality.

Amplitude based data-driven respiratory gating technique was used in dividing the list-mode PET data into four equal-count gates. 2D motion vector field from Frame 1 to Frame 4, i.e. from end-inspiration to end-expiration, was estimated from the maximum intensity projections of the reconstructed images using B-spline image registration algorithm. Optimal gating number for each location in FOV was determined by dividing the calculated motion amplitude with spatial resolution of the PET system. The gating number was used to build a Gaussian kernel in time domain for each location, based on the rationale that region of large motion requires higher temporal resolution. We designed a new image reconstruction algorithm incorporating the modeling of locally adaptive temporal resolution based on OS-EM algorithm framework. The proposed method was applied to clinical PET data collected from suspected lung tumor patients. Compared with reconstructed images of gated data using conventional method, image reconstructed using the new reconstruction algorithm was able to optimize resolution-noise trade-offs for every region and achieve higher image quality than static or conventional gated reconstruction.

In summary, we proposed a new image reconstruction method to incorporate locally adaptive temporal resolution determined by motion amplitude. While traditional gated PET reconstruction risks over-gating for small-motion area or under-gating for large-motion area, our method is able to obtain optimal gating for each location to achieve the best image quality.

Keywords: image reconstruction, PET, locally adaptive gating
Poster panel
(face) ID: 214

Poster Number:

MR-Resolution Kernel Method for PET Reconstruction (#2239)

J. Bland1, S. Ellis1, M. A. Belzunce1, A. Mehranian1, C. J. McGinnity2, A. Hammers2, A. J. Reader1

1 King's College London, Biomedical Engineering, London, United Kingdom of Great Britain and Northern Ireland
2 King's College London, Kings College London & Guy's and St Thomas' PET Centre, London, United Kingdom of Great Britain and Northern Ireland


Conventional PET reconstruction produces noisy images.  Recently proposed techniques such as the MR-guided kernel method have been employed to reduce the impact of noise, whilst retaining important image details.  However, this can lead to over smoothing of PET unique features.  To address this issue, this work extends the MR-guided kernel method to use MR resolution basis functions, which are extracted from an MR image at its native resolution.  Furthermore, this MR-resolution kernel method is modified to produce spatially constrained basis functions in order to limit the smoothing of PET-unique features whilst still reducing the impact of noise.  The MR-resolution kernel reconstruction is compared to MLEM and conventional PET resolution kernel methods for tumour contrast recovery.  These methods are applied to real patient FDG data augmented with simulated tumours.  The proposed kernel method shows an improved contrast to noise ratio compared to the conventional kernel method for all tumour sizes.  However, MLEM attained a higher contrast to noise ratio for the small tumour.  In summary, the MR resolution spatially constrained kernel method maintains the noise reduction properties of the conventional kernel method implementation, whilst better retaining the features unique to the PET data.  

Keywords: PET, reconstruction, kernel, mri
Poster panel
(face) ID: 217

Poster Number:

Three-dimensional radiopharmaceutical-excited fluorescence imaging of lymph nodes (#2575)

H. B. Guo1, 2, X. W. He1, M. H. Liu2, Z. Y. Zhang2, X. J. Shi2, Z. H. Hu2, J. Tian2

1 Northwest University, The School of Information Sciences and Technology, Xi'an, China
2 Chinese Academy of Sciences (CAS), Key Laboratory of Molecular Imaging, Institute of Automation, Beijing, China


Optical imaging techniques have been developed for localizing lymph nodes before surgical resection due to the non-invasion and high sensitivity. However, its attendant penetrability limitations and auto-fluorescence effect have greatly limited the spatial resolution and imaging precision. A novel three-dimensional optical molecular imaging technique named radiopharmaceutical-excited fluorescence imaging (REFI) was used for imaging and assessing superficial lymph node. REFI maken use of gamma-ray photon and Cerenkov radiation from radioisotopes to excite lanthanide nanophosphors europium oxide (EO) and boost the light intensity. In order to improve the lymph nodes reconstruction quality, an adaptive-steepest-descent-projection onto convex sets (ASD_POCS) reconstruction algorithm based on total variation (TV) regularization was proposed, which minimizes the TV objective using adaptive steepest descent and the data fidelity error using projection onto convex sets (POCS). Our research find that REFI can shift spectral from blue to red (620 nm), when using ultraviolet to blue spectrum to excite EO. And REFI can greatly boost the light intensity. In a common used 3D digital mouse atlas, the proper axillary lymph node and accessory axillary lymph node were regard as radiopharmaceutical-excited fluorescence source. The reconstructed lymph nodes by our proposed ASD_POCS algorithm show that the location deviation of the proper axillary lymph node and accessory axillary lymph node were 1.03mm and 0.89mm, respectively. The results indicate that REFI combined with ASD_POCS algorithm can greatly boost the light intensity and three-dimensional imaging lymph node accurately.

Keywords: Optical imaging, radiopharmaceutical-excited fluorescence imaging, localizing lymph nodes, reconstruction algorithm
Poster panel
(face) ID: 220

Poster Number:

3D reconstruction based on origin ensembles algorithm and multi-focus method for Compton camera imaging (#2711)

Z. Yao1, 2, Y. Xiao1, 2, Z. Chen1, 2

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

On behalf of THU and UCL Collaboration


Compton camera has been proposed as an imaging tool of high potential in astronomy, industry, homeland security, and medical imaging. As it allows radiation density images to be reconstructed at different depths from a single shot without tomography and without moving the patient, Compton camera is expected to be used in heavy ion treatment inline monitoring. Improve its resolution and reduce the reconstruction time is the key to facilitate the progress of the use in clinic therapy. In this paper, we present an algebraic method to acquire the point on the event conical surface using inverse and transfer matrix, which has simple form and high precision. The statistical reconstruction method based on origin ensembles (OE) was presented which adopt a probability density function (pdf) to determine ensemble expectation of numbers for detected event origins per voxel. To improve resolution in the three-dimensional, a multi-focus method presented in this paper was studied. Instead of increase the multi-views by hardware such as using multiple detectors combined image, the multi-focus method rely on software. Multi-focus method addressed in this paper uses multiple imaging planes (or imaging space ) to focus the sources, which is promising to enhance the spatial resolution in Compton camera imaging using OE algorithm, and weaken or even eliminate the effect of spatial orientation, which can cause distortion in the direction perpendicular to the imaging plane.  The preliminary numerical simulation results demonstrate the feasibility and effectiveness of the proposed method. Multi-focus method works well in two sources image reconstruction, with remarkable promotion in both image contrast and resolution capability of two sources. It's promising that multi-focus method can be used in Compton camera to improve spatial resolution and imaging accuracy in clinic therapy.

Keywords: 3D reconstruction, Compton camera imaging, multi-focus, origin ensembles algorithm
Poster panel
(face) ID: 223

Poster Number:

Data-driven Improved Sampling in PET (#2831)

P. Galve1, A. López Montes1, J. M. Udías1, S. C. Moore2, J. L. Herraiz1

1 Universidad Complutense de Madrid, Nuclear Physics Group, Madrid, Spain
2 Brigham and Women’s Hospital & Harvard Medical School, Division of Nuclear Medicine, Boston, Massachusetts, United States of America


Positron Emission Tomography (PET) scanners are usually designed with the goal to obtain the best compromise between sensitivity, resolution, field-of-view size, and cost. Therefore, it is difficult to improve the resolution of a PET scanner with a hardware modification, without affecting some of the other important parameters. Iterative image reconstruction methods such as the ordered subsets expectation maximization (OSEM) algorithm are able to obtain some resolution recovery by using a realistic system response matrix that includes all the relevant physical effects. Nevertheless, this resolution recovery is often limited by reduced sampling in the projection space, determined by the geometry of the detector.

The goal of this work is to improve the resolution beyond the detector size limit by increasing the sampling with data-driven interpolated data. A maximum-likelihood estimation of the counts in each virtual sub-line-of-response (subLOR) is obtained after a complete image reconstruction, conserving the statistics of the initial data set. The new estimation is used for the next complete reconstruction. The method typically requires two or three of these full reconstructions (superiterations). We have evaluated it with simulations and real acquisitions for the Argus and Super Argus preclinical PET scanners manufactured by SMI, considering different types of increased sampling. The procedure is able to reduce significantly the impact of depth-of-interaction in large crystals, and improve the spatial resolution. The peak-to-valley ratio of 1.5 mm rods of a Derenzo phantom acquired with the SuperArgus scanner is improved by up to 41±3% when transversal subdivisions are considered, while the noise in the resulting images is slightly increased (up to a relative 10%). The proposed method is quite general and it can be applied to other scanners and configurations.

Keywords: Positron Emission Tomography, resolution recovery, superiterative methods, maximum likelihood
Poster panel
(face) ID: 226

Poster Number:

Compressed-sensing (CS) Based Reconstruction in Region-of-interest (ROI) Digital Tomosynthesis with a Dual-resolution Voxellation Strategy (#2983)

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

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


Common digital tomosynthesis (DTS) based on filtered-backprojection (FBP) reconstruction requires a full field-of-view scan and relatively dense projections, which results in high dose for medical imaging purposes. To overcome these difficulties, we investigated region-of-interest (ROI) DTS reconstruction where the x-ray beam span covers only a small ROI containing a target area to reduce excessive dose to the patient. However, ROI-DTS measures incomplete (i.e., truncated) projection data so that conventional FBP-based algorithms are not successful in producing clinically feasible images. In this work, an iterative method based on compressed-sensing (CS) scheme was considered for better reconstruction quality in ROI-DTS. Further, we propose a dual-resolution voxellation strategy for the CS-based ROI-DTS reconstruction in which the voxels outside ROI are binned while the other voxels remain unbinned to relieve computational burden in the iterative process. We implemented the proposed algorithm and performed a systematic simulation and experiment. Our results indicate that the proposed method seems effective for reducing the patient’s dose and the computational burden considerably, keeping the reconstruction quality inside ROI not much degraded.

