IEEE 2020 NSS MIC
Availability of interactive elements is extended to December 10, 2020.
Our exhibitors and sponsors:

Please support our exhibitors by visiting their virtual booths between Monday 9:00 AM and Friday 6:00 PM. Click on the company logo in the scrolling list above to the website now and to the exhibitor booth during the conference.
Their booths will be open for viewing equipment 24 hours a day and for live discussions during conference hours for the whole conference week.


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

MIC Poster Session II

   
Shortcut: M-08
Date: Thursday, 5 November, 2020, 2:00 PM - 3:30 PM
Room: MIC - 1
Session type: MIC Session

Contents

Click on an contribution to preview the abstract content.

2:00 PM Poster panel: 2

Poster Number:
M-08-002

Multi-Resolution SiPM Array Based PET Detector (#1027)

Q. Yang1, Y. Wang2, Z. Sang1, Y. Yang1, J. Du1

1 Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China
2 Sichuan Tianle Photonics Co., Ltd., Chengdu, China

Content

Multi-resolution silicon photomultiplier (SiPM) array, which was fabricated by using a mix of SiPMs with different sizes and arranging the SiPMs with smaller size at the periphery and the SiPMs with larger size at the middle of the SiPM array, was proposed to improve the position resolution and detection efficiency of PET detector by reducing the edge effect and to reduce the SiPM channel number for a given detector area. In this paper, the flood histograms of a PET detector based on a mimic multi-resolution SiPM array with SiPMs of 3.16 × 3.16 mm2 and 6.47 × 6.47 mm2, a PET detector based on a 16 × 16 array of 3.16 × 3.16 mm2 SiPMs and a PET detector based on a mimic 8 × 8 array of 6.47 × 6.47 mm2 SiPMs were compared by using a 15 × 15 LYSO array with a pitch size of 3.4 mm. We are fabricating a 36 × 36 LYSO array with a pitch size of 1.5 mm and a cross section matching the size of the SiPM array now. The LYSO array will be measured and the results will be shown at the conference.

Keywords: Multi-Resolution SiPM, PET, edge effect, PET detector, SiPM
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 5

Poster Number:
M-08-005

Performance of Dual-Ended Readout Depth-Encoding PET Detectors Based on LYSO and BGO Scintillators (#1101)

J. Du1, G. Ariño-Estrada1, X. Bai1, S. R. Cherry1

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

Content

The performance of dual-ended readout depth-encoding PET detectors based on LYSO arrays and BGO arrays were compared using SiPM arrays as photodetectors. BGO has not been widely studied for dual-ended readout designs.  The LYSO arrays, BGO arrays and the SiPM arrays all have the same pitch size of 2.2 mm, and the crystal arrays were coupled to the SiPM array using a one-to-one configuration. The results show that dual-ended readout detectors based on polished LYSO arrays with Toray reflector and unpolished BGO arrays with Toray reflector can provide good depth-of-interaction (DOI) resolution of gamma photons, whilst a dual-ended detector based on polished BGO arrays with Toray reflector cannot provide DOI information. All the crystal elements in the crystal arrays were clearly resolved, and the detectors based on LYSO arrays have better flood histogram, better DOI resolution and timing resolution since to LYSO scintillator has higher light output and faster decay time, however, the detectors based on BGO arrays have better detector-level and crystal-level energy resolution. Our next step is to design a BGO based total-body small animal PET for low-dose studies in mouse/rat models of human disease.

Keywords: PET, DOI, Dual-ended readout, LYSO, BGO
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 8

Poster Number:
M-08-008

Improving TOFPET2c Timing Performance by Passive Filtering of SiPM Signals (#1160)

M. Profe1, V. Nadig1, H. Radermacher1, D. Schug1, 2, B. Weissler1, 2, S. Gundacker3, 4, 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 Hyperion Hybrid Imaging Systems GmbH, Aachen, North Rhine-Westphalia, Germany
3 University of Milano-Bicocca, Milan, Italy
4 European Organization for Nuclear Research (CERN), Geneva, Switzerland

Content

Time-of-flight positron emission tomography has high potential in clinical practice as it offers an increased signal-to-noise ratio leading to better image quality compared to conventional positron emission tomography. The key to time-of-flight imaging is a precise measurement of the time-of-arrival difference of a coincidence event. By using analog silicon photomultipliers and custom-designed front-end electronics, the timing resolution of a system can be improved. Coincidence resolution times below 100 ps have been measured in benchtop experiments. This work focuses on the implementation of a passive filter between the TOFPET2 ASIC readout circuit and analog silicon photomultipliers to reduce the variation of the noise floor over the calibrated channel baseline. To characterize the timing performance of the circuit with the applied filter, we use a coincidence setup based on the TOFPET2 ASIC evaluation kit (PETsys Electronics S.A.). We employ AFBR-S4N66C013 photomultipliers (Broadcom) with an active area of 6×6 mm2 and LYSO:Ce scintillators co-doped with Ca of 6.5×6.5×6.5 mm3. Firstly, the influence of the implemented filter is investigated in single-channel measurements. Secondly, coincidence experiments equipping multiple channels of the same ASIC are performed. The performance of the benchtop system is compared for direct coupling and the implemented filter in terms of coincidence resolution time and energy resolution. Single-channel measurements show an improvement of the coincidence resolution time of about (23.6 +- 5.2) ps for a filtered signal compared to direct coupling, i.e., a gain in timing resolution of about 9.3 %. For a 2×2 array of silicon photomultipliers an improvement of approximately (47.0 +- 2.8) ps is achieved, i.e., a gain of about 17.0 % in timing resolution. The results confirm that signal filtering is essential to consider and that the selected filter is a suitable component to integrate when designing an analog sensor front-end.

Keywords: ASIC, CRT, positron emission tomography, signal filter, time-of-flight
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 11

Poster Number:
M-08-011

Sub-100 ps PET Detectors Suitable for High-Resolution Brain Mouse Imaging (#1216)

E. Lamprou1, S. Aguilar1, G. Cañizares1, J. Barrio1, N. Cucarella1, M. Freire1, S. Echegoyen1, F. Sanchez1, L. F. Vidal1, J. M. Benlloch1, A. J. Gonzalez1

1 Institute for Instrumentation in Molecular Imaging, (I3M-CSIC), Valencia, Spain

Content

Efforts have been devoted over the years to enhance the Time of Flight capabilities of PET detectors. Following several technological advances, a timing resolution below 100 ps FWHM seems achievable nowadays even at the system level. However, compromises have to be made in order to reach that good resolution. In this work, we present our approaches to develop an ultra-high spatial and timing resolution PET ring. This ring will be the scatter layer in a Compton-PET configuration for mouse brain imaging. Crystals (both pixelated and monolithic) with 3 mm thickness and overall sizes around 10 mm x 10 mm are tested. Experiments using 8 x 8 SiPM arrays of 3 x 3 and 1 x 1 mm2 (both with 50 and 25 μm cell sizes) have been used aiming to find the best detector configuration for sub-100 ps resolution scanners. We found the best detector performance when using crystal arrays of 1 mm pixel (8 x 8 elements) of the type LYSO(Ca) when combined with 3 x 3 mm2 (50 μm cell size) Hamamatsu Photonics SiPM (S13360). We reached a detector resolution as good as 65 ps FWHM in some cases. Additionally, with the same SiPMs coupled to a 3×3×10 mm3 LYSO(Ca) monolithic block, following calibrations and a timestamps averaging method, a DTR of 158 ps for a Volume-Of-Interest (VOI) was achieved. Several more results are presented in this work as well as methods to improve the RAW timing resolution.

Acknowledgment

This work was supported by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant Agreement No. 695536). It has also been supported by the Spanish Ministerio de Economía, Industria y Competitividad under Grant TEC2016- 79884-C2-1-R

Keywords: TOF-PET, Crystal Arrays, Monolithic crystals, SiPMs, ASIC
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 14

Poster Number:
M-08-014

Channel-Individual Time-of-Flight Performance of the TOFPET2c ASIC (#1332)

V. Nadig1, D. Schug1, 2, H. Radermacher1, B. Weissler1, 2, 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 Hyperion Hybrid Imaging Systems GmbH, Aachen, North Rhine-Westphalia, Germany

Content

In positron emission tomography systems, the time-of-flight information of coincidence events acquired by several thousand detector channels needs to be digitized. Apart from the scintillator and photo sensor chosen, also the employed readout electronics, requiring compact circuit design to read out multiple channels at once, have a large impact on the time-of-flight performance of the system. This study evaluates the performance for twelve samples of the TOFPET2 ASIC (version 2c) designed by PETsys Electronics S.A. packaged as chip on board and ball grid array, each providing 64 individual readout channels for simultaneous event digitization. Such a broad study of the TOFPET2c performance of in total 768 channels with additional reference to its packaging option has, to our knowledge, not been conducted so far. For this purpose, two detector blocks, each consisting of a Hamamatsu S14161-3050-HS-08 SiPM array one-to-one coupled to an 8 x 8 LYSO matrix of 12 mm height using a two-component dielectric gel, are successively coupled to the twelve ASIC samples. Their performance is assessed in multi-channel coincidence experiments, employing a geometry of five Na-22 point sources with a total activity of approximately 2 MBq. The channel linearity ratios, energy resolutions as well as coincidence resolution times show Gaussian distributions for all ASIC samples. Computing the global performance, all ASIC samples achieve coincidence resolution times down to 250.7 ps and energy resolutions down to 10.7 % at a raw hit rate of approximately 3 Mcps and a coincidence rate of 67 kcps at an overvoltage of 3 V, applying an energy window of 450 to 600 keV.
For higher overvoltages, chips on board achieve slightly better timing resolution than ball grid arrays. The results obtained encourage the usage and integration of the TOFPET2 ASIC in large systems without the need for an enormous effort regarding individual configuration and thus, no risk of ASIC-dependent performance.

Keywords: time-of-flight, positron emission tomography, ASIC, CRT, energy resolution
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 17

Poster Number:
M-08-017

Study of optical reflectors used in scintillation detectors that achieve 100 ps coincidence time resolution for TOF-PET. (#1445)

A. Gonzalez-Montoro1, S. Pourashraf1, M. S. Lee1, J. W. Cates2, Z. Zhao1, C. S. Levin1, 3

1 Stanford University, Department of Radiology, Stanford, California, United States of America
2 Lawrence Berkeley National Laboratory, Applied Nuclear Physics, Berkeley, California, United States of America
3 Stanford University, Department of Bioengineering, Physics and Electrical Engineering, Stanford, California, United States of America

Content

Incorporating 511 keV photon time-of-flight (TOF) information in PET enables a significant boost in reconstructed image signal-to-noise ratio (SNR). This SNR boost depends on the 511 keV photon pair coincidence time resolution (CTR) of the PET system, which is determined by several factors including properties of the scintillation crystal and photodetector, crystal-to-sensor coupling configurations, and reflective materials. The goal of the present work is to achieve 100 picoseconds (ps) CTR for > 2-fold additional improvement in reconstructed image SNR compared to state-of-the-art PET systems that currently have 250-400 ps CTR. A critical parameter to optimize in achieving this goal is the optical reflector’s influence on light collection and transit time to the photodetector. For the experimental set-up, we made use of PET detector elements based on both 3x3x10 and 3x3x20 mm3 LGSO crystals coupled to an array of SiPMs by using a novel “side-readout” configuration that enables high scintillation light collection efficiency and low photon transit time to the photodetector. We have tested the CTR performance by applying three different reflector materials to the crystal surfaces, namely: Teflon, BaSO4and TiO2paints. Results show CRT values of 102.0, 97.3 and 95.0 ps FWHM for the 10 mm length crystals and 112.1, 107.4 and 98.2 ps FWHM for 20 mm length crystals (SiPMs only partially covered the side of the 20 mm crystal), for the Teflon, BaSO4 and TiOpaint treatments, respectively.

AcknowledgmentThis work was partially supported by NIH grants R01CA214669 and R01EB025125.
Keywords: Positron Emission Tomography, Time Of Flight, TOF-PET, SNR, SiPM
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 20

Poster Number:
M-08-020

New Generation of Large-area Direct X-ray Detectors for High Sensitivity Medical Imaging (#1538)

A. Datta1, J. Fiala1, S. Motakef1

1 CapeSym Inc., Research and Development, Natick, Massachusetts, United States of America

Content

With the advent of flat panel detector technology, flat panel x-ray imagers (FPXIs) are now widely used in digital X-ray imaging, specifically for digital radiography, fluoroscopy, digital tomosynthesis, image-guided radiation therapy, and cone beam computed tomography. These imagers are currently being heavily used during the COVID-19 crisis to assess the peculiar lung-damage that is unique to this virus. Commercial FPXIs are primarily based on scintillators and rely on indirect conversion of x-rays to photons and then to electronic signals or are based on amorphous-Se semiconductor films that directly convert absorbed x-rays to an electronic signal.  The indirect FPXIs have high detective quantum efficiency (DQE) over the energy range of interest (up to 140kVp) and are used in most of the FP imaging applications. However, due to the isotropic propagation of light in the scintillators, these systems lack the necessary spatial resolution that is needed for other imaging applications. On the other hand, a-Se based panels have excellent spatial resolution and DQE, but only up to 40kVp, due to low absorptivity of a-Se at higher energies. Thus, the use of direct conversion FPXIs is currently limited only to soft-tissue applications. In this presentation, we will be demonstrating direct conversion X-ray detection with a detector structure based on novel photo-detecting polycrystalline semiconductor Methylammonium lead iodide (MAPbI3) characterized by a high attenuation coefficient and excellent charge transport properties. The critical challenge, as is well known in the field of metal-organic semiconductors, is the repeatability in achieving low dark current under high enough electric fields. We have developed a detector structure that combines different current limiting layers to achieve this goal. These detectors not only provides low dark current and high X-ray sensitivity but also can be deposited over large areas for manufacturing FPXIs with repeatable performance.

Keywords: X-ray Detectors, Semiconductor detectors, Dark current, X-ray sensitivity., COVID-19
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 23

Poster Number:
M-08-023

This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 26

Poster Number:
M-08-026

Design and development of a compact high-resolution detector for PET insert in small animal irradiator (#1678)

X. Cheng1, K. Hu1, D. Yang1, Y. Shao1

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

Content

The goal of this project is to design and develop a high-resolution PET detector that will balance the performance, readout simplicity, and space requirement for integrated onboard PET/CT image guided translational radiotherapy. The detector consists of 4 sub-detectors; each sub-detector consists of a 15×15 1×1×20 mm3 LYSO scintillator array with each of its end coupled to an 8×8 array of 2×2 mm2 silicon photomultipliers for depth-of-interaction (DOI) measurement; all sides of scintillators were roughened with a 0.03 mm surface lapping for balanced detector performance. Based on row and column signal readout, only 64-ch electronics is required to process 512-ch output of each detector. A compact 96-ch electronics board was developed to measure the charge, timing, and position of each interaction and convert them in digital output pulses for further processing. The detector performance evaluation study shows an average ~26.3% energy resolution, ~3.2 ns coincidence timing resolution, and ~3.2 mm DOI resolution from all crystals. Each scintillator can be well identified. The maximum throughput of the detector is ~200 K events/s.

Keywords: PET, detector, FPGA, TDC, DOI
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 29

Poster Number:
M-08-029

Discretized DOI Encoding of a Trapezoid Shape PET Detector Made Using 3D Sub-Surface Laser Engraving (#1739)

H. G. Kang1, F. Nishikido1, H. Tashima1, T. Sakai2, T. Yamaya1

1 National Institute of Radiological Sciences (NIRS) in National Institutes for Quantum and Radiological Science and Technology (QST), Department of Nuclear Medicine Science, Chiba, Japan
2 Hamamatsu Photonics K.K,, The Development Center, Hamamatsu, Japan

Content

Previously for small animal positron emission tomography (PET), we developed the trapezoid geometry depth-of-interaction (DOI) detector using a sub-surface laser engraving (SSLE) technique. However, the required DOI calibration should be done with collimated beams which takes a long time. This problem can be solved by segmenting the detector crystal in the DOI direction using SSLE. In this study, we present the preliminary results obtained for a DOI detector with a laser processed trapezoid shape LYSO crystal for a small animal DOI PET scanner. The proposed DOI detector consisted of three trapezoid shape LYSO crystal plates.  Each 20 mm long LYSO plate (top =  15.3 mm, bottom = 25.6 mm, thickness = 0.9 mm) was segmented into a 15 × 10 array by using SSLE to yield a top crystal pitch of 1.02 mm, a bottom crystal pitch of 1.71 mm, and DOI bin of 2.0 mm (10 segments). Unlike our previous work, the crystal was segmented in the DOI direction to provide discretized DOI information. The LYSO plates were optically isolated by using an enhanced specular reflector (ESR). A 7×6 array of SiPMs (S13360-2050VE, Hamamatsu Photonics, Japan) with a pixel pitch of 2.4 mm and an 8×4 array of SiPMs (S13361-3050NE-04, Hamamatsu Photonics) with a pixel pitch of 3.2 mm were coupled to the top and bottom surfaces of the trapezoid LYSO array, respectively. The SiPM output signals were multiplexed by using a resistive network, and then digitized using the CAMAC DAQ. We could identify all the 15×3 crystals in the 2D flood map of the back SiPMs. Nine of ten DOI segments could be identified and each had a 2 mm DOI bin. The measured energy resolution was 11.4±1.2%. In the future, we plan to increase the number of crystals in the axial direction to use the full sensitive area of the SiPM.