Keywords: Digital tomosynthesis, Compressed sensing, Dual resolution voxellation
Poster panel
(face) ID: 229

Poster Number:

CT iterative reconstruction within the CASToR platform using GPU architecture (#3112)

D. Benoit1, T. Merlin1, J. Bert1, T. Carlier2, F. Lamare3, D. Visvikis1

1 LaTIM - INSERM UMR1101, CHRU, Brest, France
2 CRCNA - University Hospital - Nantes Hotel Dieu, Nantes, France
3 INCIA, UMR 5287, University Bordeaux,, Service de Medecine nucleaire, Bordeaux, France


Iterative CT reconstruction recently regained interest due to computational hardware architecture improvements. Among other advantages, it allows the use of more accurate acquisition models as well as better handling of incomplete data compared toanalytical reconstruction, the main disadvantage being the increased computation time. In this work, we propose a first integration of iterative CT reconstruction and GPU implementation in the open-source multi-modal CASToR reconstruction platform.

The original CASToR framework separates the computation of the system matrix elements and the operations of the optimization algorithm. This architecture had to be modified for GPU computing, as the system matrix elements could no longer be stored due to the limited amount of memory in a GPU and the number of lines being simultaneously processed. The proposed implementation merge the projection and optimization parts of the code in a single GPU kernel. The system matrix elements are computed on the fly to get the forward projected value of a specific optimizer algorithm, then recomputed for the back-projection step. All these operations are implemented inside the same GPU kernel which is specific to a projector/optimizer pair.

Assessment of the new implementation were performed on a head phantom acquired using the Varian TrueBeam CBCT system. In order to evaluate the efficiency of the new implementation in terms of computing performance, the computation time of images generated by the forward-projection operation were compared. The GPU implementation led to a significant reduction factor (11,4 on the test machine) for a forward projection in comparison with the CASToR CPU implementation.

This study proposes a new implementation of the CASToR architecture for the integration of more efficient parallel computing using GPU and iterative CT reconstruction. On-going work focuses on improving the proposed implementation in order to preserve the original platform’s generic features.

Keywords: Recontruction, CT, GPU, CASToR, PET, SPECT
Poster panel
(face) ID: 232

Poster Number:

ML and MAP PET reconstruction with MR-voxel sizes for simultaneous PET-MR (#3256)

M. A. Belzunce1, A. Mehranian1, J. R. Bland1, A. J. Reader1

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


The introduction of clinical simultaneous PET-MR scanners has brought new opportunities to use parametric MR information to assist PET image reconstruction. In this context, MR images are usually downsampled to the PET resolution before being used as anatomical priors in MR-guided PET reconstruction. However, the reconstruction of PET images at the MR-voxel size could achieve a better utilization of the high resolution anatomical information and improve the partial volume correction obtained with these methods. When the PET reconstruction needs to be done in a higher resolution matrix a number of artifacts arise in the image reconstruction, depending on the projector and system matrix used. In this work, we consider the pitfalls of performing PET image reconstruction with a smaller voxel size than the standard native pixel size of the scanner. We look at the singular values of the system matrices for standard and higher resolution voxel sizes to show that higher frequencies could be recovered with the latter. However, when using the MR-voxel sizes system matrix, artifacts are obtained. We propose a method that modifies the system matrix to overcome these difficulties and we show reconstructed images of a NEMA phantom and patient data for standard and high resolution image sizes. The higher resolution reconstructed images show a better delineation of the edges and a mild improvement of the contrast in the smallest spheres of the NEMA phantom. In addition, we evaluated the method for MR-guided MAP reconstruction, where patient data was reconstructed using a Bowsher prior computed from the T1-weighted image in its original resolution. The higher resolution reconstructed image showed a better definition of the structures of the brain, showing that MR-guided MAP reconstruction with MR-voxel sizes can enhance partial volume correction.

Keywords: PET-MR, high-resolution PET reconstruction, MR-guided PET reconstruction, System Matrix, SVD
Poster panel
(face) ID: 235

Poster Number:

Multigrid tomographic reconstruction (#3551)

B. Gao1, D. Larsson1, M. Colarieti-Tosti1, J. - W. Buurlage2, H. Kohr2

1 KTH Royal Institute of Technology, Medical Engineering, Stockholm, Sweden
2 CWI, Centrum Wiskunde & Informatica, Computational Imaging, Amsterdam, Netherlands


In this work we present a novel methodology for multigrid image reconstruction of tomographic data. With multigrid reconstructions we refer to a reconstructed volume with different discretisation domains, here accomplished by treating the different discretisation domains separately by masked forward and backward operators. The methodology is tested on both a numerical Shepp-Logan phantom, and a physical phantom (QRM low-contrast phantom). Reconstructions are performed using several reconstruction techniques, as well as for multiple setups of discretisation domains. The results indicate that accurate high-resolution ROIs can be reconstructed within an otherwise coarse surrounding grid, and that only for the iterative reconstruction schemes and only when the surrounding grid is very coarse (ratio 1:40) can visible artefacts be seen in the high-resolution ROI. This seemed to be the case for both numerical and experimental tests. Also, for the imaged QRM-phantom, accurate insert contrast (7% and 6% for CG and TV, compared to ground-truth  8%) could be seen down to multigrid setups of 1:8 discretisation differences. Overall, the work presents an easily implementable technique for multigrid image reconstruction, being of benefit when having to save computational time or resources. Also, with the technique of masked reconstruction operators, potential multigrid combinations of different reconstruction schemes in different disretisation domains could be attempted. With that said, further validation of the technique is needed prior to any large-scale implementation.

Keywords: Reconstruction, CT, Multiresolution, Multigrid, Tomography
Poster panel
(face) ID: 238

Poster Number:

SNR TOF gain in high time resolution PET systems (#3674)

D. K. Bharkhada1, L. Eriksson1, 2, M. Conti1, H. Rothfuss1

1 Siemens Medical Solutions USA, Inc., Knoxville, Tennessee, United States of America
2 University of Stockholm, Physics, Stockholm, Sweden


The key performance parameter of a Time-of-flight (TOF) positron emission tomography (PET) system is the time resolution, which has a direct effect on the TOF signal-to-noise ratio (SNR) gain. TOF gain has been modeled, simulated and measured in past literature, and advanced models have been proposed. In his work we are able to compare models and simulations with experimental data down to 250ps time resolution, using a next generation Siemens SiPM PET/CT prototype scanner. TOF gain is assessed using uniform cylindrical phantoms of different diameters (8cm, 20cm, 37cm). Filtered back projection (FBP) reconstruction was used for the reconstruction, in TOF and nonTOF version.

A SNR TOF gain up to 2.7, equivalent to a TOF sensitivity gain of about 7.5, was measured in this preliminary experiment. In addition, we will present results at variable random fraction, and an assessment of SNR TOF gain for iterative reconstruction will also be discussed.

Keywords: TOF Reconstruction; PET; Time Resolution
Poster panel
(face) ID: 241

Poster Number:

Sampling Lines of Response for the Sensitivity Image in List-Mode OSEM PET Reconstruction (#3878)

W. Dieckmann1, S. Thada1, W. C. Barker1

1 NIH, Clinical Center PET Dept., Bethesda, Maryland, United States of America


Iterative list-mode PET reconstruction requires the pre-calculation of a sensitivity image, in theory by backprojecting all possible detector lines of response (LORs). For systems with a large number of LORs, or because of subject motion compensation, computing the exact sensitivity image is computationally expensive, so sub-sampling the LORs is attractive to reduce the work. The resulting inaccuracies in the sensitivity image propagate to the reconstructed emission image, which motivates looking at different sub-sampling schemes to reduce the errors. We extended the MOLAR reconstruction, originally implemented using uniform random (UNIF) sub-sampling of LORs, by adding a quasi-random (Halton) sequence based sampler and also a previously described Monte Carlo-based (MC) non-uniform sampler. We also evaluated sampling using both techniques simultaneously (Halton-MC). A simple post-smoothing of the sensitivity image can be used to improve variability in exchange for a loss in spatial resolution. We included this method (UNIF-Smooth) in our comparison.  Applying the four sampling methods (UNIF-Smooth, Halton, MC, Halton-MC), we obtain region of interest coefficients of variation (CoV) for replicated reconstructions of a human brain scan and compare them to those obtained with the original uniform subsampling.  We observe approximately two-fold improvement in CoV (Halton) with a 4% increase in time to compute the sensitivity image, and three to six-fold improvement (Halton-MC) with a 163% increased computational investment.

Keywords: Sensitivity Image PET OSEM Listmode
Poster panel
(face) ID: 244

Poster Number:

IteResNet: Iterative Residual Artifacts Network for low-dose CT reconstruction (#3978)

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

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


Radiation exposure and the associated risk of cancer for patients during CT examinations procedure have been the major clinical concern.  The radiation exposure can be reduced effectively via lowering tube current or pulse duration.  However, these strategies of reducing radiation exposure may lead to excessive noise and streak artifacts in the conventional filtered back-projection (FBP) reconstructed images.  To address this issue, in this study, we propose a deep convolutional neural network (ConvNet) approach for low-dose CT reconstruction inspired by the recent residual learning.  However, some of the texture can be corrupted by the severe streaks especially in ultra-low-dose cases, which could be close to prosthesis hampering diagnosis.  To solve this problem, here we propose an iterative residual artifacts ConvNet (IteResNet) approach to improve the reconstruction performance against the conventional residual network.  Specifically, the proposed IteResNet estimates the high-frequency details within the noise and then remove them iteratively, after removing severe streaks in the low-dose CT images, the residual low-frequency details can be removed through the residual network.  Real patient data were used to evaluate the proposed IteResNet, and corresponding experimental results demonstrate that the proposed IteResNet can produce high-quality images that compare favorable in both reducing the noise and artifacts to the previous ConvNet-based approach, including conventional ConvNet approach, and residual ConvNet approach, suggesting that the iterative residual network can improve the CT image quality, especially in ultra-low-dose cases.