Keywords: Sub-surface laser engraving, DOI, PET
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 32

Poster Number:
M-08-032

Characterization of a PET Detector Block based on a Semi-Monolithic Crystal with DOI and TOF Capabilities (#1773)

J. Barrio1, N. Cucarella1, M. Freire1, E. Lamprou1, S. Aguilar1, V. Ilisie1, J. M. Benlloch1, A. J. Gonzalez1

1 Institute for Instrumentation in Molecular Imaging (i3M-CSIC), Valencia, Spain

Content

A detector block based on semi-monolithic scintillators consists of a monolithic crystal segmented in one direction in different pieces called slabs. In this configuration, optical photons are preferred to only propagate in the monolithic direction, but not in the segmented direction. Semi-monolithic crystals are intended to overcome the limitations of monolithic and pixelated crystals, offering accurate time of flight (TOF) and depth of interaction (DOI) information, as well as good spatial resolution. In the work presented here, the characterization of a semi-monolithic detector with DOI and TOF capabilities for PET has been carried out. The detector consists of a single LYSO slab of 50×18×3 mm3 with all faces polished. ESR foils have been added to all faces, except the two 50×3 mm2 sides. One of these faces is covered by a retroreflector layer and the other one is optically coupled to a linear array of 1×16 SiPMs of 3×3 mm2 active area each with 50 μm cell size and 3.2 mm pitch. The 16 individual signals are read out by the TOFPET2 ASIC. We reached an energy resolution of 13% FWHM at 511 keV. The interaction position has been calculated with analytical methods. The preliminary spatial resolution is around 2 mm FWHM in the monolithic and in the DOI directions. A detector time resolution (DTR) of 317 ps FWHM has been obtained for the whole detector after applying skew time and time-walk corrections. This value improves to 260 ps FHWM DTR if energy weighted averaging of the timestamps is applied. If a region of interest close to the photodetector is selected, the DTR decreases to 224 ps FWHM. New statistical methods to improve the spatial and the timing resolutions are under investigation.

Keywords: DOI, PET, semi-monolithic crystal, SiPMs, TOF
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 35

Poster Number:
M-08-035

Flexible and Large Area Direct Conversion X-ray Detector Arrays (#1835)

J. Zhao1, L. Zhao1, Y. Deng1, X. Xiao1, Z. Ni1, S. Xu1, J. Huang1

1 UNC CHAPEL HILL, Applied Physical Sciences, CHAPEL HILL, North Carolina, United States of America

Content

Metal halide perovskites are rising as a new candidate for low-dose X-ray detection because of its high mobility-lifetime product and strong stopping power. The comparable softness of perovskites with polymers makes them potentially applicable for flexible X-ray detectors. Here we report a highly sensitive, flexible, and large-area x-ray perovskite detectors. The flexible perovskite X-ray detector with areas up to 400 cm2 are formed. The good connectivity and crystallization of perovskite crystals enable a large mobility-lifetime product. The sensitivity of the X-ray detectors under a field of 0.05 V/µm reaches 8696 µC Gyair−1 cm−2, and shows no degradation after storage for over six months and exposing a total dose of 376.8 Gyair, equivalent to 1.88 million Chest X-ray scans. The flexible X-ray detector can be bent at radii down to 2 milliliters without losing performance. The stand-alone detector array is curved and put inside metal pipes for the detection of material defects with superior imaging quality to flat-panel detectors. This work points out a new direction for the application of perovskites in highly sensitive flexible X-ray detectors for low-dose medical imaging, security screening, and non-destructive material inspections.

Keywords: Direct conversion, Flexible X-ray detectors, High sensitivity, Large area, Perovskite X-ray imagers
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 38

Poster Number:
M-08-038

A 5-axis calibration stage for depth-of-interaction-correcting scintillation crystals (#1969)

O. Anderson1, L. Bläckberg3, S. Sajedi3, H. Sabet3, L. R. Furenlid1, 2

1 University of Arizona, College of Optical Sciences, Tucson, Arizona, United States of America
2 University of Arizona, Department of Medical Imaging, Tucson, Arizona, United States of America
3 Gordon Center for Medical Imaging, Department of Radiology, Massachsetts General Hospital, Harvard Medical School, Boston, Massachusetts, United States of America

Content

New laser-induced optical-barrier technology designed to reduce depth-of-interaction blurring requires a new calibration approach.  Calibration by using a collimated beam of gamma rays to acquire a mean detector response function (MDRF) is important for maximum-likelihood estimation of interaction location. We have designed a 5 axis calibration stage in order to meet this need.  The calibration system will be able to move independently in 3 spatial dimensions and 2 rotational dimensions to allow for complete control of collimated beam position and direction.  This will allow us to simulate gamma rays coming from a pinhole from any angle in the field of view.  The complexity of the movements between positions of the system motivated the use of a probe mechanism to map out the detector surfaces.  We are incorporating collision detection and automation in the software to simplify and speed up the calibration process.

AcknowledgmentThis work was supported by the National Institute of Health Grant 1R01HL145160-01  "High performance SPECT System for Cardiac Imaging"
Keywords: depth of interaction, gamma-ray detector, calibration, SPECT
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 41

Poster Number:
M-08-041

Simultaneous multi-nuclide SPECT imaging with double photon emission coincidence method (#2103)

M. Uenomachi1, K. Ogane2, K. Shimazoe1, H. Takahashi1

1 The University of Tokyo, School of Engineering, Tokyo, Japan
2 The University of Tokyo, Graduate School of Medicine, Tokyo, Japan

Content

Single photon emission tomography (SPECT) is one of important nuclear medicine imaging modalities. The conventional SPECT imaging is performed by using a mechanical collimator to localize the direction of incoming gamma-rays. Although the simultaneous multi-nuclide imaging is required for medical diagnosis or research, it is difficult due to low energy resolution and spatial resolution of the conventional scintillator-based SPECT. One of promising method to realize simultaneous SPECT imaging is double photon coincidence method. This method can increase the signal-to-noise ratio (SNR) in a reconstructed image. For example, 111In (SPECT nuclide) emits 171 keV and 245 keV cascade gamma-rays. 177Lu (therapeutic nuclide) emits 208 keV and 113 keV cascade gamma-rays. In this study, we demonstrated 111In and 177Lu simultaneous SPECT imaging with the double photon coincidence method. For the demonstration, we used 8x8 array GAGG schintillators coupled to 8x8 array SiPMs and 8x8 array parallel hole collimators. Four detectors were placed at 90 degrees and approximately 0.2 mL of 111In and 177Lu were measured. The SNRs in images of two nuclides by using the double photon emission coincidence method were drastically increased.  The imaging results shows that the double photon emission coincidence method is useful to separate images of the two nuclides.

Keywords: SPECT, multi-nuclide imaging, double photon coincidence method
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 44

Poster Number:
M-08-044

A novel method for processing photonic crystals:Femtosecond laser (#2212)

W. Jiao1, X. Zhang1, S. Xie2, G. Ying1, J. Xu1, Q. Peng3

1 Huazhong university of Science and Technology, School of Mechanical Science and Engineering, Wuhan, China
2 Pitech Company, Shenzhen, China
3 Lawrence Berkeley National Laboratory, Department of Molecular Biophysics and Integrated Bioimaging, Berkeley, California, United States of America

Content

1. Introduction
2. Method
2.1 Photonic crystal Simulation
2.2 Femtosecond laser processing
3. Results
3.1 Simulation results
3.2 Processing results
4. Discussion and conclusion

AcknowledgmentThis work was supported by the National Natural Science Foundation of China (51627807), the National Natural Science Foundation-Guangdong Joint Funds of China (U1501256), the 111 Project (B16019), Hubei International Cooperation Project (2018AHB001), and the National Institutes of Health, National Institute of Biomedical Imaging and Bioengineering (R01EB006085).
Keywords: Photonic crystal, Femtosecond laser, Light ouput, Crystal processing
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 47

Poster Number:
M-08-047

A Cost-Effective Field-Programmable-Gate-Array-based Pulse Processor for Biomedical Imaging Applications (#2371)

A. Refaey1, G. Burkett1, J. Du2, R. D. Badawi1, 2

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

Content

This work explains the design and characterization of a low-cost FPGA-based pulse-processing circuit for energy estimation and time-of-flight measurements. The design is fully digital and intended to meet the requirements of the scintillation detectors used in positron emission tomography (PET) systems. Readout is achieved using sigma-delta and tapped delay-line approaches. Two waveforms are used to test the design; an arbitrary waveform generated by a function generator, and a waveform generated by an actual PET detector. The PET detector consists of a single 3x3x20 mm3 LYSO crystal coupled to a single 3x3 mm2 silicon photomultiplier (SiPM). For a pair of simple PET detectors positioned 80 cm apart running in coincidence mode, and a Na-22 source placed at 37 cm from the start detector, the best timing resolution achieved was 242 ± 2.35 ps. The direct measurement of energy resolution was 9.0% for the start detector and 9.3% for the stop detector for 511 keV events. In conclusion, we have developed a cost-effective, compact time-to-digital converter (TDC) for energy and time estimation for PET applications.

Keywords: PET, time-to-digital converter (TDC), Timing resolution, TDL, FPGA
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 50

Poster Number:
M-08-050

A Feasibility Monte Carlo Simulation for Whole-body Study using Variable-Aperture Full-Ring CZT-SPECT (#1170)

Y. Huh1, O. U. Dim2, Y. Cui2, W. Tao1, 3, Q. Huang3, G. T. Gullberg1, Y. Seo1

1 University of California San Francisco, the Department of Radiology and Biomedical Imaging, San Francisco, California, United States of America
2 Brookhaven National Laboratory, the Department of Nonproliferation and National Security, Upton, New York, United States of America
3 Ruijin Hospital and Shanghai Jiao Tong University, the Department of Nuclear Medicine and Biomedical Engineering, Shanghai, China

Content

We previously presented a variable-aperture full-ring geometry SPECT scanner using a pixelated CZT detector (i.e., CZT-SPECT) for brain studies and compared its performance metrices with those of a conventional SPECT system. In this current study, we propose an extended full-ring CZT-SPECT concept for whole-body studies with a similar design, i.e., each detector module can move radially in and out and rotate around the common axes of rotation. The proposed SPECT system consists of eight modules of 179.2 mm × 128.0 mm large-area pixelated CZT detectors with energy-optimized parallel hole collimator, having a 230 ~ 250 mm radius of rotation in order to cover the whole body. Extended projection data for reconstruction are generated by combining two adjacent projection data taken by two detector modules, with corrections for gap and edge between the two detectors in order to overcome a truncation problem. Derenzo hot-rod phantom was simulated with GATE package and the phantom images were reconstructed using an iterative reconstruction algorithm (STIR package) for a comparison study with a conventional SPECT system. The reconstructed phantom images for the proposed scanner showed no truncation effect when corrections for the gap and edge were applied. Qualitative assessment showed rods in the edge of 7.9 mm diameter rod segment were well differentiated. Initial simulation studies demonstrate that our proposed SPECT system is a robust concept, potentially a better alternative to the conventional dual-head SPECT scanner, taking advantage of flexible geometry and a single collimator covering a broad range of gamma energies. For future work, acquisition strategy and correction methods will be investigated to improve spatial resolution and evaluated for the proposed system with the XCAT phantom including various anatomical compartments and lesions of varying sizes and activity levels relevant to radionuclide distribution of clinically used SPECT radiotracers.

Keywords: CZT, SPECT, MC simlation, whole-body, bone study
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 53

Poster Number:
M-08-053

Design study of a breast-dedicated PET system with a round-edge detector arrangement (#1241)

G. Akamatsu1, H. Tashima1, S. Ito2, H. G. Kang1, E. Yoshida1, M. Takahashi1, T. Yamaya1

1 National Institute of Radiological Sciences (NIRS-QST), Chiba, Japan
2 Furukawa Scintitech Corp., Tsukuba, Japan

Content

Breast-dedicated PET systems are categorized into two types, a ring-shaped detector arrangement or a two-panel parallel detector arrangement. Although there are some advantages in the parallel arrangement, PET images are blurred due to the limited angular coverage. In this work, to compensate for the angular coverage issue while keeping the open space to access the breast, we proposed a third type, a round-edge detector arrangement. Using the Geant4 simulation, we evaluated the imaging performance of the three detector arrangements: a ring arrangement, a round-edge arrangement, and a parallel arrangement. Adding TOF and DOI measurement capabilities was also investigated. Each simulated system was composed of the same 50 detector blocks, for which the scintillator was a 30×30 array of 1.5×1.5×15 mm3 GAGG. We evaluated sensitivity, spatial resolution and image quality (contrast and noise). TOF (timing resolution: 500-ps and 300-ps) and DOI (7.5-mm-thickness × 2-layer) information were applied in the image reconstruction to investigate their impacts on image quality. As a result, we saw that the round-edge arrangement improved the spatial resolution and slightly reduced the image blurring, compared to the parallel one. In the non-ring arrangements, the 500-ps TOF slightly reduce the image blurring, while the 300-ps TOF eliminated it. Using the 2-layer DOI information, the contrasts of small spheres were increased in all arrangements. In conclusion, our proposed round-edge arrangement could improve the spatial resolution and reduce the image blurring with the help of TOF and DOI information. The 300-ps TOF performance was needed to eliminate the image blurring and the DOI measurement capability was preferable to obtain higher contrasts of small lesions.

AcknowledgmentThis research was supported by AMED under Grant Number JP20he2202004.
Keywords: PET, breast, round-edge, TOF, DOI
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 56

Poster Number:
M-08-056

Study of the Coincidence Energy Spectra due to the Intrinsic Radioactivity of LYSO Crystals (#1294)

F. E. Enríquez-Mier-y-Terán2, A. S. Ortega-Galindo1, T. Murrieta-Rodríguez1, M. Rodríguez-Villafuerte1, A. Martínez-Dávalos1, H. Alva-Sánchez1

1 Instituto de Fisica, UNAM, Experimental Physics Department, Mexico City, Mexico
2 University of Sydney, Faculty of Engineering, Sydney, Australia

Content

In this work we studied the energy spectra due to the intrinsic radioactivity of cerium doped lutetium yttrium oxyorthosilicate (LYSO) scintillation crystals of two opposing detectors working in coincidence mode. The investigation included experimental data, Monte Carlo simulations using GATE and an analytic model. The structure of the energy spectra was completely understood as the result of the self-detection of a beta particle from 176Lu in one crystal and the detection of one or more prompt gamma rays from the isomeric transitions of 176Hf detected in coincidence by the opposing crystal. The most probable coincidence detection involves the gamma rays of 202 and 307 keV, which result in two narrow peaks, superimposed on a continuous energy distribution due to the beta particle energy deposition. The relative intensities of the gamma ray peaks depend on the crystal size and detector separation distance, as is explained by the analytic model and verified through the Monte Carlo simulations and experiments. This work will be useful to those research groups performing transmission scans with the intrinsic radioactivity of LYSO crystals and for detector calibration in coincidence mode.

Keywords: Energy spectra, Intrinsic radioactivity, GATE, LYSO, PET detector
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 59

Poster Number:
M-08-059

Delivery, beam and range monitoring in Particle Therapy in an innovative integrated design (#1438)

P. Cerello1, L. Bottura2, E. Felcini2, 3, V. Ferrero1, E. Fiorina1, V. Monaco4, 1, F. Pennazio1, G. de Rijk2

1 INFN, Torino, Italy
2 CERN, Geneva, Switzerland
3 EPFL, Lausanne, Switzerland
4 University of Torino, Department of Physics, Torino, Italy

Content

The design of a particle therapy system that integrates an innovative beam delivery concept based on a static toroidal gantry and an imaging configuration suitable for beam and online range monitoring is proposed and discussed.
Such approach would provide a smaller, faster and cost-effective layout, with a highly flexible and fast beam delivery, single particle counting capability for fast measurement of beam fluence and position and a precise real time verification of the compliance between the treatment delivery and its prescription.
The expected performance of the PET-based range monitor in the reconstruction of the activity distribution, obtained by simulating the delivery of a clinical treatment plan in two different configurations, shows that its precision is better than the state-of-the-art devices.
The feasibility of the proposed design is then discussed through an assessment of the technological improvements required to actually start the construction and commissioning of a system prototype.

Acknowledgment

This work is part of ATTRACT that has received funding from the European Union’s Horizon 2020 Research and Innovation Programme, with project name Hybrid High-precision In-vivo Imaging in Particle Therapy (H2I2). The project is co-funded by the CERN Budget for Knowledge Transfer to Medical Applications.

Keywords: Gantry, PET, UFSD
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 62

Poster Number:
M-08-062

Experimental Validation of the ANTS2 code for Modelling the Transport of Optical Photons in Monolithic LYSO Crystals (#1590)

V. M. Lara-Camacho1, E. M. Hernández-Acevedo2, F. E. Enríquez-Mier-y-Terán3, H. Alva-Sánchez2, T. Murrieta-Rodríguez2, A. Martínez-Dávalos2, M. Rodríguez-Villafuerte2

1 Instituto Politécnico Nacional, Escuela Superior de Física y Matemáticas, CDMX, Mexico
2 Universidad Nacional Autónoma de México, Instituto de Física, CDMX, Mexico
3 University of Sydney, Faculty of Engineering and Information Technologies, Sydney, Australia

Content

ANTS (Anger-camera type Neutron detector: Toolkit for Simulations) is a Monte Carlo simulation code for the transport of particles and optical photons, with the capability of including the full signal formation process in the photodetector. In this work, we study the performance of ANTS2 to model the transport of light photons in scintillation-based PET detectors. To this end, the intrinsic background of [176Lu]-LYSO monolithic scintillators was simulated for two crystal sizes (57.4×57.4×10 mm3 and 10×10×10 mm3). Three different optical reflectance properties of the crystal surfaces were considered: diffuse reflective, specular reflective and light-absorbing. The MC data were validated with experimental results obtained with detector-blocks using a setup as close as possible as the one assumed in the simulation. In all cases, the MC results have very good agreement with the experimental data, except for the small crystal covered with black paint, in which the intrinsic background structure disappears entirely.

AcknowledgmentE. M. Hernández-Acevedo acknowledges the postdoctoral scholarship from Programa de Becas Posdoctorales, DGAPA-UNAM.
Keywords: PET, Monte Carlo, Optical Transport, LYSO
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 65

Poster Number:
M-08-065

A Simulation Study on Two Readout Methods of a Monolithic Scintillator Ring PET (#1702)

Q. Xie1, Y. Xie1, S. Xie1, X. Zhang1, Q. Peng2, J. Xu1

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

Content

We simulated and evaluated two preclinical PET systems consisting of a monolithic ring based on previous study(MSR) in GATE with different readout methods: Dual-ended readout and Inner-surface readout. The Center-of-Gravity method and neural network algorithm were used to decode interaction positions of single photons and performances of both PET systems were evaluated.
The crystal of the two detectors are identical, which are cylinders made of LYSO with an inner diameter of 48.5mm, an outer diameter of 58.5mm and a height of 25mm. The dual-ended readout monolithic crystal ring detector (DERD) is uniformly coupled with 46 SiPMs on the upper and lower surfaces. As for the inner surface readout monolithic crystal ring detector (ISRD), eight rings containing 46 SiPMs are coupled to the inner surface of the crystal. The Center-of-Gravity (COG) method and the neural network (NN) method were used to decode circumferential and axial direction of DERD respectively and the COG method was adapted to decode both circumferential and axial direction of ISRD. FWHMs and MAEs of circumferential direction decoding of DERD and ISRD are 1.02±0.34mm and 0.61±0.26mm mm respectively. In the axial direction, decoding solution represented by FWHMs and MAEs of DERD and ISRD are 0.68±0.23mm and 0.38±0.26mm respectively.
To further evaluate the performance of DERD, a miniature Derenzo phantom was built and simulated. Images were reconstructed by using MLEM reconstruction algorithm. We concluded that best reconstruction image quality can be achieved when the number of iterations was 50, and rods with a diameter of 0.5mm could be clearly distinguished meanwhile.

Acknowledgment

This work was supported by the National Natural Science Foundation of China (51627807), the 111 Project (B16019), Hubei International Cooperation Project (2018AHB001), and the National Institutes of Health, National Institute of Biomedical Imaging and Bioengineering (R01EB006085).

Keywords: monolithic ring, readout method, GATE, scintillator, Positron emission tomograph
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 68

Poster Number:
M-08-068

Simulation study on the effect of depth-of-interaction and source position on time performance in monolithic crystals (#1779)

C. Thyssen1, 2, K. Deprez1, P. Mollet1, R. Van Holen1, 2, S. Vandenberghe2

1 Molecubes NV, Gent, Belgium
2 Ghent University, Medisip, Gent, Belgium

Content

Since the introduction of time-of-flight (TOF) in positron emission tomography (PET) in 2007, it has only been commercially implemented in clinical systems with pixelated scintillators. However, monolithic crystals are valuable as they allow to model depth of interaction, increase sensitivity, improve resolution and reduce production costs. In order to develop a monolithic TOF detector, it is important to understand how signal processing can be optimized to achieve good timing performance. In this study, we use optical Monte Carlo simulations to study the effect of DOI and source position on the single photon time resolution (SPTR) of a monolithic crystal. Using the SPTR’s, we make an estimate of the coincidence time resolution (CTR) and the TOF-sensitivity gain. Our results show an optimum in the SPTR of 89ps when averaging the four fastest timestamps with a source located in the center. SPTR increases to 109ps near the corner. Events that annihilate closest to the top of the crystal result in the best SPTR. The combination of two SPTR gives an estimation for the CTR: 125ps and 154ps respectively. This corresponds with a 19mm and 23mm uncertainty (FWHM) along the line-of-response which results in an effective sensitivity improvement for an average patient (~27cm diameter) with a factor ~8.9 and ~7.3

Keywords: Positron Emission Tomography, Time-of-Flight, Monolithic detector, Monte Carlo simulations, Optical simulations
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 71

Poster Number:
M-08-071

Multi-purpose Ultra-fast Monte Carlo PET simulator (#1945)

P. Galve1, F. Arias-Valcayo1, A. Lopez-Montes1, A. Villa-Araunza1, P. Ibáñez1, J. L. Herraiz1, 2, J. M. Udías1, 2

1 Universidad Complutense de Madrid, Grupo de Física Nuclear and IPARCOS, Madrid, Spain
2 Instituto de Investigación Sanitaria del Hospital Clínico San Carlos (IdISSC), Madrid, Spain

Content

Accurate Monte Carlo (MC) simulations play a major role in PET corrections and scanner design. We present the Ultra-fast Monte Carlo PET simulator (UMC-PET), a multi-purpose, accurate, fast and flexible PET simulator. The UMC-PET includes all the relevant physics related to the emission, transport and detection of the radiation in a PET acquisition, such as positron range, non-colinearity, scatter and attenuation inside the patient, photon interaction with the scanner, and detector response (energy resolution, depth of interaction, time of flight, etc.). The simulator accuracy has been extensively tested against other MC PET packages such as PeneloPET, obtaining similar results, while being more than 3000 times faster. The code can handle arbitrary scanner geometries with simple and intuitive input files. These features make UMC-PET useful beyond standard MC applications. UMC-PET has been tested to accurately predict scatter and attenuation corrections during reconstruction, or to compute system response matrix (SRM) for adaptative geometry scanners. Furthermore, the speed of UMC-PET makes it possible 3D iterative reconstruction for complex scanners, with a projection step based on, on the fly, full MC calculations, avoiding explicit storage of the SRM or any physics simplifications. On a single common GPU these fully MC reconstructions require a few hours for a scanner with > 1 billion lines of response. This provides not only a useful and flexible gold standard reconstruction, but may become a practical approach if it is combined with variance reduction methods and/or high performance multi-GPU systems.

Acknowledgment

P. Galve is supported by a Complutense University of Madrid, Moncloa Campus of International Excellence and Banco Santander predoctoral fellowship, CT27/16-CT28/16.

Keywords: GPU, Positron emission tomography, reconstruction algorithms, simulation
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 74

Poster Number:
M-08-074

Compton Decomposition and Recovery in a Prism-PET Detector Module (#1967)

E. W. Petersen1, A. LaBella1, W. Zhao2, A. H. Goldan2

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

Content

Identifying the correct Line-of-Response (LOR) in positron emission tomography (PET) requires accurate localization of the first interaction between incident gamma ray and detector. Improving the accuracy of this localization typically entails offsetting losses in detector efficiency, complexity, or cost. In this paper, we propose a solution that can localize scattered gammas without losses in detector efficiency, utilizing Prism-PET - a single-sided detector module with pixelated light guide. Using Monte Carlo simulations of gamma-detector interactions, we train a Convolutional Neural Network to predict the first interaction position of gammas incident on a detector block. 

Keywords: Prism, Light guide, PET, Compton
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 77

Poster Number:
M-08-077

Fragmentation model for Treatment Planning System of carbon ions implemented in a fast MC code (FRED) (#2081)

M. De Simoni1, 3, G. Battistoni5, F. Marta2, 3, F. Gaia1, 3, M. Michela6, 3, V. Patera2, 3, A. Sarti2, 3, A. Sciubba2, 4, G. Traini3, 6, A. Schiavi2, 3, M. Toppi4, 2, P. De Maria7

1 Sapienza, Università di Roma, Dipartimento di Fisica, Rome, Italy
2 Sapienza, Università di Roma, Dipartimento di Scienze di Base Applicate all'ingegneria, Rome, Italy
3 INFN, Section of Rome, Rome, Italy
4 INFN, Section of Frascati, Frascati, Italy
5 INFN, Section of Milan, Milan, Italy
6 Museo Storico della Fisica e Centro Studi e Ricerche E Fermi, Rome, Italy
7 Sapienza, Dipartimento di Scienze e Biotecnologie Medico-Chirurgiche, Rome, Italy

Content

The advent of general programming Graphics Processing Units (GPU) has prompted the development of MC codes that can dramatically reduce the plan recalculation time with respect to standard MC codes in CPU hardware. FRED (Fast paRticle thErapy Dose evaluator) is a software that exploits the GPU power to recalculate and optimize ion beam treatment plans. Rapidly recalculating a complete treatment plan within minutes, instead of hours, it paves the way for many clinical applications where the time-factor is important.
When developing the core algorithms, the goal is to balance accuracy, calculation time and GPU execution guidelines.  For what concerns proton beams, FRED is already used as a research tool in several clinical and research centers in Europe (Krakow, Trento, Maastricht, Lyon). The use of FRED for carbon ion, electron and photon beams is under development. In this article a new data-driven model of carbon ion fragmentation is described.
Nuclear interaction models at the energies of interest for particle therapy applications still have large experimental uncertainties. Many semi-empirical models have been developed to describe nucleus-nucleus interactions, and a considerable effort is continuously put into the development of these models and their benchmarking against the limited set of available experimental data. The development of the model implemented within FRED was based on the data taken at Ganil (laboratory of CAEN, France) and was benchmarked using the FLUKA MC software.

Keywords: TPS, Monte Carlo, GPU, nuclear fragmentation, carbon ion therapy
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 80

Poster Number:
M-08-080

Modelling of a bi-modal PET / Compton-camera system for non-pure positron emitters (#2228)

A. Bolke1, M. Zvolsky1, N. Kohlhase1, S. Seeger1, M. Schaar1, M. Rafecas1

1 Universität zu Lübeck, Institute of Medical Engineering, Lübeck, Schleswig-Holstein, Germany

Content

In this work, we investigate the combination of two imaging modalities, positron emission tomography (PET) and Compton-camera (CC) imaging. A prototype of simultaneous PET/CC acquisition was simulated. The design is based on the combination of two existing prototypes and allows a first investigation of the bi-modal acquisition of the radionuclides Zr-89 and I-124. Since the simulation framework GATE is not able to properly output coincidences of such a bi-modal system, a post-processing tool was developed. This includes custom coincidence sorting and classification of PET and CC data. The PET system is based on our small aquatic animal PET system MERMAID and consists of 18 detector modules arranged in a small ring of 28 mm radius. The CC planes are arranged parallel to the PET ring. In this paper, we present our post-processing pipeline and discuss the implications of the coincidence logic and the fraction of randoms on the data. The coincidences are classified in true, random and scattered coincidences. Random events are those coincidences which originate from two different nuclear decays. A coincidence that consists of detections caused by gamma radiation of different initial energies but originated from the same decay is classified as a random as well. First classifications in PET for I-124 show ~12 % random coincidences compared to ~5 % of Zr-89. An in-depth analysis of random coincidences with our current setup might result in improved coincidence sorting and help to pinpoint additional obstacles for future joint image reconstruction. The availability of such data is the prerequisite for developing novel image reconstruction approaches.

Keywords: PET, Compton Camera, Simulation, Multi-modal imaging, Radionuclide
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 83

Poster Number:
M-08-083

High Spatial-Temporal Resolution Electron Multiplier Equipped with Aluminum Oxide Nanopores: A Simulation Study (#2312)

M. Jamalizadeh1, S. Saramad1, A. Sanaat2, H. Zaidi2, 3

1 Amirkabir University of Technology, Faculty of Energy and Physics, Tehran, Iran (Islamic Republic of)
2 Geneva University Hospital, Division of Nuclear Medicine and Molecular Imaging, Geneva, Genève, Switzerland
3 Geneva University, Geneva University Neurocenter, Geneva, Genève, Switzerland

Content

Various studies reported on the use of Electron Multipliers (EM) as an electric device that is capable of multiplying a primary incident electron to a large shower of electrons. Recent technology for EM production with discrete dynodes suffer from moderately poor time and spatial resolution, sensitivity to magnetic fields and large dimensions. A series of simulations were carried out to simulate nanopores with a radius of 25 nm, 50 nm, and 75 nm and length of 10 μm with potential difference varying between 100 V and 450 V. The Casino toolkit was used for prediction of backscattered electron emission and secondary emission yield of the pore wall for electrons with various incident angles and energy. The Electrical Current in stationary mode and Charged Particle Tracing in time dependent mode from COMSOL software library was utilized to simulate electrons and their trajectories inside the nanopores. The MatLab software was employed for estimation of time, energy and related parameters for electron motion inside the nanopore by solving the equation of motion (non-relativistic). In nanowires, by increasing the voltage, the level of gain rises exponentially. For voltages higher than 300 V and in a channel with a radius of 25 nm, gain stability is observed because of the saturation of electrons inside the nanowire. It was observed that the peak energy of the output electrons increases from about 12 eV at 250 V to approximately 25 eV at 400 V. The time histograms of electrons for two voltages (250 V and 400 V) and oblique channel with 50 nm radius show that the highest numbers of electrons were collected at 4.7 ps and 4.4 ps, respectively. Our simulation study shows that the proposed EM can multiply the electrons by a factor of ~ 105.

Keywords: Nanopores, aluminium oxide, electron multiplier, Aluminum Oxide
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 86

Poster Number:
M-08-086

Experimental Results of Variable Pinhole Collimator for Organ Specific SPECT (#1388)

H. Cha1, K. Cho1, Y. - J. Jung1, S. Bae2, K. M. Kim3, M. Kim3, H. Lee2, K. Lee1

1 Korea University, Bio-convergence Engineering, Seoul, Republic of Korea
2 ARALE Laboratory Co., Ltd., Seoul, Republic of Korea
3 Korea Institute of Radiological and Medical Sciences, Radiation Devices Research Team, Seoul, Republic of Korea

Content

Single photon emission computed tomography (SPECT) is a nuclear medical imaging method that allows a user to view functional images of a patient. The collimator, an essential component of the SPECT limits the direction of the incident gamma rays, so that the distribution of gamma photons from the body can be observed. However, because collimators are composed of materials with high atomic number and density, it is difficult to produce complex shapes, and it is also challenging to modify the structure during the scan. Whereas a variable pinhole (VP) collimator comprises several thin tungsten layers, which have various hole sizes. Thus, pinhole parameters, such as hole diameter and acceptance angle, can be varied for the region of interest (ROI) by forming the desired pinhole shape via a combination of the holes. In this study, we implemented the concept of VP collimator and applied it in a SPECT system to enhance the performance. The collimator was composed of 5 layers of 170 mm in diameter and 1 mm thickness and had 6 holes of different sizes for each layer. Two line source phantoms (Tc-99m, 140 keV) with an internal diameter of 1 mm were used for the performance evaluation of the system. The phantoms were positioned 20 mm apart inside the ROI with a diameter of 50 mm at a position of 29 mm from the object center. By applying the VP collimator, the sensitivity and resolution performance were improved, achieving an FWHM value of 2.7 times and counts of 2.8 times those of the conventional SPECT system. In future research, we will enhance the system efficiency by conducting experimental test of the dual-head VP collimator SPECT system.

AcknowledgmentThis work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (No. NRF-2019M2D2A1A02059221)
Keywords: Variable Pinhole Collimator, single photon emission tomography, organ specific SPECT
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 89

Poster Number:
M-08-089

Hardware development of hybrid-sensor cameras and gantry for an adaptive SPECT system (#1922)

R. G. Richards1, M. Ruiz-Gonzalez3, M. May1, K. J. Doty2, 1, K. S. Kalluri4, N. Zeraatkar4, B. Auer4, M. A. King4, P. H. Kuo3, L. R. Furenlid3, 1

1 The University of Arizona, College of Optical Sciences, Tucson, Arizona, United States of America
2 The University of Arizona, Department of Biomedical Engineering, Tucson, Arizona, United States of America
3 The University of Arizona, Department of Medical Imaging, Tucson, Arizona, United States of America
4 University of Massachusetts Medical School, Department of Radiology, Worcester, Massachusetts, United States of America

Content

Stationary single photon emission computed tomography (SPECT) systems offer numerous advantages over rotating-head systems, but have one notable drawback in that their arrays of smaller cameras contain relatively more edges and gaps than a clinical two headed system. Scintillation events occurring at the edges of traditional photomultiplier-tube (PMT)-based SPECT cameras lose spatial resolution due to loss of lightsampling. Simulations using customized non-sequential raytracing scripts to model and analyze mean detector response functions (MDRFs) showed significant improvement in spatial resolution for hybrid-sensor cameras employing both silicon photomultipliers (SiPMs) and PMTs. The results inform the hardware design of AdaptiSPECT-C: a stationary clinical whole-brain SPECT imager with adaptive apertures for selective dynamic or high spatial resolution imaging. Its modular hybrid cameras use SiPMs to augment the PMTs and improve spatial resolution for position estimation tasks. SiPMs, having a small pitch and efficient fill factor, are employed in a border around the edges of each detector area. PMTs, being low cost and reliable, are packed in the center. The front end electronics are broken out into two main boards: one to drive and provide signal conditioning for the PMTs, and the other performing a similar function for the SiPMs. Ultimately, 81 total signal channels leave each camera as negative voltage pulses. AdaptiSPECT-C will have two equatorial rings of 10 cameras each and a quasi-vertex ring of 4 cameras, totaling 24. Modularity is the guiding design principle for the mechanical components of the cameras and ensures ease of assembly and field service in the completed system.