Keywords: low-dose CT, convolutional neural network, Residual Artifacts, CT reconstruction
Poster panel
(face) ID: 247

Poster Number:

Visualization of Rib and Diaphragm Motion in an Anaesthetized Mouse by Live Animal Synchrotron Imaging (#1083)

G. K. Aulakh1, W. Kuebler3, B. Singh4, D. Chapman2

1 University of Saskatchewan, SACS/WCVM, Saskatoon, Saskatchewan, Canada
2 University of Saskatchewan, Anatomy & Cell Biology/College of Medicine, Saskatoon, Saskatchewan, Canada
3 Charité - Universitätsmedizin Berlin, Institute of Physiology, Charité - Universitätsmedizin Berlin, Berlin, Berlin, Germany
4 University of Calgary, Faculty of Veterinary Medicine, Calgary, Alberta, Canada


Pulmonary research is challenging because of the lack of visualization of lung airspace by commonly used imaging modalities. Cyclic breathing and motion artifacts due to superimposed ribs and cardiac motion further confound the interpretation of lung imaging data. This is partly responsible for little progress in drug therapy or intervention for many lung diseases such as acute respiratory distress syndrome, asthma, chronic obstructive pulmonary disease, cystic fibrosis, and lung cancer. Synchrotron sources can provide soft tissue contrast not available with conventional technologies by making use of phase contrast. Given the advantage of increased flux and narrow energy-range beams of collimated x-rays, there remains challenges in data analysis due to motion artifacts and lack of supporting image analysis algorithms for the correction of them. Therefore, we performed motion analysis of ribs and diaphragm for potential longitudinal functional lung imaging. The system utilized custom x-ray optics at the Canadian Light Source, for full field mouse imaging at 30 frames per second, on the biomedical beamline. This study presents a motion analysis of ribs and diaphragm of a spontaneously breathing anaesthetized mouse. To our knowledge this is the first temporal image analysis done in a spontaneously breathing individual without the aid of ventilation or extensive transducers. This is a feasibility study for future imaging of animal models of pulmonary diseases.

Keywords: synchrotron x-ray imaging, phase contrast imaging, in-vivo imaging, pulmonary pathobiology, image analysis, motion analysis
Poster panel
(face) ID: 250

Poster Number:

A Novel Non-rigid Registration Method Based on Nonparametric Statistical Deformation Model for Medical Image Analysis (#1260)

Z. Cui1, S. Mahmoodi1, J. Conway2, 3, M. J. Guy2, 4, E. Lewis1, 4, T. Havelock2, 5, M. J. Bennett2, 5

1 University of Southampton, School of Electronics and Computer Science, Southampton, United Kingdom of Great Britain and Northern Ireland
2 Southampton University Hospital NHS Foundation Trust, Southampton NIHR Respiratory Biomedical Research Unit, Southampton, United Kingdom of Great Britain and Northern Ireland
3 University of Southampton, Faculty of Health Sciences, Southampton, United Kingdom of Great Britain and Northern Ireland
4 University Hospitals Southampton NHS Foundation Trust, Department of Medical Physics, Southampton, United Kingdom of Great Britain and Northern Ireland
5 University of Southampton, Faculty of Medicine, Southampton, United Kingdom of Great Britain and Northern Ireland


Non-rigid registration has been widely used in medical image processing for many years. In order to preserve the anatomical topology and perform the registration more realistically and reliably for image guided surgery, methods based on statistical deformation model have been receiving considerable interests. However, the shortcomings in previous work such as the empirically configured weighting parameter for the statistical term lead to a controversial and unrealistic alignment. Therefore, a non-parametric method based on statistical deformation model is proposed here to avoid the discussion of weighting parameter. Our novel method is developed through incorporating the statistical model into two indispensable terms: similarity metric and smoothing regulariser. The advantages of the proposed algorithm in terms of convergence rate and registration accuracy have been proved mathematically in methodology and evaluated numerically in experiments compared with the state of the art method. It has also laid a solid foundation for the development of multi-modality image fusion with prior knowledge in the future.

Keywords: statistical deformation model, non-rigid registration, nonparametric model, nuclear medicine imaging
Poster panel
(face) ID: 253

Poster Number:

PET analysis using features from intensity size zone matrix for group difference between mild cognitive impairment and normal control (#1532)

S. H. Lee2, 3, S. J. Son2, 3, M. Kim2, 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 currently one of the important diseases affecting many patients. Unfortunately, the disease is one of the intractable diseases with no easy cure. Many AD patients transition from an intermediate state known as mild cognitive impairment (MCI) state before the final diagnosis of AD. Many MCI patients finally convert to AD. AD is the final state of the disease, and the treatment options are limited at the final stage. Thus, it is important to detect AD early in the MCI stage where there might be more treatment options available. Positron emission tomography (PET) is representative neuroimaging methods for AD. Many studies have been conducted using PET images based on various image processing techniques to assess AD. Texture analysis is a method uses information and various features derived from the gray-level intensity and position of neighborhood voxels in a region of interest (ROI) to accurate describe a group of patients. We hypothesized that texture analysis can distinguish MCI and normal control (NC) groups well. We collected subjects with FDG and PiB PET images from Alzheimer’s Disease Neuroimaging Initiative (ADNI) database and adopted two features from Intensity size zone matrix (ISZM), which is one of texture analysis methods, to compare MCI and NC groups in both hemispheres of the hippocampus. We aimed to demonstrate the feasibility of two features derived from ISZM as biomarkers to distinguish between MCI and NC groups. As a result, we distinguished between MCI and NC groups using intensity variability (IV) and size zone variability (SZV) from ISZM in both PET images. In particular, the analysis using IV value of FDG PET images showed significant differences between the groups in the hippocampus of both hemispheres. Through this study, we confirmed the possibility that IV and SZV values could discriminate between MCI and NC groups and these features have good potential as an imaging biomarker for early detection of AD.

Keywords: Alzheimer's disease, Mild cognitive impairment, Texture analysis, Intensity size zone matrix, Early detection
Poster panel
(face) ID: 256

Poster Number:

A dictionary learning based approach to remove the Rician noise (#1796)

B. Chen1, 2, F. Xie1, J. Liang2, W. Chen1, 3, W. Zhang1, 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


Rican noise is associated with Magnetic Resonance Imaging (MRI) and removal of the noise is very important because of the wide use of MRI. Great efforts have recently been devoted to develop the corresponding noise-removal algorithms, particularly the development based on the newly-established Total Variation (TV) theorem. However, all the TV-based algorithms depend mainly on the gradient information and have been shown to produce the so called “blocky” artifact, which also deteriorates the image quality and causes image interpretation errors. In order to avoid producing the artifact, this paper presents a new de-noising model based on dictionary learning. An appropriate dictionary is designed by the use of the Kernel Singular Value Decomposition method, resulting in a new Rican noise removal algorithm. Compared with other de-noising algorithms, the presented new algorithm can achieve superior performance, in terms of quantitative measure of the Peak Signal to Noise Ratio, by a series of experiments using different images in the presence of Rician noise.

Keywords: Rician noise, dictionary learning, total variation
Poster panel
(face) ID: 259

Poster Number:

Fusing Motion Estimates for CBCT Reconstruction (#1891)

B. Bier1, M. Unberath1, T. Geimer1, N. Ravikumar1, 2, G. Gold3, R. Fahrig3, 4, A. Maier1

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


Involuntary patient motion decreases image quality of cone-beam CT acquisitions acquired under weight-bearing conditions. Thus, motion compensation is crucial for assessing knee cartilage health in the reconstructed images.

Previous methods for motion compensation used externally attached fiducial markers and 3D/2D registration of prior scans to the projection data. Recently, the combination of cone-beam CT and range-imaging has received increasing attention, as it allows motion compensation based on 3D surface data. We recently proposed an Iterative Closest Point (ICP)-based method to compensate for motion by alignment of surface data. Despite promising results, motion estimates parallel to the rotation axis proved error prone. Yet, motion in this direction is usually estimated very well with projection domain methods, such as Amsterdam Shroud (AS).

In this work, we investigate sensor fusion of ICP and AS based on particle filtering. The ICP motion estimates are improved with a motion surrogate signal obtained by the AS method.

Compared to the ICP-based approach, the proposed fusion yields superior motion estimation accuracy and reconstruction quality, improving the Structural Similarity from 0.96 to 0.99.

Our preliminary results are promising and suggest a high potential of particle filter-based sensor fusion for motion compensation in cone-beam CT. Future work will investigate possibilities to derive fusion parameters automatically from improvements in image quality achieved with a particular estimate. 

Keywords: Motion Estimation, Cone-beam CT, Reconstruction, Sensor Fusion, ICP, Weight-bearing Imaging
Poster panel
(face) ID: 262

Poster Number:

Automated Optic Nerve Head Detection Based on Different Retina Vasculature Segmentation Methods and Mathematical Morphology (#2159)

M. Tavakoli1, P. Kelley1, M. Nazar2, F. Kalantari3, A. Golestaneh4

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
4 Arizona State University, Electrical Engineering, Tempe, Arizona, United States of America


Computer techniques provide physicians assistance at any time and relieve their work load, especially iterative processes like identifying objects of interest such as lesions and anatomical structures from the image. Optic Nerve Head (ONH) detection is considered to be a crucial step in some retinal image analysis algorithms to find other retinal landmarks and lesions, and their corresponding diameters, to use as a length reference to measure objects in the retina.

The objective of this study is to apply three retinal vessel segmentation methods, Laplacian-of-Gaussian edge detector (using second-order spatial differentiation), Canny edge detector (estimating the gradient intensity), and Matched filter edge detector for detection of the ONH either in the normal fundus images or in the presence of retinal lesions like diabetic retinopathy. The steps for the segmentation are as 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 an accuracy view, comparing to manual detection performed by ophthalmologists for retinal images belonging to a test set of 240 images, by using the Laplacian-of-Gaussian vessel segmentation, our automated algorithm finds 180 ONHs in true location for 200 color images in the MESSIDOR database and all images in the DRIVE database. For the Canny vessel segmentation, our automated algorithm finds 164 ONHs in true location for 200 images in the MESSIDOR database and 32 out of 40 images in the DRIVE database. And lastly, using matched filter in the vessel segmentation, our algorithm finds 189 ONHs in true location for 200 images in MESSIDOR database and all images in the DRIVE.

Keywords: retinal image, Optic Nerve Head, Laplacian-of-Gaussian, Canny edge detector, Matched filter, vessel segmentation
Poster panel
(face) ID: 265

Poster Number:

Retinal Surface Prediction in Ophthalmic Spectral-Domain Optical Coherence Tomography (#2317)

J. Wu1

1 University of Tokyo, The Center for Global Communication Strategies, Tokyo, Japan


Examination of the retina in Spectral-domain Optical Coherence Tomography (SD-OCT) images for sight degrading diseases, such as age-related macular degeneration (AMD), relies heavily on the extraction of the retinal layers to detect and quantify fluid filled structures. However, current methods of retinal layer segmentation may perform poorly in the presence of heavy disease pathologies. Presented here is an automated method to predict the retinal layers, based on cubic spline based interpolation, to compensate for potentially missing layer information. Simulated retinal layer segmentation error is used to validate this automated method, showing a high similarity between predicted and ground truth retinal layers of 0.9818±0.0031 based on the Jaccard Index. 

Keywords: optical coherence tomography, sd-oct, retina, segmentation, prediction, retinal disease, interpolation
Poster panel
(face) ID: 268

Poster Number:

Combined PET and MRI Radiomics with Breast Cancer Outcomes (#2504)

S. - Y. Huang1, B. L. Franc1, R. Harnish1, T. Copeland1, V. Arasu1, E. F. Jones1, N. M. Hylton1, Y. Seo1

1 University of California, San Francisco, Radiology and Biomedical Imaging, San Francisco, California, United States of America


We investigated how PET and MRI radiomics features are associated with disease phenotypes and outcomes in breast cancer. In the datasets of 116 breast cancer FDG-PET and dynamic contrast-enhanced (DCE)-MRI, segmented tumor volumes were standardized and normalized prior to extracting 37 radiomics features. The extracted features were further normalized to z-scores, and consensus clustering using multiple clustering algorithms and sampling iterations was performed to associate the feature clusters with disease phenotypes and prognostic factors such as tumor grade, tumor histology, T-stage, N-stage, overall stage, hormone status, recurrence, and recurrence sites. We performed the c2 tests between the cluster labels and the outcome measures, and determined how the radiomics features were associated with the disease characteristics. Our results show that both PET and MRI radiomics features have significant associations with tumor grade, T-stage, and hormone status, and more importantly, only PET radiomics features have significant associations with N-stage and overall stage. Our results, although preliminary, imply that both FDG-PET and DCE-MRI have great potential to radiomics-based disease characterization and prognosis correlation with FDG-PET having a slight edge in our datasets obtained from breast cancer patients.

Keywords: breast cancer, PET-MRI, radiomics
Poster panel
(face) ID: 271

Poster Number:

Parametric LV Model Fitting to Coronary Arteries (#2806)

T. Geimer1, 2, J. Höhn1, M. Unberath1, 2, A. Maier1, 2

1 Friedrich-Alexander University Erlangen-Nürnberg, Pattern Recognition Lab, Department of Computer Science, Erlangen, Bavaria, Germany
2 Friedrich-Alexander University Erlangen-Nürnberg, Erlangen Graduate School in Advanced Optical Technologies, Erlangen, Bavaria, Germany


X-ray angiography is the gold standard in assessing coronary artery diseases. With research focus shifting towards 3D+t applications, heart models that allow for the extraction of functional parameters from the heart motion receive increasing attention. We present an approach to fit a parametric left ventricular heart model originally developed for tagged MRI to the centerlines of coronary arteries reconstructed from rotational coronary angiography. This 3D model fitting process must accommodate the sparse point set conditional to the underlying angiography data. Using a coarse-to-fine optimization based on simulated annealing and ellipsoid pseudo-distances, we achieve a reprojection error of 0.794 mm for the surface points compared to 0.422 mm of the 3D centerline ground truth. Results are promising and form the basis to extend the model to 3D+t in order to monitor radial and longitudinal contraction as well as left ventricular twist over the cardiac cycle.

Keywords: Angiography, Optimization, 3D Surface Model
Poster panel
(face) ID: 274

Poster Number:

Semi-Automatic Algorithm for Breast MRI Lesion Segmentation Using Marker-Controlled Watershed Transformation (#2905)

S. Vesal1, A. Diaz-Pinto2, N. R. Kumar1, S. Ellmann3, A. Davari1, A. Maier1

1 Friedrich-Alexander University Erlangen-Nürnberg, Computer Science Department/ Pattern Recognition Lab, Erlangen, Bavaria, Germany
2 Universitat Politècnica de València, Instituto de Investigación e Innovación en Bioingeniería, Valencia, Spain
3 Universitätsklinikum Erlangen, Radiologisches Institut, Erlangen, Germany


Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is an effective tool for determining breast lesion in women. An accurate and precise segmentation of the breast lesion in a Computer Aided Detection (CAD) system is a crucial step in evaluating tumor volume and in the quantification of tumor characteristics. However, this is a challenging task, since breast lesions have sophisticated shape, topological structure, and variation in the intensity distribution. In this paper, we proposed a novel Marker-controlled Watershed method which uses brightest pixels as markers to overcome this challenge and accurately segment the breast MRI lesions through predetermined Region of Interest (ROI) by an expert. The proposed approach was evaluated on 106 lesion cases which include 59 malignant and 47 benign. The segmentation results compared with ground truth through Jaccard and Dice Coefficient metrics, in which the method achieved an overall Jaccard of 0.7808 ± 0.1729 and Dice Coefficient of 0.6704 ± 0.2167. The results illustrate that the proposed method significantly improve segmentation accuracy and it will help further in improving classification performance between malignant and radiologically suspicious benign lesions. 

Keywords: Breast MRI, Lesion Segmentation, Marker-Controlled Watershed, Magnetic Resonance Imaging
Poster panel
(face) ID: 277

Poster Number:

Unsupervised Learning in PET Radiomics (#3185)

G. Liu1, D. Mitra1, S. - Y. Huang2, Y. Seo2, B. Franc2

1 Florida Institute of Technology, School of Computing, Melbourne, Florida, United States of America
2 University of California San Francisco, Radiology Department, San Francisco, California, United States of America


In this project we have developed a pipeline for large scale radiomics analyses. We are particularly interested in unsupervised learning to cluster patients using the image features in order to associate such clusters with the disease characteristics. This is performed in order to understand usefulness of different features sets against their capability to predict disease outcomes. We have run a pilot study with 116 breast cancer patients’ PET dataset and present our findings here. We have used two different sets of image features for clustering patients and aligned them with three different types of disease outcomes: recurrence of tumors, histology results and tumor stages. We observe some predictability of these outcome types from clustering results and possibly a unknown radiomics characteristic for which we have no known disease outcome type. The objective of radiomics is to develop reliable classifier that would be able to predict patient outcomes given a patient’s radiology image.

Our work is a step in that direction as well as in finding unknown radiomics types. We provide a future directions toward which our work will be guided.

Keywords: Radiomics, Workflow, Unsupervised Clustering, PET, Breast Cancer
Poster panel
(face) ID: 280

Poster Number:

Motion Compensated Compressed Sensing Reconstruction in Cardiac Cine MRI (#3829)

A. Tolouee1, J. Alirezaie1, P. Babyn2

1 Ryerson University, Electrical and Computer Engineering, Toronto, Ontario, Canada
2 University of Saskatoon Health Region, Medical Imaging, Saskatoon, Saskatchewan, Canada


Magnetic Resonance Imaging (MRI) provides unique advantages over other well-known imaging technologies such as X-ray Computed Tomography. Unfortunately the adoption of MRI in clinical studies has been slow until recently due to the slow rate of data acquisition and high expense of the equipment. This problem has been partially remedied with the introduction of compressed sensing in which the image can be reconstructed from fewer samples reducing the total scan time. The breakthrough of compressed sensing opened the door for MRI to new applications such as cardiac imaging which was not possible with slower speeds. More recently, matrix recovery schemes using the linear dependencies of pixel time profiles using low rank image priors have been proposed. While these strategies show successful recovery when the inter frame motion is negligible, the main challenge is the sensitivity of these techniques to large inter frame motion.

In this work, we introduce a novel formulation for the joint estimation of the deformation and the images in dynamic cardiac cine MR imaging. Low rank plus sparse decomposition reformulates decomposition as an optimization problem to recover the sparse and low rank components of the input data. Our hypothesis is that low rank plus sparse decomposition coupled with a registration algorithm provides accurate registration of dynamic time series in a broad range of organs and for various breathing protocols. We used a variable splitting frame work with continuation to minimize the objective function, and thus derive the deformation and the dynamic images. The validation of the proposed algorithm using numerical phantom and in-vivo cine MRI data show the feasibility in precisely recovering cardiac MRI data from extensively under-sampled data.

Keywords: Compressed sensing, Motion compensation, Low rank plus sparse decomposition, Dynamic cardiac MRI
Poster panel
(face) ID: 283

Poster Number:

The influence of sampling schedule for image-derived input function and Patlak plot estimation in whole body PET studies (#1188)

M. Naganawa1, J. - D. Gallezot1, V. Shah2, A. M. Smith2, R. E. Carson1

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


Dynamic whole body positron emission tomography (PET) imaging has recently been proposed for 18F-FDG. This imaging employs a multi-bed acquisition protocol. The limited number of dynamic frames per bed position, compared to a single-bed position protocol, affects the estimation of the kinetic parameters and the image-derived input function (IDIF). The aim of this study was to investigate the effect of different sampling schedule designs for both IDIF and Patlak plot estimation. We conducted two simulations. In simulation 1, tumor and normal liver time-activity curves (TACs) were simulated with noise using a noise-free input function by changing the number of passes (3 to 6). Standard (2-parameter) and generalized (3-parameter) Patlak analyses (sPatlak, gPatlak) were applied to estimate the tracer uptake rate Ki with t*=30 min. In simulation 2, noisy IDIFs were simulated using 6-min cardiac scans, to capture the peak of the input function, and various numbers of bed passes (2 to 42 passes) for the rest of the scan. The tail of the IDIF was smoothed by a two-exponential function. The 2 Patlak models were applied to the noise-free tumor and normal liver TACs and the obtained IDIFs. Simulation 1 showed that 3 passes after t* are enough to generate tumor and normal liver TACs for Ki estimation. Simulation 2 showed that at least 3 passes are required to generate the IDIF. Overall, there were negligible effects of number of passes on the bias of Ki in both Patlak analyses. The standard deviation of Ki, which increased with higher image noise level, was also not affected by the number of passes.