AcknowledgmentThis work was supported in part by NIH/NIBIB Grant R01-EB022521. R. Garrett Richards was partially supported by the Biomedical Imaging and Spectroscopy Fellowship, NIH grant T32-EB000809.
Keywords: gamma-ray detectors, molecular imaging, ray tracing, silicon radiation detectors, single photon emission computed tomography.
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 92

Poster Number:
M-08-092

Influence of pulse pile-up effects on material decomposition with photon-counting CT (#1402)

K. Murata1, K. Ogawa1

1 Hosei University, Faculty of Science and Engineering, Tokyo, Japan

Content

The aim of this study is to reveal an influence of pulse pile-up effects on material decomposition with a photon-counting CT system. The photon-counting CT has great advantages compared with a conventional system. Among them we focused on the ability of a material decomposition. However, a photon- counting CT system also has many issues to be solved. The most serious problem is a pulse pile-up effect. When multiple X-ray photons are simultaneously incident on a detector, the recorded spectrum is distorted. It should significantly degrade the material decomposition accuracy. Hence, we investigated influence of the pile-up effect on material decomposition, and feasibility of a spectral distortion-correction method. Using an analytical pile-up model, we performed simulations and found that accuracy of material-density measurements decreased with increasing X-ray intensity. We also found that the number of incident photons should be less than 0.1 in a detector deadtime in order to accurately measure densities of gold nanoparticle in 0.1wt% concentration solution. This low intensity is not practical and suggests a requirement of spectral distortion correction. Hence, we provided a correction method based on a least square method. In a simulation, our method successfully corrected the spectral distortion even in case of wide energy windows. The remained uncertainty was less than a few percent for a moderate X-ray intensity.

Keywords: medical imaging, computed tomography, photon counting, pulse pileup
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 95

Poster Number:
M-08-095

Rotatable LYSO-GAPD DEXA Detector for Providing High-Resolution (#1832)

H. Heo1, J. Yang1, J. Kang1

1 Chonnam National University, Department of Biomedical Engineering, Yeosu, Republic of Korea

Content

A rotatable LYSO-GAPD DEXA detector that can be configured into either a normal resolution/fast scan mode or a high resolution/slow scan mode, was proposed and examined. 3 × 3 × 2 mm3 LYSO pixel crystal was coupled to a 3 × 3 mm2 GAPD with optical grease. The versatile transformation for the high resolution mode is possible by rotating the DEXA detector (90 degrees) on its own axis, and the intrinsic resolution in this mode improves by ~33% from that of the regular mode. Dual energy X-ray spectra generated by K-edge Cerium filter were acquired for both acquisition modes. Imaging capability was evaluated by using the phantom with different hole sizes. No considerable changes in peak positions and peak-to-valley ratio (PVR) were observed for both acquisition modes. Dual energy peaks were located at around 588 mV and 1093 mV. PVR were ~3.4 and ~1.6 for low-energy and high-energy, respectively. However, there were noticeable changes in the count rate performance. The measured count per second values for the high resolution mode decreased from 869 to 613 and from 663 to 444 for low- and high-energy band, respectively, compared to the normal resolution mode one. Phantom images were successfully obtained to demonstrate the potential of proposed transformable design. Overall, the visible image quality obtained using high resolution mode were better than to those of normal resolution mode. These results demonstrate the rotatable LYSO-GAPD DEXA detector can allow the improved spatial resolution and a number of potential advantages.

Keywords: DEXA, LYSO-GAPD, High-spatial resolution
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 98

Poster Number:
M-08-098

Towards Unsupervised Domain Adaptation for Multi-sequence Cardiac MRI Segmentation (#1019)

S. Vesal1, M. Gu1, A. Maier1, N. Ravikumar1, 2

1 Friedrich-Alexander-Universität Erlangen-Nürnberg, Pattern Recognition Lab, Erlangen, Bavaria, Germany
2 University of Leeds, CISTIB, Centre for Computational Imaging and Simulation Technologies in Biomedicine,School of Medicine, Leeds, United Kingdom

Content

Deep learning methods are generally sensitive to domain shifts and often fail when deployed on a new set of data with different data distribution. Data annotation for every new domain is an expensive and time-consuming task, particularly in the medical imaging domain that requires clinical expertise. In this study, we introduce an unsupervised domain adaptation (UDA) approach that utilizes image and feature adaptation to segment cardiac anatomies in Late Gadolinium Enhanced-MR (LGE-MR) images without any annotation. We adapt the image features from the labeled balanced steady-state free precession (bSSFP) domain to LGE-domain. Our model has an imageto-image translation module using Cycle-GAN, which produces translated bSSFP images stylized as LGE, and a Dilated-Residual U-Net to segment style-transferred images. We also introduce a data augmentation technique based on extracting intermediate Cycle-GAN predictions to increase training data and diversity. The proposed UDA method along with the data augmentation approach achieved better segmentation accuracy and outperformed a supervised network that trained with limited training data.

Keywords: Multi-model Cardiac MRI, MRI, Cardiac Segmentation, Deep Learning, Domain Adaptation
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 101

Poster Number:
M-08-101

MRI-Compatible Data and Synchronization Links for PET Systems (#1352)

J. Lemay1, J. Bouchard1, A. Samson1, R. Lecomte2, 3, R. Fontaine1

1 Université de Sherbrooke, Interdisciplinary Institute for Technological Innovation 3IT, Sherbrooke, Québec, Canada
2 Université de Sherbrooke, Sherbrooke Molecular Imaging Center, Departement of Nuclear Medicine and Radiobiology, Sherbrooke, Québec, Canada
3 IR&T, Sherbrooke, Québec, Canada

Content

Multi-modal PET-MRI systems offer simultaneous structural and molecular imaging. The LabPET II is a PET technology based on several Embedded Signal Processing Unit (ESPU) boards synchronized with a central clock signal in order to enable coincidence detection between detectors. However, the data and synchronization links are not MRI-compatible. As a first experiment, copper cables with double isolation were tested. PCBs were assembled to simulate the path of the synchronization signals from a distribution unit to an ESPU. The delay and jitter between each end were measured with a high-speed oscilloscope when an Ethernet connection was turned off or transmitting at full speed on the other differential pairs of the same cable. The downside of this is approach is having to use isolators that are not MRI-compatible. A second experiment used the WhiteRabbit protocol and a PCB was designed to test its performance. This PCB primarily contains a ZYNQ system-on-module, an SFP connector, a mezzanine connector for different oscillators and houses 4 PET detection modules. The form factor was chosen to allow for a coil or an RF antenna to surround it to emulate the electromagnetic environment of an MRI. A jitter result of 10.7 ps RMS was obtained with the cables. The tests for the White-Rabbit system are on-going.

Keywords: Positron Emission Tomography, Magnetic Resonance Imaging, White-Rabbit Protocol
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 104

Poster Number:
M-08-104

Feasibility study on a single-channel microstrip transmission-line RF coil integrated with PET detector module for a 7T whole-body MRI system (#2170)

M. S. H. Akram1, M. Fukunaga2, F. Nishikido1, S. Takyu1, T. Obata1, T. Yamaya1

1 National Institute of Radiological Sciences (NIRS-QST), Chiba, Japan
2 National Institute for Physiological Sciences (NIPS), Aichi, Japan

Collaboration between National Institute of Radiological Sciences (NIRS-QST) and National Institute for Physiological Sciences (NIPS) in Japan.

Content

In this study, we developed a single-channel microstrip transmission-line RF coil with PET detector RF shield cage as the ground plane of the coil for a 7T whole-body MRI system. The shield cage was mounted with a 4-layer DOI detector and front-end readout electronics. The coil performance was compared with a conventional coil with one-layer copper plate as the ground plane of the coil. A cesium point source was used for the PET detector study. For MRI measurements, a cylindrical homogeneous doped water (NiSO4 and NaCl solution) phantom was used. Experiments were conducted inside the MRI bore for both single modality and simultaneous multimodality conditions. The PET measurement system was just outside the MRI room and RF shielded cables were carried out through the penetration panel of MRI room. PET data were taken for 10 minutes duration and for simultaneous measurements, a 10 minutes long MR gradient echo imaging sequence was used. We also used vendor provided RF pickup noise sequence to analyze the contamination of MR receive channel by outside noise that might come through the PET detector measurement system. In our analysis, we did not find any significant changes in MR image, PET energy and crystal flood diagram for both the single modality and simultaneous multimodality conditions. However, an approximate 10% increase in noise was found in the MR images for the simultaneous measurement condition. In the RF pickup noise study, we saw some pickup noise signals, however they were not visible in the MR images. This study was conducted without any noise reduction filter for the PET detector power sources. In future we plan to use noise reduction filter for this study.

Keywords: 7T MRI, PET/MRI, PET insert, microstrip RF coil
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 107

Poster Number:
M-08-107

Preliminary Performance Measurement of New Small Animal PET Based on GAGG Scintillator and SiPM. (#1029)

D. Kim1, O. Yutaka2, A. Choghadi2, K. Shimazoe2, H. Takahashi1, 2

1 University of Tokyo, Department of Nuclear Engineering and Management, Tokyo, Japan
2 University of Tokyo, Department of Bio Engineering, Tokyo, Japan

Content

Positron Emission Tomography (PET) is a functional imaging tool to obtain molecular mechanism and physiology. We are constructing a new PET system with small GAGG crystals with SiPMs. The focus of this work is to preliminarily measure the performance of the PET system using Geant 4 simulation.
The high-resolution PET scanner with SiPM and GAGG for small animals that we are building consists of 8 blocks of detectors composing an octagon. Each block is composed of 12x12 array of 1.6 mm x1.6 mm x 15 mm (1.9 mm pitch) GAGG:Ce crystals. KETEK PM1125 SiPM array is used for one detector module. Crystal and SiPM are coupled one to one to SiPM .The length between the center and detector surface is 30 mm so the radius of the system is approximately 60 mm. The radial and axial field of view (FOV) are 23 mm and 21.85 mm.
Multi-ray approach similar to the method was used for geometrical projection. Each crystal is divided into 12 pieces axially. Then, 144 rays are generated with a pair of crystals. Corresponding weight for each ray has been calculated with attenuation in the detector modules. Siddon’s ray tracing algorithm was used to find affected voxels per LOR. Geometric correction factor was calculated with plate source simulation in Geant 4. Image reconstruction was done with MLEM algorithm on 120×120×23 (0.5 mm × 0.5 mm × 0.95 mm voxel).
The best and the worst resolutions are 1 mm and 2.95 mm. The resolution is less than 2.5 mm in most of the region. Position error is shown to be negligible.
In addition to the resolution evaluated in this work, real measurement should contain more blur such as beta ray induced resolution degradation. Next, resolution recovery methods such as space variant deconvolution.

Keywords: PET, GAGG
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 110

Poster Number:
M-08-110

A High Resolution MRI Compatible Human Brain PET Insert Using Dual-ended Readout Detectors (#1147)

Z. Sang1, Z. Kuang1, X. Wang1, S. Wu1, N. Ren1, J. Gao1, D. Gao1, Q. Yang1, L. Cong1, T. Zeng1, Z. Liu1, Z. Hu1, J. Du1, Y. Yang1

1 Shenzhen Institutes of Advanced Technology, CAS, Shenzhen, China

Content

Combined PET/MRI is a powerful multi-modality tool for brain imaging. Dedicated brain PET has much higher sensitivity and spatial resolution that are required in neurological disease diagnosis and brain science research as compared to whole-body PET scanner. This report describes the design and development of a high-resolution MRI compatible brain PET insert using dual-ended readout detectors. The brain PET insert is cylindrical with 377 mm detector ring diameter and 330 mm axial field of view. The PET insert has 224 detector modules arranged in 8 detector rings with 28 detectors per ring. The detector module consists of a 26 × 26 LYSO array with a crystal size of 1.4 × 1.4 × 20 mm3 read out by two 10 × 10 SiPM arrays of 3.4×3.4 mm2 pixel size and 3.9 mm pitch from both ends. In total the PET insert uses 151424 crystals and 44800 SiPM pixels. The electronics of the PET insert consists of SiPM readout boards, front-end amplifier boards, single event boards, coincidence process board, synchronizer board and power supply boards. The performance of the detector was measured and the results show that all but the edge crystals can be clearly resolved, an average energy resolution of 16.12.3%, DOI resolution of 2.330.26 mm and timing resolution of 1.600.18 ns were achieved. The PET insert is expected to achieve a sensitivity of more than 10% at the center of the field of view and a uniform spatial resolution of 1.5 mm. The integration of the PET insert is ongoing and the first image is expected to be obtained by the time of the conference.

AcknowledgmentThis work is funded by the Peacock Innovation Team of Shenzhen (KQTD2016053117113327).
Keywords: Brain PET, MRI compatible, high spatial resolution, high sensitivity
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 113

Poster Number:
M-08-113

NEMA Characterization of the SAFIR Prototype PET-MR Scanner (#1423)

P. Bebie1, R. Becker1, A. Buck2, V. Commichau1, J. Debus1, G. Dissertori1, L. Djambazov1, A. Eleftheriou3, P. Fischer4, P. Khateri1, W. Lustermann1, C. Ritzer1, M. Ritzert4, U. Röser1, M. Rudin5, I. Sacco4, C. Tsoumpas6, G. Warnock3, B. Weber3, M. Wyss3, A. Zagozdzinska-Bochenek1

1 ETH Zürich, Institute for Particle Physics and Astrophysics, Zürich, Zürich, Switzerland
2 University of Zürich, Clinic of Nuclear Medicine, Zürich, Zürich, Switzerland
3 University of Zürich, Institute for Pharmacology and Toxicology, Zürich, Zürich, Switzerland
4 University Heidelberg, Institute of Computer Engineering, Heidelberg, Baden-Württemberg, Germany
5 ETH Zürich, Institute for Biomedical Engineering, Zürich, Zürich, Switzerland
6 Leeds Institute of Cardiovascular and Metabolic Medicine, University of Leeds, Leeds, United Kingdom

Content

The SAFIR (Small Animal Fast Insert for MRI) prototype insert is a pre-clinical Positron Emission Tomography scanner built to  acquire dynamic images simultaneously with a 7T Bruker Magnetic Resonance Imaging scanner. The insert is designed to perform with an excellent coincidence resolving time of 194 ps full-width-half-maximum and an energy resolution of 13.8% full-width-half-maximum. These properties enable the insert to acquire data at high activities of the order of 500 MBq. The axial coverage of the prototype is 35.6 mm.
In this study, we  have assessed the SAFIR prototype insert according to the National Electrical Manufacturers Association (NEMA) standard in terms of spatial resolution, sensitivity, scatter fraction and noise equivalent count rate. Data have been sorted with an energy window of 391-601 keV and a coincidence time window of 500 ps.
The spatial resolution (full-width-half-maximum) in the center of the scanner is  2.1 mm, 2.6 mm and 2.1 mm in radial, tangential and axial direction, respectively. The peak sensitivity is 1.1%. The noise equivalent count rate has been measured up to 537 MBq - not yet reaching its maximum. The total count rate increases with the activity resulting in 1.572 Mcps at the highest activity.
We acquired an in-vivo cardiac PET image from a mouse with an activity of 84.9 MBq and very short acquisition time of 5 s. The image clearly depicts the myocardium.

AcknowledgmentThis work was supported by the ETH Zurich Foundation through ETH Research Grant ETH-30 14-2. Charalampos Tsoumpas was supported by a Royal Society Industry Fellowship(IF170011). Geoff Warnock was funded by the Clinical Research Priority Program for Molecular Imaging of the University of Zurich (MINZ)
Keywords: PET/MRI, Pre-clinical PET, Dynamic Imaging, NEMA Characterization
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 116

Poster Number:
M-08-116

No gates, Real-Time Cardiac Reconstruction of Small Animal PET Data (#1507)

F. Arias-Valcayo, J. L. Herraiz, A. López-Montes, J. J. Vaquero, M. Desco, P. Galve, J. M. Udias1, 2, J. M. Udías

1 Universidad Complutense de Madrid, Nuclear Physics Group (GFN) and IPARCOS, Madrid, Spain
2 Hospital Clinico San Carlos, Health Research Institute (IdISSC), Madrid, Spain, Madrid, Spain

Content

Cardiac studies, especially in small animal, require very short frames in the ms range. Image modalities such as Magnetic Resonance (MRI) and Ultrasound provide high frame rate (15-1000 Hz) but their functional capabilities are very limited. Positron Emission Tomography (PET) provides true metabolic images, but the frame rate, so far, was limited and cardiac studies were performed rebinning counts from gated acquisitions. High sensitivity preclinical scanners enabling true high frame rate metabolic images can open a new horizon in cardiac studies thanks to the capability to quantitatively measure tracer uptake, revealing physiological phenomena not seen on low time resolution gated studies. We introduce a list-mode iterative reconstruction with image-guided effective filtering and time median filtering after automatic self-gating. The proposed method provides up to 50 Hz un-gated high resolution cardiac PET images.