Keywords: Whole body positron emission tomography, IDIF, Sampling schedule, Patlak analysis
Poster panel
(face) ID: 286

Poster Number:

Distributed non-invasive system for measuring the arterial input function in PET (#1881)

C. Pereira1, A. Abrunhosa2, A. Cruz2, N. Ferreira1, 2, A. Moreira2, F. Oliveira1, 2, M. Patrício1, F. Caramelo1

1 University of Coimbra, Institute for Biomedical Imaging and Life Sciences - Faculty of Medicine, Coimbra, Portugal
2 University of Coimbra, Institute for Nuclear Sciences Applied to Health – ICNAS, Coimbra, Portugal


Usually, PET images are evaluated statically and qualitatively, seeking for regions with abnormally high or low uptake. Quantification allows for a more thorough understanding of the physiological process, benefiting not only diagnoses but also allowing for longitudinal studies and inter subject normalization. Quantification can be performed using compartmental analysis that requires the knowledge of concentration of the radiotracer present in the arterial blood at each instant - the arterial input function (AIF). Measuring the AIF is usually an invasive and challenging process, making this a real obstacle to performing in patients and in healthy volunteers. In this work, a prototype for measuring the AIF non-invasively is presented. The system is composed by several gamma detectors that are placed near superficial arteries. The main idea is to obtain signals from arteries and their surroundings using different locations and combining the signals in order to accurately estimate the AIF. We assembled proportional radiation detectors comprising essentially a bismuth germanium oxide (BGO) crystal and a photodiode. The first prototypes are being tested for energy resolution, efficiency, linearity and adaptation to the patient body during real PET exams. Energy resolution was estimated to be 14% FWHM (Full Width at Half Maximum), efficiency 1.4% and proportionality between response and activity of radioactivity sources was observed. The detectors were placed on the wrist and close to the ankle in two patients that underwent PET scans. Measures from the wrist are compatible to the AIF but movements of the patients show that shielding the detectors is necessary to prevent contamination from other sources other than the radial artery. Regarding the ankle detector, it is difficult to have it fixed during the exam, which has yet to be tackled. Results so far permit to conclude that the system exhibits suitable characteristics for measuring the AIF for PET quantification.

Keywords: PET, Quantification, Arterial input function
Poster panel
(face) ID: 289

Poster Number:

Quantification of TSPO expression with [18F]DPA-714 PET using an image derived input function (#2974)

C. Wimberley1, L. Ngyuen1, Y. Fontyn1, R. Boisgard1, V. Bouilleret1, I. Buvat1

1 Inserm/CEA/University Paris Sud, Laboratoire Imagerie Moleculaire In Vivo (IMIV), Orsay, France


The translocator protein (TSPO) is of great interest as a biomarker for the study of diseases that have a component of neuroinflammation, such as epilepsy and Alzheimer’s disease. TSPO PET imaging can follow neuroinflammation associated with neurodenenerative disorders using [18F]-DPA-714, but there are quantification challenges due to the small size of the brain structures in the mouse causing partial volume effects, the difficulty of arterial sampling in mice, especially in longitudinal studies, and to the absence of a reference region for some pathologies and disease models. The aim of this study was to investigate the use of an image derived input function (IDIF) extracted using factor analysis (FA) for estimation of binding parameters.

The model was an injection of kainic acid (KA) in the right dorsal hippocampus of adult male C57/Bl6 mice (n=4). A dynamic [18F]DPA-714 PET/CT (60 min) was performed at 2 time points: i) baseline scan before & ii) one month after KA injection (KA1m). FA was applied to all images using 4 factors to extract the IDIF and used in the Logan plot to estimate the total volume of distribution (VT). The VT was compared to the estimated binding potential (BPND) from the Logan Reference model using the averaged striatum as reference region which was previously shown to have low TSPO expression compared to other regions in this model.

The extracted IDIFs had a peak that was highly correlated with the injected dose (r2=0.82), and there were strong regional correlations between the Logan IDIF estimates and the Logan reference estimates for baseline (r2=0.9) and KA1m datasets (r2=0.99).

In conclusion, our results suggest that we can quantify TSPO PET in the mouse brain when there is no reference region available by extracting an IDIF using FA. This will be useful for longitudinal studies of neuroinflammation with high levels of TSPO expression throughout the whole brain, as well as in pathologies where the pattern of neuroinflammation is unknown.

Keywords: TSPO, neuroinflammation, quantification, kinetic modelling, image derived input function, factor analysis
Poster panel
(face) ID: 292

Poster Number:

Evaluation of 18 F-FET-PET and perfusion MRI texture features in brain tumor grades (#3277)

S. Assili1, L. Caldeira1, P. Lehman1, A. Shahbazi2, C. P. Filss5, N. J. Shah1, 3, K. J. Langen5, 4

1 Forschungszentrum Jülich GmbH, Medical Imaging Physics Department, Institute of Neuroscience and Medicine (INM-4), Jülich, North Rhine-Westphalia, Germany
2 University of Notre Dame, Department of Computer Science and Engineering, South Bend, United States of America
3 RWTH Aachen University, Department of Neurology, Aachen, Germany
4 University of Aachen, Department of Nuclear Medicine, Aachen, Germany
5 Forschungszentrum Jülich GmbH, Institute of Neuroscience and Medicine (INM-4), Jülich, Germany


The aim of this study is evaluating PET using the 18F-labeled amino acid O-(2-18F-fluoroethyl)-L-tyrosine (FET-PET) and perfusion-weighted MRI (PWI) texture features for the differentiation between high- and low-grade gliomas. Twenty-seven patients with gliomas underwent 18F-FET-PET and perfusion MR Imaging. FET-PET images and the perfusion maps were co-registered with perfusion images. The tumor’s volume of interest (VOI) was delineated by a threshold-based method. In the next step, texture feature analysis was done on the created VOIs based on the gray levels. For the evaluation of the extracted texture parameters, principal component analysis was applied to identify the best features for the classification between low- and high-grade gliomas by calculating the score and coefficient of each feature for further use. With PCA we precisely calculated the score and weight of each feature and showed how these top score features are correlated and enough to making a decision on machine learning feature extraction/selection. We prepared an MLDB (machine learning database) feature extraction based on these top scored feature to train and test further data from this field.

Keywords: FET-PET, DSC-MRI, Brain tumors, classification, grading
Poster panel
(face) ID: 295

Poster Number:

Modeling and Correlation Between CT Calcification and Glucose Metabolism in Atherosclerosis (#3549)

M. S. Al-enezi1, 2, F. A. A. Slimani1, A. Khalil3, M. Bentourkia1

1 Université de Sherbrooke, Nuclear Medicine and Radiobiology, Sherbrooke, Québec, Canada
2 University of Hail, College of Applied Medical Science, Hail, Saudi Arabia
3 Université de Sherbrooke, Medicine, Sherbrooke, Québec, Canada


Atherosclerosis is an inflammatory disease characterized by the accumulation of lipids in an arterial wall, leading to accumulation of foam cells and lipids contributing to plaque formation. The plaque narrows the arteries, hardens and gets calcified, and it can rupture and provokes cerebral or cardiac vascular accidents. The plaque prone to rupture is always accompanied by inflammation where the cells are active. The inflammation can be measured with 18F-FDG-PET. Several imaging techniques have been proposed like MRI, CT, ultrasound and angiography, as they appropriately locate the plaque but they lack the detection of its vulnerability to rupture. Also, CT imaging can detect the calcification and hardening in the plaque. In this work, we report a quantitative analysis based on the calcification as measured with CT imaging and the glucose metabolism as measured with 18-FDG in normal volunteers and subjects with chest angina. The CT and PET images were analyzed slice by slice to account for different artery calcifications. The volume of the calcified plaque was found to vary from occupying some pixels in the artery transaxial slice to the whole circumference of the artery. The results show that even the arteries use glucose as demonstrated by the decomposition of the artery images with factor analysis, there were no correlation between 18F-FDG uptake and calcification extent or intensity in CT images. By clustering the CT artery images depending on their intensity in Hounsfield units, these clusters were not preserved in the PET artery images. These results will be further assessed with compartmental analysis in a future work.

Keywords: Atherosclerosis; Calcification; Glucose metabolism; Kinetic modeling; Clustering; arteries.
Poster panel
(face) ID: 298

Poster Number:

Proton Imaging with Novel Glass Calorimeters (#1013)

U. Akgun1

1 Coe College, Physics Department, Cedar Rapids, Iowa, United States of America


In recent years, proton therapy has achieved remarkable precision in dose delivery to cancerous cells while avoiding healthy tissue. However, to utilize this high precision treatment, a greater accuracy in positioning the patient is needed. A 3% range of uncertainty exists in the current practice of proton therapy due to the conversion of units for x-rays to stopping power. This study focuses on the use of protons instead, which would eliminate this conversion entirely, determine a more accurate stopping power, and lessen exposure to the patient.

There have been two separate detectors designed with unique glasses. The first detector design utilizes the scintillating high density glass bars, the other semiconducting glass fibers. The unique geometry of these detectors allows for the measurement of both the position and residual energy of pencil beams of protons, eliminating the need for trackers in the system. The simplicity, compactness, and efficiency was the major objectives in these models for the purpose of presenting a novel imaging technique that is both precise and practical for a clinical setting. The same devices can also be suitable to detect the prompt gammas (2-15 MeV) for in vivo tracking of the Bragg peaks.

This report summarizes the optical and electrical detector designs, resolution of the imager, data collection and the image reconstruction methods, as well as the properties of the materials specifically developed for these systems. Images created via Geant4 simulations, and the approaches using prompt gammas during the proton therapy will be reported.

Keywords: proton imaging, in-situ imaging, hadron therapy, novel glasses
Poster panel
(face) ID: 301

Poster Number:

Development Of The Bragg-Peak Position Monitoring System In Particle Therapy By Using e+ e- Pair Production Events (#1663)

S. Kimura1, Y. Emoto1, K. Fujihara1, H. Ito1, H. Kawai1, 2, A. Kobayashi1, H. Matsunaga2, T. Mizuno1, T. Nakamura1, T. Tanaka2, T. Yuzawa2

1 Chiba University, Graduate School of Science, Chiba, Chiba, Japan
2 Chiba University, Faculty of Science, Chiba, Chiba, Japan


In particle therapy, it is important to monitor the Bragg-peak position. We are developing the system to monitor the Bragg-peak position which can measure pair production events occurred in the detector by gamma rays from irradiation points. The momentum direction of the gamma ray can be determined by measuring passing points and energy of e+ and e- generated by pair production. This system has 5 parts. The first is the conversion part. This part consists of several layers. Each layer is composed of a La-GPS ((Gd0.75La0.24Ce0.01)2Si2O7) scintillator plate and wavelength-shifting fibre (WLSF) sheets. The scintillator plate is sandwiched between sheets, where the directions of the sheets are in orthogonally x and y directions. In this part, gamma rays are converted to ee- pairs and the position where the conversion occured is determined. The second is the tracking part. This part consists of 2 layers of scintillating fibre tracker. Each layer has 6 scintillating fibre sheets for x, x', u, u', v, and v'. The third is the energy measurement part. It measures the energy of e+ and e- by scintillator array and Silicon Photomultipliers. The fourth is the veto counter for bremsstrahlung gamma rays from e+ and e-. The fifth is the beam monitor. It was confirmed by GEANT4 Monte Carlo Simulation Code that the gamma rays of 10 MeV or more, where the pair production event dominates, are generated intensively from the Bragg-peak position. By experiment, the number of photoelectrons of La-GPS with a WLSF (B-3(300)MJ, Kuraray) sheet and scintillating fibre (SCSF-78, Kuraray) when charged particle passed was measured as 9.7 and 7.6 respectively. Position resolution of a WLSF sheet and a scintillating fibre sheet are being measured. In addition, we are simulating angular resolution of this system.

Keywords: particle therapy, pair production
Poster panel
(face) ID: 304

Poster Number:

Feasibility Study of All-digital PET Monitoring Proton Therapy (#2439)

M. Gao1, C. - M. Kao4, H. - H. Chen3, X. Lyu1, J. - H. Hong3, Y. Hua1, T. - C. Yen3, Q. Xie1, 2

1 Huazhong University of Science and Technology, Wuhan, China
2 Wuhan National Laboratory for Optoelectronics, Wuhan, China
3 Chang Gung Memorial Hospital, Taoyuan, China
4 The University of Chicago, Chicago, United States of America


Positron Emission Tomography (PET) is one of the most promising non-invasive in vivo imaging techniques for the proton therapy verification. In this paper we investigate the feasibility of using an all-digital PET technology platform for in-beam or in-room imaging for proton therapy. We performed a series of experiments using phantom and biological tissue. Our initial results demonstrate the PET technology platform is promising for verification of proton therapy.

Keywords: all-digital PET, Proton Monitoring
Poster panel
(face) ID: 307

Poster Number:

Development of a low-energy X-ray camera for the imaging of secondary electron bremsstrahlung X-ray emitted during proton irradiation for range estimation (#2661)

K. Ando1, A. Kajita1, S. Yamamoto1, M. Yamaguchi2, Y. Nagao2, T. Toshito3, J. Kataoka4, N. Kawachi2

1 Nagoya University, Graduate School of Medicine, Nagoya, Japan
2 National Institutes for Quantum and Radiological Science and Technology (QST), Takasaki, Gunma, Japan
3 Nagoya City West Medical Center, Nagoya Proton Therapy Center, Nagoya, Japan
4 Waseda University, Research Institute for Science and Engineering, Tokyo, Japan


Imaging of secondary electron bremsstrahlung X-ray emitted during proton irradiation is a possible method for measurement of the proton beam distribution in phantom. However it is not clear that the method is used for range estimation of protons. For this purpose, we developed a low-energy X-ray camera and imaging was conducted for the bremsstrahlung X-ray produced during irradiation of proton beams. We used a 20 mm x 20 mm x 1 mm thick finely grooved GAGG scintillator. The scintillator was optically coupled to a 1-inch square high quantum efficiency (HQE) type position sensitive photomultiplier tube (PSPMT) to form an imaging detector. The imaging detector was encased in a 2 cm thick tungsten container and a pinhole collimator was attached to its camera head. After performance of the camera was evaluated, secondary electron bremsstrahlung X-ray imaging was conducted during irradiations of the proton beams for three different proton energies and the results were compared with calculated value. The spatial resolution and sensitivity of the developed X-ray camera with 1.5mm diameter pinhole collimator were estimated to be 32 mm FWHM and 5.2 x 10-7 for 32~37 keV X-ray photons at 100 cm from the collimator surface, respectively. We could image the proton beam tracks by measuring the secondary electron bremsstrahlung X-ray during irradiation of the proton beams and the ranges for different proton energies could be estimated from the images. We confirmed that the imaging of the secondary electron bremsstrahlung X-ray emitted during proton irradiation with the developed X-ray camera has a potential to be a new tool for proton range estimations.

Keywords: proton therapy, range estimation, X-ray camera, secondary electron bremsstrahlung
Poster panel
(face) ID: 310

Poster Number:

Proton Computed Tomography: pre-clinical tests with an anthropomorphic head phantom (#2990)

M. Bruzzi1, 2, C. Civinini2, M. Scaringella1, 2, D. Bonanno3, M. Brianzi2, M. Carpinelli5, 4, G. Cuttone4, G. Maccioni6, N. Randazzo3, M. Rovituso7, V. Sipala5, 4, C. Talamonti1, 2, F. Tommasino7, 8

1 Universita' di Firenze, Sesto Fiorentino, Italy
2 INFN Firenze, Sesto Fiorentino, Italy
3 INFN Catania, Catania, Italy
4 LNS INFN, Catania, Italy
5 Universita' di Sassari, Sassari, Italy
6 INFN Cagliari, Cagliari, Italy
7 TIFPA, Trento, Italy
8 Universita' di Trento, Trento, Italy


A large-view proton Computed Tomography (pCT) apparatus based on a silicon tracker and a YAG:Ce calorimeter has been developed. Tests with a 220MeV proton beam at the “Trento Proton Therapy Center” (Trento, Italy) aiming at collecting data for reconstructing radiography and tomography images have been carried out. Algebraic iterative reconstruction methods (ART), together with the most likely path calculation, have been used to extract density and spatial resolutions. A pre-clinical test of this pCT system has been performed with an anthropomorphic head phantom. Certified phantoms have been tested to investigate spatial and density resolutions of this technique in a realistic scenario.

Keywords: Instrumentation for hadron therapy; silicon microstrip tracker; scintillating calorimeter; proton computed tomography
Poster panel
(face) ID: 313

Poster Number:

Development of prompt gamma rays imaging detector using LaBr3(Ce) scintillator for boron neutron capture therapy (#3192)

H. Tanaka1, T. Takata1, K. Akabori2, Y. Sakurai1, K. Okaszaki1, T. Watanabe1, M. Suzuki1, S. - I. Masunaga1, K. Ono1

1 Kyoto university Research Reactor institute, Department of Radiation Life Science and Radiation Medical Science, Osaka, Japan
2 Sumitomo Heavy Insustiries,Ltd, Tokyo, Japan


At Kyoto University Research Reactor Institute, clinical studies of over 500 of Boron Neutron Capture Therapy (BNCT) have been performed using a research reactor. To improve the quality of treatment, it is necessary to perform the detection of boron concentration during the treatment as quality control. Generally, boron concentration is measured by the method of prompt gamma ray analysis. However, the information of boron concentration is not able to be obtained by this method during the treatment. It is desired to measure the boron concentration in real-time. Therefore, we propose the method using prompt gamma rays with the energy of 478 keV emitted by the reaction between boron-10 and thermal neutron. There are some agendas, which should be overcome in order to apply a prompt gamma rays imaging detector in BNCT irradiation field. Most important agenda is the discrimination from 511 keV annihilation gamma ray as back-ground.

The energy resolution of less than 6.5 % FWHM at 511 keV is needed in order to discriminate 511 keV gamma ray. We constructed a prompt gamma rays imaging detector consisting of LaBr3(Ce) scintillator, 8 x 8 MPPC arrays, 64 channel channels amplifier and ADC. Energy resolution of each channel was evaluated using Na-22 gamma-ray source. Average energy resolution FWHM was 6.6 +- 0.48 %. Furthermore, two dimensional 511 keV gamma ray distribution was demonstrably obtained.

Keywords: boron neutron capture therapy, SPECT, prompt gamma, real-tme, BNCT
Poster panel
(face) ID: 316

Poster Number:

Potential of Novel Optical Fibers for Proton Therapy Dosimetry (#3444)

C. Hoehr1, A. Morana2, O. Duhamel4, B. Capoen3, M. Trinczek1, C. Duzenli5, P. Paillet4, H. El Hamzaoui3, M. Bouazaoui3, G. Bouwmans3, Y. Ouerdane2, A. Boukenter2, S. Girard2

1 TRIUMF, Vancouver, British Columbia, Canada
2 Univ-Lyon, Saint-Etienne, France
3 Univ-Lille, Lille, France
4 CEA, Arpajon, France
5 British Columbia Cancer Agency, Vancouver, British Columbia, Canada


Novel optical fiber materials with excellent light output and radiation hardness have been tested for their use as detectors for dosimetry at the Proton Therapy facility of TRIUMF, Vancouver, Canada. The fibers are made by the sol-gel technique out of amorphous silica (a-SiO2) doped with either Copper (Cu) or Cerium (Ce) ions. Both fibers show great promise for the characterization of the small fields typical in the treatment of ocular tumors due to their small diameter of only ~ 500 micrometers. In addition to spatial resolution, a detector for radiotherapy dosimetry also needs to be linear in dose rate and energy independent. The dose-rate dependence was found to be linear, as desired. Both fibers do show however a dependence on the proton energy, especially at lower energy which corresponds to a higher linear energy transfer. Work is underway to overcome this effect.

Keywords: inorganic fiber, dosimetry, proton therapy
Poster panel
(face) ID: 319

Poster Number:

Verification of Mulitleaf Collimator Position Reconstruction for Intensity Modulated Radiotherapy using the Upstream TRAPS detector (#3676)

R. F. Page1

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

On behalf of TRAPS-Collaboration


In Intensity Modulated Radiotherapy (IMRT) there is a move to more dynamic and intense therapies. These are complex and require intricate fields that are formed using Multileaf Collimators (MLC). It is becoming accepted that to monitor these therapies a realtime beam monitor and verification system is needed. One possibility is to place a very thin detector upstream of the patient. TRAPS is an upstream detector built using a Monolithic Active Pixel Sensor and the first with hardware and software specifically designed for upstream detection. A major challenge for any upstream device is measuring the MLC positions. To measure the performance of the TRAPS an IMRT prostate plan was delivered on an Elekta linac and captured with both TRAPS and an EPID. The MLC positions for the treatment fields where reconstructed using both devices and projected to the isocentre where they were compared. The reconstruction method first generates a maximum gradient contour that is caused by MLCs in the beam and then employs a peak finder to determine the position. The precision on the MLC reconstruction for TRAPS was 0.07+/-0.04mm, whilst for the EPID it was 0.25mm limited by the pitch of the detector. The agreement between both methods was excellent showing the expected linear correlation and verifying that TRAPS can reconstruct MLC positions at the head of the linac. To test the accuracy of the TRAPS the difference between the EPID and TRAPS measurements were calculated. The expected result was a normal distribution centred at 0 with a width of 0.26mm. A fit to the data yielded a mean of -0.03+/-0.02mm and a width of 0.254+/-0.004mm. This result shows that with the EPID as a reference the TRAPS has an accuracy of 0.254mm. To test the TRAPS more accurately would require a more precise reference. Even with this limitation the result is below the 1mm requirement for verification. These results demonstrate that the TRAPS detector has overcome the challenges of upstream MLC reconstruction.