AcknowledgmentWe acknowledge support from the Spanish Government (FPA2015-65035-P, RTC-2015-3772-2 and RTI2018-095800-A-I00), from Comunidad de Madrid (S2013/MIT-3024 TOPUS-CM, B2017/BMD-3888 PRONTO-CM) and European Regional Funds. This work is also supported by EU's H2020 under MediNet, a Networking Activity of ENSAR-2 (grant agreement 654002) and by NIH R01 CA215700-2 grant.
Keywords: List-mode, PET, Small animal, Cardiac imaging, high frame rate
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 119

Poster Number:
M-08-119

Parallelizing the Characterization and Optimizing Energy and Timing Resolutions of a Time-over-Threshold APD-based PET Scanner (#1625)

P. Buteau1, C. Thibaudeau2, E. Gaudin4, A. Samson1, R. Lecomte4, 3, R. Fontaine1, J. - B. Michaud1, 3

1 Université de Sherbrooke, Interdisciplinary Institute for Technological Innovation, Sherbrooke, Québec, Canada
2 IR&T Inc., Sherbrooke, Québec, Canada
3 CIUSSSE-CHUS, Centre de Recherche du CHUS, Sherbrooke, Québec, Canada
4 Université de Sherbrooke, Department of Nuclear Medicine and Radiobiology, Sherbooke, Québec, Canada

Content

The LabPET II avalanche photodiode (APD) based scanner is designed for submillimetric spatial resolution in Positron Emission Tomography (PET). The mouse scanner global performance can be characterized through the timing and energy resolutions of its ~6000 channels. By varying the channel parameters, their impact on the timing and energy resolutions can be assessed and their optimal operating point for each independent channel found. A fully automated timing probe-based method is proposed to allow a thorough characterization of the detection modules. Using only probe-to-channel coincidences, the approach parallelizes the timing resolution measurement of different configurations on all scanner channels. For energy resolution, we mathematically show that we can directly measure it from the time-over-threshold (ToT) histograms. This removes the need for a time-consuming non-linear transformation estimation that must be repeated for every new configuration. Our method shows optimal resolutions at minimal gain at the expense of sensitivity.

Acknowledgment

This work was partly supported by the Natural Science and Engineering Research Council of Canada (NSERC) and Le Fonds de recherche du Québec–Nature et technologies (FRQNT)

Keywords: positron emission tomography, performance characterization, timing resolution, energy resolution, optimization
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 122

Poster Number:
M-08-122

A Staggered 3-Layer DOI Detector with a 1 mm Crystal Pitch and 20 mm Thickness for a Spherical PET Geometry (#1704)

H. G. Kang1, M. Nitta2, G. Lovatti2, P. Thirlof2, K. Parodi2, F. Nishikido1, E. Yoshida1, T. Yamaya1

1 National Institute of Radiological Sciences (NIRS) in National Institutes for Quantum and Radiological Science and Technology (QST), Department of Nuclear Medicine Science, Chiba, Japan
2 Ludwig-Maximilians-Universitat Munchen (LMU), Department of Medical Physics, Garching, Germany

Content

The in-beam positron emission tomography (PET) scanner can provide useful information about particle beams such as the beam range and biological washout effect. Last year, a novel PET geometry was proposed for the Small animal proton Irradiator for Research in Molecular Image-guided Radiation-Oncology (SIRMIO) at LMU. Their simulation study showed that the spatial resolution and sensitivity could be optimized by employing a staggered 3-layer depth-of-interaction (DOI) detector with a 1 mm crystal pitch. In this study, we present a staggered 3-layer depth-of-interaction (DOI) detector for the SIRMIO project that is being developed at NIRS, in collaboration with LMU. The proposed detector consists of a LYSO array, acrylic light guide, and 8×8 SiPM array (Hamamatsu, S14161-9055, Japan). The LYSO crystal array has a pixel pitch of 1 mm and total thickness of 20 mm.  For the reflector material, BaSO4 is used. The SiPM anode signals are multiplexed by using a resistive network (multiplexing ratio 64:4). The timing signal is extracted from the common cathode signal. The multiplexed SiPM analog signals are digitized by using a CAMAC DAQ in coincidence mode. In tests, we were able to resolve most of the crystals in the flood map except the corner regions using this highly multiplexed staggered 3-layer DOI detector with 1 mm crystal pitch and 20 mm total crystal thickness. In the future, we will employ the detector for the realization of a small animal in-beam PET scanner prototype for the SIRMIO project.

Acknowledgment

This work is supported by the ERC under grant agreement 725539.

Keywords: Small animal PET, DOI, in-beam PET
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 125

Poster Number:
M-08-125

Preclinical PET scanners based in easyPET technology: new developments for full body mice imaging (#1765)

P. M. M. Correia1, A. L. M. Silva1, F. M. Ribeiro1, I. F. Castro1, 2, P. M. C. C. Encarnação1, I. Mohammadi1, 3, R. G. Oliveira1, F. M. Rodrigues2, A. Azevedo2, A. Sá2, A. I. Veloso4, A. C. Santos5, J. F. C. A. Veloso1

1 University of Aveiro, Physics Department, Institute for Nanostructures, Nanomodelling and Nanofabrication (i3N), Aveiro, Portugal
2 RI-TE - Radiation Imaging Technologies, Lda, Aveiro, Portugal
3 Islamic Azad University, Department of Basic Sciences, Faculty of Medicine, Sari, Iran (Islamic Republic of)
4 University of Aveiro, DigiMedia, Departament of Comunication and Art, Aveiro, Portugal
5 University of Coimbra, Institute for Clinical and Biomedical Research (iCBR), Area of Environment Genetics and Oncobiology (CIMAGO), Faculty of Medicine, Coimbra, Portugal

Content

EasyPET is an affordable PET scanner technology capable of real-time high quality in-vivo images. While the high price of preclinical PET scanners makes them unaffordable to many research centers, easyPET-based systems, using an innovative scanning method based on two rotation shafts for the movement of detector arrays, reduce the overall costs without compromising image quality.
This concept achieves high and uniform position resolution over the whole field of view (FoV), by minimizing scattered events and parallax errors due to depth of interaction (DoI) uncertainty, which are typical of ring-based PET systems. Also, this feature allows setting the detector modules very close to the subject, favouring sensitivity.  Full body mouse imaging is possible using only a small number of detector elements, capable of scanning billions of lines of response (LoRs) in few minutes. The easyPET.3D prototype uses two arrays of 64 detector cells, based on LYSO scintillators coupled to SiPMs, covering a 50 mm diameter × 73 mm long FoV. A dedicated frontend board processes the SiPM signals and sends the information to a microprocessor, responsible for the computer interface.
LoRs are organized in a List-Mode format and processed by a dedicated algorithm based in Maximum-Likelihood Estimation Methods (MLEM), running in GPU CUDA kernels, delivering results in few seconds. This speed gain allows real-time visualization of the reconstructed images during in-vivo PET scanning, with multiple advantages such as the possibility of identification and correction of mispositioning of the animal in the scanner FoV. An overview of the easyPET-based scanners, their features and results will be presented, as well as examples of preclinical images obtained with EasyPET.3D.

Patent, Universidade de Aveiro: PCT/IB2016/051487

Acknowledgment

This work was supported by projects POCI-01-0145-FEDER-016855, CENTRO-01-0247-FEDER-039880 (iPET), UID/CTM/50025/2019 and through COMPETE, FEDER, POCI and FCT (Lisbon) programs.

Keywords: Positron Emission Tomography PET, Preclinical, High Resolution, SiPM, GPU
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 128

Poster Number:
M-08-128

CardioPET, A High Resolution TOF-PET for Limited Angle Tomography (#1829)

E. Lamprou1, G. Cañizares1, M. Vergara1, 2, S. Aguilar1, F. Sanchez1, L. F. Vidal1, L. Hernandez1, A. Mascarell1, M. J. Rodriguez-Alvarez1, J. M. Benlloch1, A. J. Gonzalez1

1 Institute for Instrumentation in Molecular Imaging (i3M), Valencia, Spain
2 Department of Imaging and Pathology, Division of Nuclear Medicine, KU Leuven, Leuven, Belgium

Content

Recent developments in the fields of photodection, microelectronics and scintillation materials, made possible the development of PET detectors with accurate Time of Flight (TOF) capabilities. TOF information, not only results in an increased image Signal-to-Noise ratio, but also permits the development of novel PET scanners concepts, such as suitable for Limited Angle Tomography (LAT). Without TOF, these concepts show a poor performance due to the missing angular information. In this work, we describe a dedicated TOF-PET scanner, designed for heart imaging of patients under stress condition. The system is called CardioPET and is composed by 2 panels with an active area of 165 x 109 mm2. Each panel is based on 4 x 6 SiPM arrays of 8 x 8 elements each, directly coupled to a LYSO crystal of 3 x 3 x10 mm3. A total of 3072 resulted channels are read out using the commercially available TOFPET2 ASIC. The first evaluation of CardioPET shows that is capable of providing a system time resolution of 238 ps FWHM. In this work, we are showing system details as well as we introduce all performance evaluation planned to be carried out. Pilot images are also provided showing the performance of the system in terms of spatial resolution as well as SNR improvement in LAT applications. An in-depth evaluation of system capabilities will be carried out in the framework of this work.

Acknowledgment

This work was supported by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant Agreement No. 695536). It has also been supported by the Spanish Ministerio de Economía, Industria y Competitividad under Grant TEC2016-79884-C2-1-R.

Keywords: TOF-PET, Dedicated PET, LAT, SiPMs, TOFPET2 ASIC
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 131

Poster Number:
M-08-131

Design and Performance of a Preclinical PET Insert for Simultaneous PET/MRI at 9T (#1943)

A. J. Gonzalez1, M. Freire1, A. Gonzalez-Montoro1, L. F. Vidal1, L. Hernadez1, V. Carrilero2, R. Polo2, G. Pastor2, C. Molinos2, C. Correcher2, S. Junge2, P. Bruyndonckx2, M. Wideroe3, M. Heidenreich2

1 Institute for Instrumentation in Molecular Imaging (i3M-CSIC), Valencia, Spain
2 Bruker BioSpin, Ettlingen, Germany
3 Norwegian University of Science and Technology, Trondheim, Norway

Content

There exists several PET insert designs which are able to simultaneously work in the strong magnetic field of an MRI system. There are designs for human brain imaging, but specially for small animal imaging. Among these designs, there is only one design that uses monolithic crystals, the Bruker PET insert. In this work, we describe the design and performance of a new PET system that also uses monolithic crystals but with a very small design suitable for gradient coils with just 100 mm aperture. The new PET insert uses 24 LYSO blocks of 25x33x10 mm3, distributed in three rings. The PET inner diameter is about 60 mm, with an axial FOV of 101 mm.
The PET insert has been evaluated, and the results reported here, as a function of different MRI sequences. We did not observe any significant reduction in performance with respect to count rates or image quality. The performance and quality of the MRI was also preserved. The PET insert achieves a spatial resolution across the whole FOV of 0.65-0.75 mm FWHM, combined with a sensitivity of almost 12%. We will provide details of these tests. Regarding count rate capabilities, a NECR peak of 380 kcps was found at about 19.4 MBq. Initial tests with mice have also been carried out, confirming our quality findings with phantoms.

Keywords: PET/MRI insert, SiPM, monolithic crystals
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 134

Poster Number:
M-08-134

Optimized acquisition energy window in 89Zr imaging with multi-pinhole PET (#1997)

M. P. Nguyen1, M. C. Goorden1, C. Kamphuis2, F. J. Beekman1, 2

1 Delft University of Technology, Faculty of Applied Sciences, Section Biomedical Imaging, Delft, Netherlands
2 MILabs B.V., Utrecht, Netherlands

Content

89Zr is an important isotope for applications like immuno-oncology as it possesses a half-life that is very compatible with the pharmacokinetics of antibodies. Its major decay modes are positron emission (resulting in 511 keV annihilation photons, 23% probability) and prompt gamma emission (909 keV, 99% probability). Imaging of 89Zr is conventionally done with coincidence PET using the 511 keV photons, but image quality is affected by its positron range (1.23 mm mean, 3.9 mm max) and down-scatter from the 909 keV photons into the 511 keV photon peak. Recently, 0.75 mm resolution 89Zr imaging has been demonstrated based on positron range-free 909 keV prompt gamma imaging with the high-energy multi-pinhole VECTor system (MILabs B.V.). Here we optimize the acquisition energy window setting, taking into account the detector’s energy resolution and the fact that some scattered photons may also contain useful information. Monte Carlo simulations of a Derenzo phantom filled with different amounts of activity (2.5 and 40 MBq) were performed. Several energy window settings were investigated: windows with a 15%, 20%, and 25% width centered at 909 keV and a broad window of 600-1000 keV. The results indicated that for all activity levels, the energy window of 15% at 909 keV resolved the hot rods slightly better than the energy windows of 20% and 25% at 909 keV. Compared to the energy window of 15% at 909 keV, the energy window of 600-1000 keV offered a lower resolution for 40 MBq, but a better image for 2.5 MBq activity. Therefore, a narrow window centered at 909 keV is optimal in the case of high-count imaging, while a wide window of 600-1000 keV should be chosen for short scans or scans with low activity levels. This can be readily implemented since the system acquires projection data in list mode.

Keywords: Zr-89, pinhole, energy window
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 137

Poster Number:
M-08-137

Physical performance of SynchroPET ArterialPETTM, a human wrist PET prototype scanner for non-invasive arterial input function evaluation (#2077)

N. A. Karakatsanis1, E. K. Fung1, M. I. Akerele1, L. Pollenz2, A. Conticello2, A. McFarland2, D. Schlyer3, R. Gross2, Y. Sinelinkov2, T. Mariner2, M. Alessi2, J. W. Babich1, S. A. Nehmeh1

1 Weill Cornell Medical College, Department of Radiology, New York, New York, United States of America
2 SynchroPET, Inc., Stony Brook, New York, United States of America
3 Brookhaven National Laboratory, Upton, New York, United States of America

Content

Arterial blood sampling (ABS) is considered the gold-standard method for evaluating the arterial input function in dynamic PET studies. However, ABS is invasive and associated with a high complexity and risks. Alternatively, image-derived input function (IDIF) methods can be employed to simplify protocols and enhance adoption of dynamic PET in research and clinical practice. However IDIF methods require scanning for longer periods at the scanner bed, can be limited by the short axial field-of-view (FOV) of the vast majority of modern clinical PET systems and are susceptible to noise and partial volume effects due to the large ring diameter and large detector elements size of modern clinical PET systems. In this work we are evaluating the physical performance of the SynchroPET ArterialPETTM scanner (Stony Brook, NY, USA), a human wrist standalone cylindrical PET prototype system of 9 cm diameter ring and 1.73 cm axial FOV, capable of 4D PET imaging of the human wrist to enhance existing IDIF methods. The NEMA NU4-2008 performance evaluation protocol was selected to assess spatial resolution. system sensitivity and image quality. All PET data were acquired in list-mode, later sorted into 3D sinograms and reconstructed with an open-source 3D-OSEM algorithm with all data corrections included (decay, randoms, normalization, attenuation and scatter) when applicable. The average of radial and transaxial resolution was 1.49mm and 2.78mm FWHM at 5mm and 25mm radial distance from the center, respectively, while the respective axial resolution was 2.84mm and 4.69mm FWHM. The system sensitivity was 3.54 kcps/MBq. In terms of image quality, all hot rod sources were distinguishable with recovery coefficients of 21.71% and 96.21% for the 1mm and 5mm rod sources, respectively. ArterialPET can be utilized to extract IDIF data from human arteries in the wrist where a 2-5mm diameter is expected. Further clinical studies are needed to assess potential enhancement of IDIF methods.

Keywords: PET, input function, dynamic, NEMA, performance
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 140

Poster Number:
M-08-140

This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 143

Poster Number:
M-08-143

Results from a further scaled up prototype of a 1-millimeter resolution clinical PET system (#2185)

M. Chin2, 1, D. Innes1, C. S. Levin1, 2

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

Content

We have been developing a 1-millimeter resolution clinical positron emission tomography (PET) system dedicated to loco-regional (e.g. breast, head/neck, etc.) cancer imaging. Due to a dual-panel, patient-adjustable geometry and position-sensitive, pixelated scintillation detector design, the system enables high sensitivity in order to realize 1 mm reconstructed resolution. The basic detector unit comprises two position-sensitive avalanche photodiodes (PSAPD), each coupled to an array of 8 × 8 lutetium- yttrium oxyorthosilicate (LYSO) crystals which we refer to as a dual-LYSO-PSAPD module, and each panel consists of such 512 detector modules. Recently, we scaled up the system to increase a field-of-view (FOV) by 25%, and in this work, we characterize the scaled-up prototype system. Energy spectrum and coincidence time spectra are presented, and the 511 keV positron annihilation photopeak energy resolution and coincidence time resolution were measured to be 11.0 % and 25.6 ns, respectively. Lastly, we imaged an 22Na rod source, with 1.6 mm diameter, and the resulting reconstructed images are shown. Full-width-half- maximum (FWHM) values of line profiles on X-Z, X-Y, and Y-Z plane were 0.86 mm, 0.94 mm, and 1.25 mm, respectively.

Acknowledgment

This work is supported in part by Wallace H. Coulter Foundation, and Emerson Collective Cancer Research Fund (Award ID: 643319).

Keywords: PET, PSAPD, Organ-specific, LYSO
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 146

Poster Number:
M-08-146

Synthesizing Neuro MRI from PET/CT using Deep Learning (#2277)

G. Kim1, J. H. Jung1, Y. Choi1, H. K. Lim1, S. Lee2, M. Yun2

1 Sogang University, Department of Electronic Engineering, Molecular Imaging Research & Education (MIRE) Laboratory, Seoul, Republic of Korea
2 Yonsei University College of Medicine, Severance Hospital, Departments of Nuclear Medicine, Seoul, Republic of Korea

Content

PET scans are widely performed in conjunction with CT or MRI because it can provide complementary diagnostic information of various of diseases. However, due to repeated PET/CT and MRI examinations, patient's inconvenience and diagnostic costs are increased. The purpose of this study was to synthesize FLAIR MRI from FDG PET/CT of human brain using deep learning. For this purpose, convolutional neural network based on the U-net architecture, which had a series of stages between down-sampling layer and up-sampling layer and skip connections through each stage, was designed and implemented. The proposed network was trained and tested using 150 patient studies of spatially normalized FDG PET, CT, and FLAIR MRI. Six-fold cross-validation was performed with 90 studies excluding 60 studies for testing. To evaluate the performance of the proposed network, FLAIR MRI were synthesized using PET, CT and PET/CT and then the qualities were compared to those acquired using 3T MRI (reference) by using the peak signal to noise ratios (PSNRs) and the structural similarity indexes (SSIMs). The quality of FLAIR MRI synthesized using both PET and CT was superior to those obtained using PET or CT alone, and was comparable to that of the reference MRI. The PSNRs and SSIMs between the reference MRI and synthesized FLAIR MRI generated using PET, CT or PET/CT were 26.59 / 0.81, 27.28 / 0.81 and 27.73 / 0.83, respectively. The results of this study indicate that it is feasible to synthesize FLAIR MRI from FDG PET/CT images of human brain using deep learning method proposed in this study.