Keywords: IMRT MAPS VMAT MLC EPID Upstream Detectors
Poster panel
(face) ID: 322

Poster Number:

A Highly Accelerated Parallel Multi-GPU based Reconstruction Algorithm for Generating Accurate Relative Stopping Powers (#3947)

P. Karbasi1, R. Cai1, B. Schultze1, H. Nguyen1, J. Reed1, P. Hall1, V. Giacometti2, 3, V. Bashkirov3, R. Johnson4, N. Karonis5, 6, J. S. Olafsen7, C. Ordoñez8, K. E. Schubert1, 3, R. W. Schulte3

1 Baylor University, Electrical and Computer Engineering, Waco, Texas, United States of America
2 University of Wollongong, Medical Radiation Physics, Wollongong, Australia
3 Loma Linda University, Department of Basic Sciences. Division of Biomedical Engineering Sciences, Loma Linda, California, United States of America
4 University of California, Santa Cruz, Santa Cruz Institute for Particle Physics, Santa Cruz, California, United States of America
5 Northern Illinois University, Computer Science Department, DeKalb, Illinois, United States of America
6 Argonne National Laboratory, Mathematics and Computer Science Division, Argonne, Illinois, United States of America
7 Baylor University, Department of Physics, Waco, Texas, United States of America
8 Northern Illinois University, Center for Research Computing and Data, DeKalb, Illinois, United States of America


Low-dose Proton Computed Tomography (pCT) is an evolving imaging modality that is used in proton therapy planning which addresses the range uncertainty problem. The goal of pCT is generating a 3D map of Relative Stopping Power (RSP) measurements with high accuracy within clinically required time frames. Generating accurate RSP values within the shortest amount of time is considered a key goal when developing a pCT software. The existing pCT softwares have successfully met this time frame and even exceeded this time goal, but  require  clusters with  hundreds of processors.

This paper describes a novel reconstruction technique using two Graphics Processing Unit (GPU) devices. The proposed reconstruction technique is tested on both simulated and experimental datasets and on three different systems namely Nvidia K40 and P100 GPUs from IBM and Cray. The experimental results demonstrate that our proposed reconstruction method meets both the timing and accuracy with the benefit of having reasonable cost and efficient use of power.

Keywords: Computed tomography, Proton computed tomography, Image reconstruction, Iterative reconstruction, Relative stopping power, RSP, PCT, GPU
Poster panel
(face) ID: 325

Poster Number:

A layered collimator for radiographic imaging (#1068)

I. Kusakari1, K. Ogawa1, K. Usui2

1 Hosei University, Department of Applied Informatics, Tokyo, Japan
2 Juntendo University, School of Medicine, Tokyo, Japan


The aim of our research is to develop a new layered collimator for x- and gamma-ray imaging. In the case of x-rayimaging, a collimation of x-rays is performed with thin lead foils and filling-up materials between the foils. Filling-up materials with a low attenuation coefficient are used for forming the collimator. Even though the attenuation coefficient is low, the filling-up materials reduce the number of primary photons passing through the collimator hole. The idea in this paper is to use a thin tungsten plate with a thickness of 0.1 mm, and insert this plate to a collimator frame with a guide. The tungsten plate is somewhat hard despite being thin, and so we can make a one dimensional collimator easily. Moreover, we can easily design a parallel-hole collimator or cone beam collimator by changing the shape of the guides. And if we stack these one dimensional collimators alternatively in x- and y-directions, we can eliminate scattered photons without losing the primary photons that enter the collimator holes. The performance of the proposed layered collimator was confirmed by Monte Carlo simulationswith several phantoms.

Keywords: collimator, Monte Carlo simulation, radiographic imaging
Poster panel
(face) ID: 328

Poster Number:

An Interactive Tool to Compute Customized Crystal Reflectance for Optical Monte Carlo Simulation in GATE V8.0/GEANT4 (#1875)

M. Stockhoff1, S. Cherry1, E. Roncali1

1 UC Davis, BME, Davis, California, United States of America


An important parameter in the optimization of positron emission tomography (PET) is the collection efficiency of optical photons in γ-ray scintillation detectors. Effects of scintillator material, size and finish can be studied with common simulation software such as GATE or GEANT4. Simulating the surface finish is supported by several models, yet remains a challenging task especially for non-polished surfaces. Our model for optical transport at the crystal boundaries, the LUT Davis model, has recently been made available in GATE V8.0. It is based on 3D data of surfaces measured with atomic force microscopy (AFM). In our custom Monte Carlo simulation code individual photons are tracked to the surface extracted from the AFM image. Reflection probabilities and directions are then predicted using Fresnel equations and Snell’s law. They are saved in look-up-tables (LUTs) that are accessed by GATE/GEANT4 when a photon hits an optical surface defined with the LUT Davis model. Currently the database includes eight LUTs for rough and polished lutetium oxyorthosilicate (LSO) surfaces that are combined with either a pure Lambertian reflector (Teflon) or a specular reflector (ESR, air and grease coupled, respectively). The LUTs were generated from our specific crystals and reflector wrapping technique; however the degree of roughness is strongly dependent on the manufacturer and is difficult to standardize. Furthermore, the application of reflector material to the crystal surface is difficult to reproduce between research facilities. Thus, we have developed a graphical user interface (GUI) to compute custom LUTs. The user can provide surface data measured at high spatial resolution (~ 100 nm), such as AFM or scanning electron microscopy (SEM) and extend the existing LUT database in GATE V8.0. This tool will enable advanced optical simulation with exact surface definition for studies of timing resolution, spatial resolution, and depth of interaction (DOI) encoding.

Keywords: GUI, GATE, GEANT4, Optical Simulation, Optical Modeling, Scintillation detectors
Poster panel
(face) ID: 331

Poster Number:

Design and Implementation of a Graphical User Interface for Dosimetry Calculation in Radiotherapy (#2079)

F. A. A. Slimani1, M. Hamdi2, M. Bentourkia1

1 Université de Sherbrooke, Nuclear Medicine and Radiobiology, Sherbrooke, Québec, Canada
2 Université de Mostaganem, Département de Génie Électrique, Mostaganem, Algeria


There are several computer programs or combination of programs for radiation tracking and other information in tissues. Among these are GEANT4 programs provided as classes that can be incorporated in C++ codes to achieve different tasks in radiation interactions with matter. GEANT4 made the physics easier but the programming remains a tedious task even for physicists familiar with computer programming, while it is obviously out of reach of biological and clinical researchers. The aim of the present work was to report the design and development of a Graphical User Interface (GUI) for absorbed dose calculation and for particle tracking in humans, small animals and phantoms. The GUI is based on the open source GEANT4 for the physics of particle interactions, on the QT cross-platform application for combining programming commands and for display, on GEANT4-GATE for the absorbed dose calculation, and finally on C++ programming. The calculation of the absorbed dose can be performed based on 3D CT images in DICOM format, from images of phantoms or from solid volumes that can be made from any pure or composite material to be specified by its molecular formulas. The GUI has several menus relative to the emitting source which can have different shapes, positions, energy as mono- or poly-energy such as X-ray spectra; the types of particles and particle interactions; energy deposition and absorbed dose; and the output results as 3D images, matrices or histograms. As application results, a lung tumor in a mouse irradiated with seven energy beams at different angles and the absorbed dose was assessed in the tumor and in 8 other tissues mostly not traversed by any radiation beam. Other results were obtained in a human thorax and in solids. In conclusion, the GUI we developed can be easily used by any researcher without the need to be familiar with computer programming, and it will be freely proposed as an open source.

Keywords: Monte Carlo simulation; Dosimetry; Radiotherapy; Geant4; Gate; Graphical user interface.
Poster panel
(face) ID: 334

Poster Number:

Investigating Angular Sampling in Multi-Pinhole AdaptiSPECT-C with XCAT Phantoms (#2154)

K. S. Kalluri1, A. Könik1, J. M. Mukherjee1, J. C. Goding1, S. Banerjee1, Y. He1, J. M. Mukherjee1, J. Dey2, G. Zubal3, L. R. Furenlid3, M. A. King1

3 UNIVERSITY OF ARIZONA, DEPT OF OPTICAL SCIENCES, Tucson, Arizona, United States of America


Investigators at the University of Massachusetts Medical School and the University of Arizona are designing a novel multi-detector, multi-pinhole modular dedicated Brain-SPECT imaging system called AdaptiSPECT-C. Goals of the AdaptiSPECT-C imaging system are to achieve an improvement in sensitivity and resolution, and address static and dynamic imaging needs in comparison to the clinically available imaging systems. The system will be capable of imaging patients with a wide variety of radio-pharmaceuticals, automatically adapt its imaging characteristics (aperture size and number of pinholes open for imaging) in response to the imaging tasks and individual patients. In its preliminary design phase, AdaptiSPECT-C is being modeled as a stationary system with a total of 23 pinhole-detector modules arranged in 3 rings forming a hemispherical volume. The patient bed is positioned such that the patient’s head (brain) is positioned within the AdaptiSPECT-C imaging system and completely sampled by the multiple pinholes. The focus this study is to investigate the design parameters of the system geometry using computer simulations. The results of this study will determine if sufficient angular sampling of the XCAT brain is achieved for reconstruction using the current design, and if not then allow investigation of alternative designs. We have simulated projection images of multiple phantoms using GATE and an analytical forward projector software. We are currently adapting our reconstruction algorithm to match the AdaptiSPECT-C system geometry. Once reconstruction has been adapted to the AdaptiSPECT-C geometry we will investigate for potential issues in angular sampling and sampling uniformity, and investigate ways to address potential issues such as adding more apertures and using shuttering to control multiplexing, minor rotation of the system or slight translation motion of patient bed.