Keywords: Deep Learning, Neural Network, Image Generation, Machine Learning, Generative Adversarial Networks
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 149

Poster Number:
M-08-149

A New Image Reconstruction Technique with Limited View-angle Projection Data for BNCT-SPECT (#1054)

H. Inamoto1, S. Takeishi1, Y. Morizane2, S. Tamaki1, S. Kusaka1, F. Sato1, I. Murata1

1 Osaka University, graduate school of engineering, Suita, Japan
2 Osaka University, school of engineering, Suita, Japan

Content

BNCT is a new cancer therapy using 10B (n, α)7Li reaction induced by low energy neutrons. For practical use of BNCT, BNCT-SPECT system is being developed worldwide, to evaluate the treatment effect. In this study, we proposed a new image reconstruction technique for BNCT-SPECT with limited view-angle projection data using the Bayesian estimation, and performed numerical tests. As a result, it was confirmed that the proposed image reconstruction method works properly when the T / N ratio is infinite. On the other hand, when the T / N ratio was finite, it had been found that there are two kinds of errors: (1) "blur" due to interference between data, and (2) "artifact" that extended the image in the direction of the visual field. These errors act in a trade-off relation. To examine these errors, in the next step, it is necessary to develop a common index of the error evaluation for this kind of technique and to find the optimal solution for the number of screens.

Keywords: Bayesian estimation, BNCT-SPECT, image reconstruction, limited view-angle projection
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 152

Poster Number:
M-08-152

Multigrid Reconstruction Method for X-ray Fluorescence Computed Tomography (#1093)

B. Gao1, L. Van Hoorebeke1, L. Vincze2, M. Boone1

1 Ghent University, Department of Radiation Physics and Astronomy, Gent, Belgium
2 Ghent University, Department of Chemistry, Gent, Belgium

Content

X-ray fluorescence computed tomography is an imaging modality known for its capacity on mapping elemental distributions inside the sample. However, the self-absorption effect has made it challenging to obtain reconstructions of elemental density with high accuracy. To tackle this issue, the state of the art algorithms reconstruct not only the density distribution of elements, but also variables necessary for the modeling of self-absorption effect. Nevertheless, reconstructing these extra variables is computationally expensive and time-consuming, which has hindered the usage of those state of the art algorithms in 3D space. To ease the computational burden, a multigrid reconstruction technique has been proposed for X-ray fluorescence computed tomography. Multigrid refers to the different discretization grids, on which the elemental distributions and variables relevant to the modeling of self-absorption effect have been reconstructed. Through test, we find that an accurate density distribution can also be obtained by reconstructing the associated variables on a much coarser grid. As a significant computational speed-up can be achieved with multigrid reconstruction, such strategy could be beneficial for the full 3D reconstruction of the elemental density.

AcknowledgmentThe Special Research Fund of Ghent University is acknowledged for the financial support for B.G. and M.N.B under project number BOF17-GOA-015; We gratefully acknowledge NVIDIA for the donation of the GPU Titan Xp.
Keywords: X-ray Fluorescence Computed Tomography (XFCT), Tomography Reconstruction, Multigrid Reconstruction
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 155

Poster Number:
M-08-155

An Improved Fused EM Algorithm for PET Image Reconstruction (#1187)

Z. Hu1

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

Content

This paper presents an improved PET image reconstruction method based on an iterative method. PET is an advanced clinical imaging technique in nuclear medicine. Patients can be specifically injected with a radioactive tracer to detect activity at the molecular level in the injected tissue, but the tracer injection poses a potential radiation risk. However, the limitation of the injection dose and the acquisition time leads to a relatively poor spatial resolution and high noise level in the positron emission imaging, which makes the quantitative interpretation of PET images difficult. Therefore, the research and development of new low-dose PET imaging methods can not only ensure the quality of PET imaging but also reduce the dose of harmful radiation, which would have important scientific significance and application prospects in the field of medical diagnosis. The reconstruction framework proposed in this paper is based on the traditional maximum likelihood expectation maximization (ML-EM) method and the bootstrapping filter method. First, the ML-EM method is used to directly reconstruct an image, which is used as a guide. Second, the Anscombe transform and a filter are applied to the projection data, followed by inverse transformation. The processed data are also reconstructed with ML-EM, and the resulting image is used as a guided image. The two images generated in the first two steps are fed into a guide filter to obtain the final image. For simplicity, the proposed improved fusion EM method is named IF-EM. In this paper, we present qualitative and quantitative results from simulation data, which show that the proposed method has good statistical performance.

Acknowledgment

This work was supported by the National Natural Science Foundation of China (81871441), the Shenzhen International Cooperation Research Project of China (GJHZ20180928115824168).

Keywords: Positron emission tomography (PET), Image reconstruction, Anscombe transform, Guided filter.
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 158

Poster Number:
M-08-158

Improvement and Evaluation of General Simultaneous Motion Estimation and Image Reconstruction (G-SMEIR) (#1225)

S. Zhou1, Y. Chi1, J. Wang2, M. Jin1

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

Content

To pursue the ultimate performance of 4D cone-beam computed tomography (CBCT), a general simultaneous motion estimation and image reconstruction (G-SMEIR) method has been proposed to mitigate the local optimum trapping problem of the original SMEIR method. In this work, we investigated the two strategies to improve the efficiency of image domain motion estimation based on the demons algorithm: multi-resolution strategy (MRS) and multi-step strategy (MSS). Both strategies achieved about 2.5 times faster convergency than the original demons method, while MRS works better for registration of large motions at the same iteration number. Using MRS of demons algorithm, G-SMEIR with flexible combinations of projection-domain and image-domain motion estimation achieved much better 4D reconstruction performance compared to SMEIR using both 4D XCAT phantom data and real patient data. The quantitative measures of root mean squared error (RMSE) and structural similarity index (SSIM) averaged over all ten respiratory phases of XCAT are 9.018x10-4 and 0.9648 for G-SMEIR, which are much better than SMEIR with statistical significance (p < 0.5). For the patient data, G-SMEIR provides less blurred images than SMEIR, thus leading to a better-defined tumor.

Acknowledgment

This work was supported in part by the U.S. National Institutes of Health under Grant No. NIH/NCI R15CA199020-01A1 and NIH/NIBIB R03EB021600-01A1.

Keywords: 4D Cone-beam computed tomography (CBCT), 4D reconstruction, Motion estimation
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 161

Poster Number:
M-08-161

SPECT Angle Interpolation Based on Deep Learning Methodologies (#1411)

C. Chrysostomou1, L. Koutsantonis1, C. D. Lemesios1, C. N. Papanicolas1

1 The Cyprus Institute, Nicosia, Cyprus

Content

A novel method for SPECT angle interpolation based on deep learning methodologies is presented. Projection data from software phantoms were used to train the proposed model. For evaluation of the efficacy of the method, phantoms based on Shepp Logan, with various noise levels added were used, and the resulting interpolated sinograms are reconstructed using Ordered Subset Expectation Maximization (OSEM) and compared to the reconstructions of the original sinograms. The proposed method can quadruple the projections, and denoise the original sinogram, in the same process. As the results show, the proposed model significantly improves the reconstruction accuracy. Finally, to demonstrate the efficacy and capability of the proposed method results from real-world DAT-SPECT sinograms are presented.

AcknowledgmentWe will like to thank Prof Dr Maintas Demetrios from the Athens Medical Center, Institute of Isotopic Studies for providing the DAT-SPECT sinograms.

 
Keywords: SPECT, Denoise, Sinogram, Angle Interpolation, Deep Learning
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 164

Poster Number:
M-08-164

Investigation of spatial-temporal kernel method for dynamic imaging in short and long range PET scanners (#1430)

Y. Li1, Y. Zhao2, Y. Lv2, J. Zhao1

1 Shanghai Jiao Tong University, School of Biomedical Engineering, Shanghai, China
2 United Imaging Healthcare, Shanghai, China

Content

Dynamic PET imaging has a limited time resolution, and always suffers from poor signal to noise ratio (SNR) due to low-count statistics from short time frames and the ill-posed nature of Expectation-Maximization (EM) based reconstruction algorithm. Incorporating of prior information from long time frames and/or anatomical images, the kernel method effectively improves the SNR of the dynamic images from very low-count data. In this study, we have implemented a spatial-temporal kernel method and applied it to the data acquired from the 194-cm long range total body PET/CT scanner (uEXPLORER). We also simulated dynamic data of short range PET scanner by using partial units of the uEXPLORER.  The results demonstrated that the spatial-temporal kernel method outperformed the kernelized EM and Ordered Subsets Expectation Maximization (OSEM) for frames as short as 0.1-second, and the respiratory and cardiac motions could be observed visibly.

Keywords: Dynamic imaging, spatial-temporal kernel method, PET image reconstruction
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 167

Poster Number:
M-08-167

Modeling of Depth of Interaction with Inter-crystal Scattering for PET Reconstruction (#1459)

L. Szirmay-Kalos1, D. Varnyú1, M. Magdics1, B. Tóth1

1 Budapest University of Technology and Economics, Budapest, Hungary

Content

Depth of Interaction (DOI) PET scanners extend the data stored about photon detection events by the depth of the photon's absorption in the detector crystals.
Popular DOI-based reconstruction methods disregard photon scattering in the detectors and simply modify the endpoints of the Lines of Responses (LORs) according to the depth of the absorption. However, this simplification reduces the accuracy of the resulting image. This paper proposes the incorporation of inter-crystal scattering into DOI-based PET reconstructions. The transfer probabilities caused by inter-crsytal scattering are determined off-line with Monte Carlo simulation and are built into the system matrix as a factored component.

AcknowledgmentThis work has been supported by OTKA K--124124, Vekop-221-16, and by the EFOP-3.6.1-16-2016-00014 project financed by the Ministry of Human Capacities
Keywords: PET, DOI, ML-EM, Monte Carlo
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 170

Poster Number:
M-08-170

Scatter correction for missing LORs in list-mode rigid motion correction (#1486)

A. Miranda1, S. Beer2, S. Stroobants1, S. Staelens1, D. Elmenhorst2, T. Kroll2, A. Bauer2, J. Verhaeghe1

1 University of Antwerp, Molecular Imaging Center Antwerp, Antwerp, Belgium
2 Forschungszentrum Jülich, Institute of Neuroscience and Medicine (INM-2), Jülich, Germany

Content

Scatter correction is a standard correction performed in PET reconstruction to improve quantification. The single scatter simulation (SSS) generates a sinogram with the estimated number of scatter events in the standard sinogram space. However, when list-mode motion correction is performed using line of response (LOR) repositioning, some LORs might fall outside the standard sinogram space, and thus the standard SSS does not provide scatter correction factors for these LORs. Instead of extending the SSS sinogram to contain all these possible LOR outside the sinogram space we propose here an approximation. First, from the standard SSS sinogram a scatter image is reconstructed and then forward projected to the LORs outside the standard space to estimate the corresponding scatter correction factors. We validated this approach using a contrast phantom scanned at 3 different orientations. In addition, a moving image quality phantom was scanned and corrected for motion, with scatter correction using the proposed approach. For the contrast phantom, the approximated scatter sinograms resembled the original SSS scatter sinogram. As a result, the quantification in hot and cold regions minimally changed when the approximated scatter sinogram was used (<2 kBq/cc difference) instead of the original SSS scatter sinogram. For the moving image quality phantom, compared to only scatter correction for LORs in the standard sinogram space, spill-over in cold regions of water and air was reduced by 18% and 69% respectively, using our approach. To conclude, approximating the scatter correction for LORs that fall outside the sinogram space using our proposed approach improves image quantification and allows the use of the scatter sinogram calculated with standard methods.

AcknowledgmentThis project is part of the ERA-NET NEURON project SleepLess supported by BMBF (01EW1808) and FWO under the frame of Neuron Cofund.
Keywords: PET, Scatter Correction, Motion Correction
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 173

Poster Number:
M-08-173

Hyper Parameter Selection for Bayesian Image Reconstruction by Mimicking Physical Crystallization (#1510)

Y. Gao1, Z. J. Liang1, 2, S. Lu1, Y. Shi1, S. Chang1, W. Hou3

1 StonyBrook University, Radiology, Stony Brook, New York, United States of America
2 StonyBrook University, Biomedical Engineering, Stony Brook, New York, United States of America
3 StonyBrook University, Family, Population and Preventive Medicine, Stony Brook, New York, United States of America

Content

Although Bayesian theory has been successfully applied for count-limited medical image reconstruction in the past two decades, its wide applications in clinic has been hampered by its hyper parameter β , which is determined by a trial-error style.  To eliminate the cumbersome style, this work aims to present a selection method by mimicking the physical model of cooling down the temperature steadily for a high-quality crystal.  From the basic Bayes’ law, the physical meaning of hyper parameter can be interpreted as the ratio of the data uncertainty (or variance α) and the prior tolerance (or σ) by formulating the probability distribution functions of the data fidelity and prior expectation.  Inspired by this idea, the prior tolerance σ can be treated as the temperature of the texture patterns, and one look-up table σ-β was constructed to guide the strength of β in reconstruction.  Similar with crystallization, the reconstruction time (or iteration number) is also an important factor in decreasing the temperature, and thus the iteration-σ was also investigated.  Both simulated sinograms from high-quality patient CT images and real clinical sinograms from patients were used for the look-up table construction and testing.  The adaptive β strategy suggest smaller prior strength at starting point of iterative reconstruction process and stronger prior at the ending point.  For testing with the simulated phantom/clinical sinograms, comparable quantity images were obtained with that using the empirical trial-error selection of hyper parameter.  Testing on the real clinical sinograms, comparable quantity images were obtained.  The adaptive β strategy showed a speed up of the reconstruction by several times than the fixed β strategy.

AcknowledgmentThis work was partially supported by the NIH/NCI grant #CA143111 and #CA206171.
Keywords: Bayesian image reconstruction, hyper parameter selection, iteration number
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 176

Poster Number:
M-08-176

Monte Carlo Simulation of Myocardial Perfusion Imaging of Tracer Dynamics with Cardiac and Respiratory Motion Using Continuous Rotating Gamma Camera (#1568)

U. M. Shrestha1, Y. Huh1, G. T. Gullberg1, Y. Seo1

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

Content

Myocardial perfusion imaging (MPI) with single photon emission computed tomography (SPECT) is routinely used for cardiac stress testing. Recently, our group showed its potential going from 3D visual qualitative image analysis to 4D spatiotemporal reconstruction of dynamically acquired data to capture the time variation of the radiotracer concentration and estimated quantitative myocardial blood flow (MBF) and coronary flow reserve (CFR). In dynamic cardiac imaging with conventional slowly-rotating clinical SPECT scanners, the emitted photons are detected and recorded continuously by the camera detector as it rotates. It is difficult to obtain consistent tomographic data in order to solve the dynamic inverse problem frame-by-frame. However, by estimating coefficients of spatiotemporal basis functions directly from projection measurements, one can fairly accurately measure the wash-in and wash-out kinetics of a myocardial perfusion SPECT tracer. In this work we develop a spatiotemporal image reconstruction algorithm using splines to explicitly model the temporal change of the tracer and Gaussian basis functions to take into account the cardiac and respiratory motions for the data. A list-mode Monto Carlo simulation is performed such that each event has spatiotemporal information as well as tags for cardiac and respiratory phases. The accumulated photons are binned into projections for cardiac and respiratory phases over short time intervals. Reconstruction results are presented, showing the dynamics of the tracer in the myocardium as it deforms due to the cardiac beating, and is displaced due to the respiratory motion. These results in 6D are then compared with our 4D-spatiotemporal reconstruction method that models only the temporal changes of the tracer activity.  

Acknowledgment

The study was supported in part by the National Institutes of Health under grant R01 HL135490.

Keywords: SPECT, Myocardial Perfusion Imaging, Monte Carlo Simulation, Reconstruction Algorithm, List Mode
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 179

Poster Number:
M-08-179

Epipolar-Constrained Optical Flow Triangulation for the Interior Problem in CBCT (#1620)

D. Punzet1, 4, R. Frysch1, 4, E. Khosroshahi1, 4, O. Beuing3, 4, O. Speck2, 4, G. Rose1, 4

1 Otto-von-Guericke University Magdeburg, Institute for Medical Engineering, Magdeburg, Saxony-Anhalt, Germany
2 Otto-von-Guericke University Magdeburg, Department of Biomedical Magnetic Resonance, Magdeburg, Saxony-Anhalt, Germany
3 University Hospital Magdeburg, Institute for Neuroradiology, Magdeburg, Saxony-Anhalt, Germany
4 Forschungscampus STIMULATE, Magdeburg, Saxony-Anhalt, Germany

Content

When looking at the sequence of cone-beam x-ray projection images of an object, humans can intuitively make some assumptions about the shape and size of the scanned object judging by its motion field during the acquisition. This works even in highly truncated scenarios by intuitively extrapolating the observed part of the motion field to the occluded outer regions. Based on this observation, we present a method that allows to infer knowledge about the shape and size of the object under investigation even far outside of the scan field of view of a truncated cone-beam x-ray acquisition. By making use of the optical flow vectors computed from a sequence of projection images, the correspondences between points of interest in consecutive projection frames are established and triangulated back to locate their projective origins in 3D space. In a previous work we have shown as a proof of concept that this can be used as an estimation of the patient’s maximum extent. Here, we add further epipolar constraints to the general method which allows for obtaining more precise structural and shape information about the patient outside of the commonly reconstructable scan field of view. Results show that from severely truncated clinical projection data of the human skull, this method enables to locate parts of the anterior skull far outside of the scan field of view and therefore can be used to e.g. optimize the fit of an extrapolation or initialize iterative reconstruction methods.