Poster panel
(face) ID: 337

Poster Number:

An accurate 3D positioning method for monolithic scintillator detector based on analytical model (#2301)

J. Zhuang1, P. Fan2, H. Liu2, Y. Xia3, T. Ma2

1 Yale University, Biomedical Engineering, New Haven, Connecticut, United States of America
2 Tsinghua University, Department of Engineering Physics, Beijing, Beijing, China
3 Institute of Spacecraft Environment Engineering, Institute of Spacecraft Environment Engineering, Beijing, Beijing, China


In PET applications, monolithic scintillator detector has 3D positioning capability, better energy resolution and sensitivity compared with discrete scintillator detector. However, position calculation of gamma photon events is more complicated. Current methods such as statistical based positioning (SBP) and artificial neural network (ANN), require large training datasets from measurements to approximate behavior of optical photons.

In this paper, we propose an analytical-model based positioning method (AMBP) which requires smaller training dataset. Optical photon distribution is approximated by weighted average of light distribution from two ideal models: scintillator covered with ideal specular film, and scintillator covered with ideal diffuse film whose reflection follows Lambert’s cosine law. Weighting factor of these two models, and amplification factor of each SiPM channel, are estimated from measurements. In AMBP, positioning is performed by maximizing linear correlation between model distribution and SiPM outputs.

We compared performance of SBP and AMBP in experiment with a 25x25x14 mm3 LYSO scintillator detector. SBP uses training data with 1mm scan step size, and AMBP method uses data with 2mm and 4mm scan step size. Positioning accuracy is measured by error (mean of absolute distance between true position and estimated position). In x,y and z (Depth of interaction) direction, SBP achieves error of 0.86mm, 0.87, 1.59mm, while AMBP with 2mm scan step achieves error of 0.95mm, 0.97mm, 1.43mm, and AMBP with 4mm scan step achieves error of 0.98mm, 1.01mm, 1.43mm. We also discussed how to adjust our model for different detector configurations, such as multi-readout, and use different reflection films.

Compared to SBP, AMBP uses only 1/16 of taining data, and achieves similar accuracy in x,y direction, and slightly better accuracy in z direction. AMBP is attractable for its accuracy, simplicity and being easily adaptable to different detector configurations.

Keywords: monolithic scintillator detector, analytical model, positioning method
Poster panel
(face) ID: 340

Poster Number:

A combinatorial method based on simulated annealing to design a biplanar MRI gradient system. (#2469)

D. Grau-Ruiz1, H. Sanchez1, J. P. Rigla2, E. Diaz-Caballero2, J. M. Gonzalez1, G. Puchalt1, A. Aguilar1, A. J. González1, S. Sánchez1, J. M. Benlloch1

1 Universitat Politecnica de Valencia, Instituto de Instrumentación para Imagen Molecular (I3M), Valencia, Spain
2 Tesoro Imaging S.L., Alicante, Spain


Gradient coils are a main part of a Magnetic Resonance Imaging (MRI) system. The aim of gradient coils in a MRI system is to encode spatially the Field of View (FoV). Different gradient coil design methods have been presented in the last years. Some factors such as linearity, inductance and resistance must be taken in account to obtain an optimized design for the gradient coils.

A combinatorial method to design the MRI gradient coils is presented. The method developed uses very simple shapes (rectangles and squares) to obtain the coil wire pattern. The algorithm presented divides the coil in as many portions as turns it contains, and all the possible wire combinations are analyzed using combinatorics. This method allows finding the best coil design in a time less than 2 minutes. Path combination which generates the highest gradient field is chosen for the X/Y gradient coil, and also the most homogeneous magnetic field is considered to select the Z-gradient wire combination. Efficiencies values of 9.3 and 19.4 mT/m/A are obtained for X/Y and Z gradient coil, respectively.

The gradient system has been tested in a biplanar permanent magnet system. Experimental gradient coil evaluation shows the ability to generate a strong gradient field (0.56 and 0.75 T/m for X/Y and Z gradient coil, respectively). A preliminary 2D MRI image with 1mm accuracy is obtained.

Keywords: gradient system, MRI, biplanar, combinatorial method
Poster panel
(face) ID: 343

Poster Number:

Comparison between Anger and Compton cameras for medical imaging: a Monte Carlo simulation study (#2837)

M. Fontana1, D. Dauvergne2, J. Krimmer1, J. - M. Létang3, J. L. Ley1, V. Maxim3, E. Testa1

1 IPNL, Villeurbanne, France
2 LPSC, Grenoble, France
3 CREATIS, Villeurbanne, France


Single Photon Emission Computed Tomography (SPECT) is at present one of the major techniques for non-invasive diagnostics in nuclear medicine. Clinical routine is mostly based on collimated cameras, originally proposed by Anger. Due to the presence of mechanical collimation, detection efficiency and energy acceptance are limited by the system geometrical features. In order to overcome these limitations, the application of Compton cameras for SPECT is being investigated for several years.

We propose in this study to compare a commercial Anger device, the GE Healtcare Infinia system, and the Compton camera prototype under development by our collaboration, through Monte Carlo simulations (GATE v7.1 and Geant4 9.6 respectively). Given the possible introduction of new radio-emitters at higher energies intrinsically allowed by the Compton camera detection principle, the detectors are exposed to point-like sources at increasing primary gamma energies, from actual isotopes already proposed for SPECT applications. The detector performances are studied in terms of radial event distribution, detection efficiency and final image, obtained by gamma transmission analysis for the Anger system, and with an iterative LM-MLEM algorithm for the Compton reconstruction. The Compton camera prototype is also characterized in terms of rate of random coincidences and at different energy resolutions.

Although the rate of random coincidences appears to be close to 50 % at clinical source activities, preliminary results show for the Compton camera a detection efficiency increased by a factor greater than an order of magnitude, associated with an enhanced spatial resolution for energies beyond 500 keV.

We discuss then the proven advantages of Compton camera application with particular focus on dose delivered to the patient, examination time and spatial uncertainties.

Keywords: SPECT, Compton camera, Anger camera, Simulation, Detection efficiency, Spatial resolution, Detector performance
Poster panel
(face) ID: 346

Poster Number:

Generation of Personalized Computational Phantoms Using Only Patient Metadata (#3190)

X. Zhong1, N. Strobel2, J. C. Sanders1, M. Kowarschik3, R. Fahrig3, A. Maier1, 4

1 Friedrich-Alexander University Erlangen-Nürnberg, Pattern Recognition Lab, Erlangen, Bavaria, Germany
2 Hochschule für angewandte Wissenschaften Würzburg-Schweinfurt, Schweinfurt, Bavaria, Germany
3 Siemens Halthcare GmbH, AT Innovation, Forchheim, Bavaria, Germany
4 Erlangen Graduate School in Advanced Optical Technologies, Erlangen, Bavaria, Germany


The use of personalized computational phantoms (CPs) describing a patient’s anatomy could enable attenuation correction for emission tomography without the need for a separate CT acquisition, thus reducing radiation dose to patients and cost to clinics. We propose a method to estimate such a phantom using only patient entry information (PEI) comprised of gender, height, and weight. The method uses a two-step machine learning-based approach whereby a joint subspace linking PEI to boundaries describing anatomical cavities is learned. Average organs from the training database are then deformed to fit within the cavities and used to populate them. We validate our method against two existing CPs and nine patient CT scans. The results show mean organ center of gravity displacement and volume errors in patients of less than 3 cm and 40%, respectively, in the lungs, liver, spleen, and kidneys. The relatively large volume error was most likely due to large inter-patient variations in organ mass. Nevertheless, our approach represents a step towards personalized CT-less AC and warrants future work to determine clinical relevance.

Keywords: Personalized computational phantom, computational phantom estimation, patient metadata, cavity estimation
Poster panel
(face) ID: 349

Poster Number:

Performance evaluation in three geometric detection arrangement for Clinical PET:Simulation study (#3907)

E. R. Moraes1, D. L. Franzé1, A. A. F. Martins1

1 University Of Sao Paulo, Departamento Of Physics-FFCLRP, Ribeirão Preto, SP, Brazil


Considering the human cross section is not exactly a circle and the fact that sensitivity and spatial resolution of PET images formed in coincidence may show better values with reduction of the distance between the coincident detectors three different detection geometries of a PET system have been simulated: one circular and two elliptical. The performance evaluation was done by system parameters: sensitivity, resolution and the Noise Equivalent Count Rate curve. The elliptical systems presented in general better sensitivity up to 63%, 25% lower expenditure with detectors, higher values for Noise equivalent Count Rate to the same or lower activity concentration, which suggest better signal to noise rate however and resolution test in general presented worst results, showing slightly improvement at coronal axes, and loss of resolution at sagittal and longitudinal axes. We believe the iterative reconstruction against filtered back projection could propitiate improvements on resolutions of elliptical against circular system. The present works shows preliminaries results of current study

Keywords: nuclear medicine, Monte Carlo Simulation, GaTE, PET systems
Poster panel
(face) ID: 352

Poster Number:

Evaluation of Photon Processing Detectors using the Fourier Crosstalk Matrix (#4222)

N. P. Henscheid1, A. K. Jha4, H. H. Barrett2, 3

1 University of Arizona, Program in Applied Mathematics, Tucson, Arizona, United States of America
2 University of Arizona, College of Optical Sciences, Tucson, Arizona, United States of America
3 University of Arizona, Department of Medical Imaging, Tucson, Arizona, United States of America
4 Johns Hopkins University, Department of Radiology and Radiological Sciences, Baltimore, Maryland, United States of America


We compare the parameter estimation performance of the photon processing emission imaging detector model to finite dimensional pixelated detector models using the Fourier crosstalk matrix formalism. The comparison is made for simulated SPECT systems using high accuracy parallel numerical integration to compute system models and the entries of the crosstalk matrices. Preliminary results indicate that a photon processing detector can be made equivalent to a pixelated detector with arbitrarily many pixels.

Keywords: Photon processing, Fourier crosstalk, Simulation, Emission imaging, Detector models