AcknowledgmentThis work was conducted within the context of the International Graduate School MEMoRIAL at Otto von Guericke University (OVGU) Magdeburg, Germany, kindly supported by the European Structural and Investment Funds (ESF) under the programme Sachsen-Anhalt WISSENSCHAFT Internationalisierung
(project no. ZS/2016/08/80646). The work of this paper is also funded by the German Ministry of Education and Research (BMBF) within the Forschungscampus STIMULATE under grant number 13GW0095A.
Keywords: optical flow, triangulation, interior problem, epipolar geometry, CBCT
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 182

Poster Number:
M-08-182

Study on reconstruction accuracy of compressed sensing-based algorithm for translational trajectory CT (#1684)

Y. Mitsuya1, M. Uesaka1, H. Takahashi1

1 The University of Tokyo, School of Engineering, Bunkyo-ku, Japan

Content

We have developed a new reconstruction algorithm based on compressed sensing for CT using a translational trajectory. In this study, we further investigated the effect of the projection conditions on the reconstruction accuracy through numerical calculations using the developed algorithm. In particular, the effects of the fan- beam source aperture angle and the number of projections were evaluated quantitatively. In addition, for the future experimental demonstration, we are implementing an X-ray detector using a photodiode-array and a scintillator which are available for projection in translational trajectory.

Keywords: CT, Reconstruction, Compressed Sensing
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 185

Poster Number:
M-08-185

Anatomically-aided PET image reconstruction using GLCM-based kernel method (#1776)

D. Gao1, J. Gao1, C. Zhang1, Z. Kuang1, X. Wang1, Q. Yang1, S. Wu1, D. Liang1, Y. Yang1, Z. Hu1

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

Content

Positron emission tomography (PET) can image biological metabolism in vivo using radioactive tracers. It’s significant for reducing the injection dose of radioactive tracers, scan time, scan cost and improving the temporal resolution of dynamic PET to study image reconstruction method of low-count PET projection data. Inspired by the feature extraction methods of radiomics, we proposed GLCM-based (Gray-Level Co-occurrence Matrix) kernel method for low-count PET image reconstruction. The proposed method applies texture features, contrast, homogeneity and dissimilarity, of anatomical image as priori information to the forward model of PET projection of kernel-based image reconstruction method. Computer simulations show that the presented anatomically aided algorithm give better image quality and lesion Contrast Recovery Coefficient performance of low-count projection data than the MLEM with post Gaussian filter algorithm. Furthermore, our work provides a novel method of applying texture features of anatomical images to PET image reconstruction.

Acknowledgment

The authors would like to thank Dr. Jinyi Qi and Dr. Guobao Wang of the Department of Biomedical Engineering at UC Davis for providing some program code. This work was supported by the National Natural Science Foundation of China (81871441), the Natural Science Foundation  of  Guangdong  Province  in  China (2017A030313743), the Guangdong International Science and  Technology  Cooperation  Project  of  China (2018A050506064), the Guangdong Special Support Program of China (2017TQ04R395) and the Basic Research Program of Shenzhen in China (JCYJ20160608153434110).

Keywords: positron emission tomography (PET), image reconstruction, Gray-Level Co-occurrence Matrix (GLCM), kernel method
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 188

Poster Number:
M-08-188

Simultaneous Attenuation Correction, Scatter Correction, and Denoising in PET Imaging with Deep Learning (#1899)

J. Hu1, W. Whiteley1, X. Zhang1, C. Zhou1, V. Panin1

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

Content

Low radiation dose is desirable in PET/CT imaging. The delivered dose originates from both CT scans and injected PET radioisotopes. CT data is used for attenuation and scatter corrections in PET image formation. A standard PET dose is usually needed to generate PET images of clinical quality so that physicians can make diagnosis with confidence. In this work, we eliminated CT scans and reduced the PET dose while maintaining image quality by performing simultaneous attenuation correction, scatter correction, and denoising using a deep learning approach. We trained a multi-layer convolutional neural network (CNN) with non-attenuation corrected, non-scatter corrected, and low dose PET images as input, and fully corrected standard dose PET images as labels. After the CNN is trained, it is used to generate fully corrected standard dose PET images from low dose PET data alone. This capability may make CT scans unnecessary and reduce dose from radioisotopes significantly in PET imaging. We validated our methodology with patient data. The results showed that attenuation correction, scatter correction, and denoising can be performed simultaneously using the deep learning method.

Keywords: Positron emission tomography, Deep learning, Attenuation correction, Scatter correction, Denoising
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 191

Poster Number:
M-08-191

A Newton-Raphson Accelerated Iterative Reconstruction Method (#2026)

A. Sideri1, 2, E. Stiliaris1, 3

1 National & Kapodistrian University of Athens, Department of Physics, Athens, Greece
2 National & Kapodistrian University of Athens, Medical School, Athens, Greece
3 Institute of Accelerating Systems & Applications (IASA), Athens, Greece

Content

The greatest advantage of iterative techniques in the computed tomography is the ability to produce better images than other analytical methods in cases, where fewer and not uniformly distributed projection data are available.  Following the general iterative approach of the Algebraic Reconstruction Technique (ART) enhanced with convergence acceleration schemes derived from the Newton-Raphson methodology, a new type of a reconstruction algorithm for tomographic images is proposed. The Cost Function connected to the minimization procedure of this algorithm operates on the squared differences between the measured and reconstructed sinograms, taking into account matrix derivatives and their correlations with respect to neighborhood rays and projection angles. In addition to the formalism, the quality of the proposed reconstruction technique tested with sparse and noisy data will be presented and its convergence speed will be discussed with respect to other, well established methods.

Keywords: Iterative Reconstruction Techniques, Cost Function, Matrix Derivative, Newton-Raphson Optimization
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 194

Poster Number:
M-08-194

Histo-Projection TOF Data Non-Rigid Motion Estimation and Correction (#2078)

V. Panin1, D. Bharkhada1, W. Whiteley1

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

Content

With the advancement of PET technology, scanners with a very good time-of-flight (TOF) resolution became available. Now scanners of long axial extent are gaining attention. Although raw data sizes increase significantly with such scanners, good TOF resolution allows substantial data compression without any loss of spatial resolution. For example, the histo-projections with relatively coarse sampling in the TOF direction in the new Siemens Biograph Vision PET/CT scanner.
The patient motion during the scan is unavoidable. The pattern caused by breathing may result in a relatively large displacement of organs and the consequent blurring of clinically relevant features in regions affected by the motion. As was shown in our previous work, non-rigid motion correction can be performed in the quasi image space of histo-projections. The TOF locality property can be used to locally perform motion correction; that is, the approximation of motion as locally rigid on a scale of TOF resolution. In this work we investigate motion estimation in the TOF data space from histo-images by histogramming the data into a 3D array consistent (in geometry and size) with the reconstruction image. The histo-image estimated motion is then used in histo-projections motion correction.
Initial results using an XCAT digital phantom showed that the presented methodology accommodates for changes in non-rigid body movements for a typical pattern of patient motion.

Keywords: PET, motion correction, time-of-flight
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 197

Poster Number:
M-08-197

Joint Sparse Coding-Based Super-Resolution PET Image Reconstruction (#2157)

X. Ren1, S. - J. Lee1

1 Paichai University, Department of Electronic Engineering, Daejeon, Republic of Korea

Content

This paper presents a comparative study of the effects of using joint sparse coding (JSC) for regularized super-resolution (SR) PET reconstruction. With an assumption that a limited number of high-resolution (HR) PET images are available for a joint training dataset for JSC, we attempt to improve the accuracy of sparse coding (SC) based SR reconstruction in conventional non-HR PET imaging. Here we also assume that the anatomical (CT or MR) and PET images acquired from the same patient lie in coupled feature spaces. The images in one feature space can be transformed into corresponding images in the other feature space by a common mapping function. In this case, the images in the coupled feature spaces have a common sparse representation in terms of the specific dictionaries that are jointly trained, which is the main key to the JSC method. We implemented the penalized-likelihood SR reconstruction algorithm whose penalty term is modeled as JSC, and compared with our previous method using the single dictionary-based SC penalty. The experimental results demonstrate that our proposed JSC method clearly outperforms the standard SC method by more accurately restoring the fine details that are often missed by the standard SC method.

Acknowledgment

This work was supported by the National Research Foundation (NRF) of Korea under Grant 2016R1D1A3B04933319.

Keywords: tomographic reconstruction, super resolution, sparse coding, joint sparse coding, penalized-likelihood methods
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 200

Poster Number:
M-08-200

A Comparison of Deep Learning Approaches to Attenuation Map Generation for Myocardial Perfusion SPECT (#2388)

P. Faley1, 2, X. Ding1, W. Gohn1, F. Massanes1, H. A. Vija1

1 Siemens Medical Solutions, USA, Molecular Imaging, Hoffman Estates, Illinois, United States of America
2 University of Notre Dame, Notre Dame, Indiana, United States of America

Content

Single photon emission computed tomography (SPECT) is a useful imaging technique with many applications. Computed tomography (CT) can enhance SPECT imaging by providing attenuation correction, increasing the accuracy of scans. However, CT is not always a viable option due to differences in the availability of machines and increased patient exposure to radiation. The purpose of this study is to explore the use of multiple different deep learning models for constructing attenuation maps directly from SPECT scans. Specifically, we examine the 2D U-Net, 3D U-Net, and V-Net architectures. These models are each fully convolutional networks with an encoder and decoder component. The main difference between these approaches is that U-Net architectures utilize concatenation of downsampled features during upsampling, while V-Net architectures instead utilize addition. Additionally, we explore the use of an adversarial loss metric. A dataset of SPECT/CT images from 160 myocardial perfusion studies conducted at Brigham and Women's Hospital in Boston, Massachusetts was used for training and testing the models. SPECT images were used as model inputs, while CT-based attenuation maps were used as targets for the model. We found that all models provided accurate attenuation maps, with the 3D U-Net with no adversarial component performing the best on the task.

Keywords: Deep Learning, SPECT, SPECT-CT, Myocardial Perfusion Imaging, Synthetic Attenuation Map
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 203

Poster Number:
M-08-203

Crystal Area Segmentation for a Scintillation Detector based on Convolutional Neural Network (#1109)

S. Leem1, B. Yu2, H. Cha1, K. Cho1, C. Kang1, 2, J. Lee2, S. Bae2, H. Lee2, K. Lee1

1 Korea University, School of Biomedical Engineering, Seoul, Republic of Korea
2 ARALE Laboratory Co., Ltd, Seoul, Republic of Korea

Content

Crystal area segmentation is one of the critical procedures for decoding the detector module coupled with scintillation crystal. However, the blurring effect makes the decoding procedure challenging. For precise decoding, we propose a crystal area segmentation method based on convolutional neural network (CNN). The method is divided into training stage and evaluation stage. In the training stage, data set was extracted from five flood maps in blocks. These blocks went over preprocessing with bandpass filter (BPF) and thresholding. Then the processed blocks were used to train and test the CNN. In evaluation stage, flood map from positron emission tomography (PET) scanner ‘micro crystal element scanner (MiCES)’ was tested. The method detected 461 peaks out of 484 peaks while existing method detected 441 peaks. The proposed algorithm detected center peaks almost perfectly and improved detectability of boundary peaks. However, due to the quality of training data sets, boundary peak detection was dependent on noise component of the flood map. In further studies, we will improve the boundary peak detection and evaluate more flood maps acquired by detectors based on silicon photomultipliers (SiPM) under development in our research group.

AcknowledgmentThis work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (No. NRF-2019M2D2A1A02059221) and National Institutes of Health under grants CA-74135, CA-86892 and EB0217. The authors would like to acknowledge the helpful support from Dr. Wiliam Hunter and Dr. Robert Miyaoka of the University of Washington.
Keywords: Crystal segmentation, Convolutional Neural Network, Classification
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 206

Poster Number:
M-08-206

Automatic attenuation map estimation from SPECT data only for DaTscan using a convolutional neural network (#1267)

Y. Chen1, M. C. Goorden1, F. J. Beekman1, 2

1 Delft University of Technology, Radiation Science and Technology, Delft, Netherlands
2 MIlabs B.V., Utrecht, Netherlands

Content

For brain SPECT, correction for photon attenuation in the patient head is essential to obtain images which provide quantitative information on the activity concentration. As attenuation maps from a CT or MRI may not always be available and could suffer from registration errors, an attenuation correction strategy based solely on SPECT data is beneficial. Earlier, we showed that a CNN-based approach using emission-data only allows to estimate the attenuation maps for full brain perfusion imaging with G-SPECT-I, a focusing stationary multi-pinhole SPECT system. Here we validate if a similar CNN approach can be applied to focused DaTscan imaging in which scans are confined in the axial direction to increase count yield from the striatum. Our work was validated in a Monte Carlo simulation study of the G-SPECT-I geometry. Data acquired in list mode was used to fully utilize photons from a wide energy range (primary and scattered) that together contain unique information about the tissue attenuation coefficient. From this data, multiple tomographic image reconstructions were performed at different energy windows. Locally extracted 4D patches (3D spatial plus 1D energy) centered at each voxel of these images were used as input for the CNN which was trained to predict the attenuation coefficient of the corresponding voxel. Results show that attenuation maps essential for DaTscan attenuation correction could be estimated using the CNN despite the fact that only part of the brain was imaged for a focused striatum scan; the attenuation corrected DaTscan images using the CNN estimated maps (CNN-AC) or the ground truth attenuation maps (GT-AC) achieve comparable accuracy for the striatal uptake value (mean deviation of 1.8%). Thus the CNN-based method can be an automatic and accurate tool for DaTscan attenuation correction that is independent of data from other imaging modalities or human interpretation.

Keywords: CNN, attenuation correction, pinhole SPECT, DaTscan, quantitative clinical SPECT
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 209

Poster Number:
M-08-209

Data-driven gating for PET using unsupervised deep clustering framework (#1393)

T. Li1, W. Qi2, E. Asma2, J. Qi1

1 University of California, Davis, Department of Biomedical Engineering, Davis, California, United States of America
2 Canon Medical Research USA, Inc., Vernon Hills, Illinois, United States of America

Content

Respiratory motion is one of the main sources of motion artifacts in positron emission tomography (PET) imaging. In this study, we proposed a data-driven approach using an unsupervised deep clustering network that employs an autoencoder (AE) to extract latent features for respiratory gating. The deep AE is first trained using reconstructed short-time frame images. To reduce noise and image reconstruction time, attenuation and scatter corrections are not performed in the short-time frame reconstructions. K-means clustering is then used to perform respiratory gating based on the latent features extracted by the deep AE. The effectiveness of our proposed method was evaluated using physical phantom data and real patient data. The result showed our proposed approach can provide higher lesion contrast than phase-based gating based on an external signal.

Acknowledgment

This work is supported in part by the National Institutes of Health under Grant R01EB000194.

Keywords: deep learning, clustering, PET, respiratory gating
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 212

Poster Number:
M-08-212

Air Fraction Correction Optimisation in PET Imaging of Lung Disease (#1447)

F. Leek1, 2, A. P. Robinson3, R. M. Moss4, F. J. Wilson5, B. F. Hutton1, K. Thielemans1

1 University College London, Institute of Nuclear Medicine, London, United Kingdom
2 University College London, EPSRC Centre for Doctoral Training in Intelligent, Integrated Imaging in Healthcare (i4health), London, United Kingdom
3 National Physical Laboratory, Teddington, United Kingdom
4 University College London, Department of Medical Physics and Biomedical Engineering, London, United Kingdom
5 GlaxoSmithKline Research & Development Limited, Stevenage, United Kingdom

Content

Accurate quantification of radiopharmaceutical uptake from lung PET/CT is challenging due to large variations in fractions of tissue, air, blood and water. Air fraction correction (AFC) uses voxel-wise air fractions which can be determined from the CT acquired for attenuation correction (AC). However, resolution effects can cause artefacts due to AC and/or AFC. In this work, we hypothesise that the resolution of the CT image used for AC should match that of the PET system but should approximate the reconstructed PET resolution for AFC. Simulations and reconstructions were performed with the Synergistic Image Reconstruction Framework (SIRF) using phantoms with inhomogeneous attenuation (mu) maps, mimicking the densities observed in lung pathologies. MLEM reconstruction was performed with a Gaussian smoothing applied to the mu-map that varied between 0 and 10mm full-width-half-maximum (FWHM). A Gaussian post-reconstruction filter of varying FWHM (0-10mm) was also applied to the PET image and independently to the mu-map used for AFC (FWHM 0-15mm). The RMSE between the simulated uptake density and the reconstructed images for each triple of blurring kernels was determined. The measured spatial resolution of the post-filtered PET was determined via a point-source insertion-and-subtraction method. Results show that the RMSE was minimised when the kernel applied to the mu-map for AC matched that simulated in the acquisition model and the kernel applied to the mu-map for AFC matched the spatial resolution of the reconstructed PET image. This supports the initial hypothesis. Further validation with Monte Carlo simulations is in progress. This work is a first step towards the construction of application-specific test-objects to optimise parameters for PET/CT lung imaging corrections.

Acknowledgment

F. L. is supported by EPSRC iCASE studentship (EP/T517628/1), NPL Management Limited, GlaxoSmithKline (BIDS3000035300) and UCL EPSRC Centre for Doctoral Training in i4health (EP/S021930/1).

Keywords: Phantoms, Positron emission tomography, Lung, Air fraction
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 215

Poster Number:
M-08-215

Improved Sensitivity at High Specificity for Signal Discrimination in CMOS Intraoperative Probes by Deep Learning (#1491)

J. Moo1, P. K. Marsden1, K. Vyas2, A. J. Reader1

1 King's College London, Biomedical Engineering and Imaging Sciences, London, United Kingdom
2 Lightpoint Medical Ltd, Chesham, United Kingdom

Content

The challenge in delineating the boundary between cancerous and healthy tissue during cancer resection surgeries can be addressed with the use of intraoperative probes to detect cancer cells labelled with radiotracers to facilitate excision. In this study, deep learning algorithms for background gamma ray signal rejection were explored for an intraoperative probe utilising CMOS monolithic active pixel sensors optimised towards the detection of internal conversion electrons from 99mTc. Two methods utilising convolutional neural networks (CNNs) were explored for beta-gamma discrimination: 1) classification of event clusters isolated from the sensor array outputs (SAOs) from the probe and 2) semantic segmentation of event clusters within an acquisition frame of an SAO. The feasibility of the methods in this study was explored with several radionuclides including 14C, 57Co and 99mTc. Overall, the classification deep network is able to achieve an improved area under the curve (AUC) of the receiver operating characteristic (ROC), giving 0.93 for 14C beta and 99mTc gamma clusters, compared to 0.88 for a more conventional feature-based discriminator. Further optimisation of the lower left region of the ROC by using a customised AUC loss function during training led to an improvement of 33% in sensitivity at low false positive rates compared to the conventional method. The segmentation deep network is able to achieve a mean dice score of 0.93. Through the direct comparison of all methods, the classification method was found to have a better performance in terms of the AUC.

Keywords: Intraoperative probes, Convolutional Neural Networks, Cancer surgeries, CMOS, Deep learning
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 218

Poster Number:
M-08-218

PET Image Denoising using a Deep-Learning Method for Extremely Obese Patients (#1563)

H. Liu1, H. Yousefi1, N. Mirian1, M. Lin1, 2, D. Menard1, M. Gregory1, M. Aboian1, A. Boustani1, M. - K. Chen1, M. Kulon1, L. Saperstein1, D. Pucar1, C. Liu1

1 Yale university, Department of Radiology and Biomedical Imaging, NEW HAVEN, Connecticut, United States of America
2 Visage Imaging, NEW HAVEN, Connecticut, United States of America

Content

In clinical PET scan, the image quality is severely degraded with high noise for the extremely obese (EO) subjects. Our work aimed to reduce the noise for clinical PET images of those EO subjects to the noise level of not EO subject images, to ensure the consistent imaging quality. Deep learning-based noise reduction method with a fully 3D patch-based U-Net was used. The training datasets, including down-sampled low-count images as the input and full dose images as the output, were derived from 100 not EO subjects. The down-sampling ratio of the input images was selected by comparing the normalized standard derivation (NSTD) inside 3D liver ROIs between various count levels of the not EO datasets and EO subjects. For U-Net training, to generate the input images with a consistent count level with the images of EO subjects, the optimized down-sampling ratio of the not EO subjects of 40% was used. The clinical PET images of 10 EO subjects were denoised using the above-trained network. For comparison, we also obtained the denoised images for these subjects using another network trained by the datasets with inconsistent count level of 10%. The results showed the U-Net trained by the images of the not EO subjects with 40% count level effectively reduced the noise in the images of the EO patients while preserving the fine structures without over-smoothing. The liver NSTD reduced from 0.13±0.04 to 0.08±0.03 after noise reduction (p = 0.01). After denoising, the image noise level of EO dataset was similar to that of not EO subjects, in terms of liver NSTD (0.08±0.03 vs. 0.08±0.02, p = 0.74). In contrast, the U-Net trained by the images with inconsistent count level over-smoothed the images of these EO patients with blurred fine structures. In conclusion, the U-Net trained by datasets from not EO subjects with matched count level can provide promising denoising performance for EO subjects while maintaining image resolution.

Keywords: noise reduction, extremely obese patient, FDG PET, deep learning
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 221

Poster Number:
M-08-221

Comparison of CNN-based approaches for detection of COVID-19 on chest X-ray images (#1690)

K. H. Leung1, 2, S. P. Rowe2, M. G. Pomper1, 2, Y. Du2

1 Johns Hopkins University School of Medicine, Department of Biomedical Engineering, Baltimore, Maryland, United States of America
2 Johns Hopkins University School of Medicine, The Russell H. Morgan Department of Radiology, Baltimore, Maryland, United States of America

Content

Coronavirus disease 2019 (COVID-19) is a highly contagious respiratory illness that has been declared a global pandemic by WHO. Chest X-ray imaging has been shown to serve an important role in early diagnosis and staging of the disease. The aim of this study was to develop a deep learning approach for detection of COVID-19 in chest X-ray images. Data were extracted from an opensource COVID-19 database developed by Cohen JP (https://github.com/ieee8023/covid-chestxray-dataset). The data consisted of X-ray images of patients with COVID-19, other pneumonias or no findings and were randomly partitioned into training, validation and test datasets containing 143, 32, and 30 images, respectively, using a 70%/15%/15% split. The performance of several commonly used deep convolutional neural network (CNN)-based architectures, including VGG16, ResNet50, DenseNet121, and InceptionV3, were evaluated on the disease detection task. These networks were first pretrained on ImageNet, a large dataset of natural images, and then further fine-tuned on the task of detecting COVID-19 in chest X-ray images. The networks were evaluated on the test set by assessing overall accuracy, area under receiver operating characteristic curve (AUROC), sensitivity and specificity. The performance of the networks trained from scratch without pretraining was also compared to the performance of the networks that were first pretrained on ImageNet and then fine-tuned on the detection task. Among the four CNNs tested, DenseNet121 had the highest performance on the test set with an overall accuracy of 90.0% (95% confidence interval (CI): 78.6%, 100%), an AUROC of 0.95, a sensitivity of 91.3% and a specificity of 85.7%. The pretrained DenseNet121 also significantly outperformed the DenseNet121 trained from scratch (paired sample t-test P-value<0.05) with a 30.0% improvement in overall accuracy. The proposed CNN-based approach showed significant promise for detection of COVID-19 in chest X-ray images.

Keywords: Coronavirus, COVID-19, Chest X-ray, Deep Learning, Disease Detection
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 224

Poster Number:
M-08-224

Rigid Motion Tracking using Moments of Inertia in TOF-PET Brain Studies (#1771)

A. Rezaei1, M. Spangler-Bickell2, G. Schramm1, K. Van Laere1, J. Nuyts1

1 KU Leuven, Department of Imaging and Pathology / Nuclear Medicine & Molecular imaging, Leuven, Belgium
2 IRCCS Ospedale San Raffaele, Nuclear Medicine Unit, Milan, Italy

Content

A data-driven rigid motion tracking approach for TOF-PET brain studies is proposed. The motion tracking approach extends previous works using moments of inertia for motion estimation. Since the raw emission measurements are used directly to estimated the motion trace, back-projections or reconstructions can be avoided. We find that patient motion traces can be estimated with good accuracies, which could possibly even be used for motion detection and correction of MR in hybrid PET/MR scanners.

AcknowledgmentThis work was supported by the Research Foundation Flanders (FWO) 12T7118N.
Keywords: rigid motion detection, TOF-PET
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 227

Poster Number:
M-08-227

Deep Learning PET Epilepsy Detection with a Novel Symmetric Loss Convolutional Autoencoder (#1827)

R. L. Smith1, E. Alsyed1, 2, L. Bartley1, H. Chandler1, P. Fielding1, C. Marshall1

1 Cardiff University, Wales Research and Diagnostic Positron Emission Tomography Imaging Centre (PETIC),School of Medicine, University Hospital of Wales, Cardiff, United Kingdom
2 king Abdulaziz University, Faculty of Engineering, Jedda, Saudi Arabia

Content

Positron Emission Tomography imaging (PET) with 18F-fluoro-2-deoxyglucose (FDG) is a well-established method to image brain glucose metabolism in healthy and disease states with application areas in neuro-degenerative and seizure disorders. Epilepsy is one of the most common neurological disorders effecting approximately 1-2% of the population. Advances in deep learning have revolutionized quantitative analysis and interpretation of medical images. Increasing availability of large medical imaging databases and sophisticated methods of extracting their discriminative features allows the potential for a greater understanding of the alterations of medical images in varying disease states. Probing medical images latent state presentations allows compressed representations with descriptive attributes to be explored. The auto-encoder and its variants serve this purpose; herein they naively attempt to learn the identity function to reconstruct the input. Further constraints on the autoencoder avoid over-fitting to the training data hence boosting the discovery process. An autoencoder trained on a normal healthy database has recently found application areas in anomaly detection of FDG uptake with the magnitude of the reconstruction error serving as a proxy for abnormal brain patterns. In this work we take advantage of the fact that FDG uptake in the healthy subject is usually homogeneous and symmetrical with left-right asymmetries in activity concentration present in patients with hypometabolic epileptogenic regions. We therefore construct a novel autoencoder for anomaly detection in FDG brain PET and utilize a regularizing symmetry term in the loss function. This has the effect of producing increased reconstruction error in the event of an anomaly. The autoencoder is trained on 120 normal volunteers and tested on 3 patients with diagnosed epilepsy, demonstrating an average increase in reconstruction error of 6%,  hence greater discriminative power in anomaly detection.

Keywords: Deep Learning, PET imaging, epilepsy detection
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 230

Poster Number:
M-08-230

A Self Organizing Map for Exploratory Analysis of PET Radiomic Features (#1889)

E. Alsyed1, R. L. Smith2, S. Paisey2, C. Marshall2, E. Spezi1

1 Cardiff University, School of Engineering, Cardiff, United Kingdom
2 Wales R&D PET Imaging Centre, Cardiff, United Kingdom

Content

Texture analysis for quantification of intratumor uptake heterogeneity in PET/CT images has received increasing attention.  This allows the extraction of a large number of ‘radiomic’ features to be correlated with end point information such as tumor type, therapy response, prognosis.  The conventional complex workflow for calculation of texture features introduces numerous confounding variables. This non exhaustively includes, imaging time post administration of radiopharmaceutical and the method and extent of functional volume segmentation.  A lack of understanding on the dependency of texture features with these variables serves as a detriment to the urgent need to standardize texture measurements to pool results from different imaging centers.  The utilization of machine learning techniques for feature (and their combinations) selection serves as a promising method to alleviate redundancy in radiomics.  To this avail, we introduce for the first time the application of a Kohonen self-organizing feature map to identify the emergent properties present when performing texture analysis.  The application of the self-organizing map to radiomic analysis serves as a powerful general-purpose exploratory instrument to reveal the statistical indicators of texture distributions.  For this purpose, texture features from PET-CT images of 8 pre-clinical mice with mammary carcinoma xenografts were analyzed with varying post injection imaging time and tumor segmentation contour size.  This varying distribution of texture parameters were interpreted by the self-organizing map to reveal two distinct clusters of texture features dependent on contour size providing additional evidence that contour size is a confounding variable when performing texture analysis. Furthermore, the self-organizing map can be utilized as a method to incorporate this revealed dependency in a prediction model in the presence of end point information, which will be an area of future work.

Acknowledgment

E. Alsyed is supported by king Abdulaziz University, Jeddah, Saudi Arabia (grant number # KAU1938).

Keywords: PET, Artificial Intelligence, Self-Organizing Map, Texture Analysis, Radiomics
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 233

Poster Number:
M-08-233

An Adaptive Multi-channel Feature-learning Model for Polyp Classification (#1959)

W. Cao1, M. J. Pomeroy1, Z. J. Liang1

1 State University of New York at Stony Brook, Department pf Radiology, Stony Brook, New York, United States of America

Content

Extracting effective texture features from computed tomographic colonography (CTC) and merging them to form a much powerful descriptor are two critical challenges in computer-aided detection (CADe) and diagnosis (CADx). In this paper, we introduce multiscale analysis into grey level co-occurrence matrix (GLCM) to construct texture features from different image domains, i.e. intensity, gradient and curvature. Thus, nine texture descriptors are generated and form a descriptor pool sorted by AUC (area under the curve of receiver operating characteristics) scores. Then an adaptive feature learning method is designed and implemented in a hierarchical framework where every layer consists of two nodes, i.e. the baseline descriptor and its complement which are always the top two descriptors in the descriptor pool. Their merging will be performed using forward stepwise method where some positive variables with gains in classification are preserved. After feature merging, the descriptor pool will be updated by removing the two candidates and adding the new baseline descriptor. This procedure will be performed iteratively till the final descriptor is obtained. Obviously, this is a greedy procedure which guarantees the monotonicity of the classification. Experimental outcomes testify the effectiveness of this method and the proposed method outperforms the pre-merging descriptor over 4% by AUC scores.

Acknowledgment

This work was partially supported by the NIH/NCI grant #CA206171.

Keywords: colorectal polyp, texture, machine learning, feature selection, classification.
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 236

Poster Number:
M-08-236

This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 239

Poster Number:
M-08-239

Respiratory motion compensation on TOF Listmode or sinograms (#2200)

T. Feng1, G. Yang2, H. Liu2, Y. Lv2, Y. Dong2

1 UIH America, Inc., Houston, Nevada, United States of America
2 Shanghai United Imaging Healthcare, Shanghai, China

Content

A dedicated reconstruction algorithm, usually 4D image reconstruction, is usually required for respiratory motion compensation. The use of a dedicated reconstruction method may make it harder for implementation and can be time-consuming especially with the use of 4D reconstruction. In this study, we present a motion compensation approach that directly applied to Listmode or sinogram data. Using the proposed approach, a regular reconstruction method with minimal or no modifications can be applied to achieve motion compensation. The first step of the proposed approach is to generate image-based motion vector fields (MVF). With the use of the motion signal, the MVF for every time point is generated. The MVF along the time-of-flight (TOF) line-of-responses (LOR) was calculated using rigid approximations, and the event was transformed with a new set of detector pairs and TOF coordinates. The same principle was used when applying sinogram-based transformations. The sensitivity map was generated using the same transformations and used in the reconstruction. For easier implementations, only the translational motion was applied to the TOF LOR at the moment, and the rotational motion was ignored at the current stage. Motion compensation using the conventional approach (Gated reconstructions with matched attenuation map, and averaged after transformation) was also used to compare the effectiveness of the proposed approach. Reconstructed images from the patient data show that the proposed approach was able to reduce the motion blurring effects. However, it is less effective compared with the conventional method, which is likely caused by the approximations used in the current implementation. We have shown that with the proposed compensation method, respiratory motion compensation can be directly applied to TOF sinograms before image reconstruction. A regular reconstruction method can be used directly to achieve motion compensation.

Keywords: Respiratory motion, TOF-PET
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 242

Poster Number:
M-08-242

Deep-learning based dose kernel for pre-treatment quality assurance using the machine log-file (#2272)

B. J. Min1, H. K. Lee1, S. Choi1, Y. - S. Seo1, W. - D. Kim1, W. - Y. Park1

1 Chungbuk National University Hospital, Department of Radiation Oncology, Cheongju, Republic of Korea

Content

The aim of this study is to predict the delivered dose distribution [Ddelivered(x, y)] with the use of a fluence-to-dose network (FDNet) to conduct patient-specific IMRT quality assurance (pQA). To acquire the information of the fluence map, a complete dynamic multileaf collimator (MLC) log file (Dynalog file) was used. The pQA that is based on the Dynalog file has advantages compared to the uncertainty of various dosimeters and human errors that may occur in the positioning of the dosimeters. To date, the previous method that used the Dynalog file is compiled according to the direct comparison of the actual and the planned fluence maps, or is based on the recalculation of the dose distribution by the treatment planning system (TPS) using partial information obtained from the Dynalog file and planned dose distribution [Dplanned(x, y)]. We propose a new method to perform pQA by predicting the Ddelivered(x, y) in a water phantom using the actual machine parameters stored in the Dynalog file for all the timesteps based on the FDNet. The FDNet was based on the training of a convolution neural network (CNN). Clinical IMRT plans from 37 cases with four disease sites were used for FDNet training. The average gamma passing rates based on the 3%/3 mm gamma criterion were respectively equal to 98.56%, 97.11%, 97.23%, and 98.03%, for the proposed method (which used all the available information), proposed method (which used only partial information), EBT3 film, and MatriXX. According to this study, the feasibility of the pQA method with the use of the FDNet with complete Dynalog information was verified. The respective differences of the average gamma passing rates for the proposed method using complete and partial Dynalog information were 1,46% and 3.19% according to the 3%/3 mm, and 2%/2 mm gamma criteria.

Acknowledgment

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (No. NRF-2020R1A2C4001910).

Keywords: radiotherapy, treatment planning, quality assurance
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 245

Poster Number:
M-08-245

Multi-Level PET and CT Image Fused Radiomics Feature-based Survival Analysis of NSCLC Patients (#2332)

M. Amini1, M. Nazari1, I. Shiri2, G. Hajianfar3, M. R. Deevband1, H. Abdollahi4, H. Zaidi5

1 Shahid Beheshti University of Medical Sciences, Department of Biomedical Engineering and Medical Physics, Tehran, Iran (Islamic Republic of)
2 Geneva University Hospital, Division of Nuclear Medicine and Molecular Imaging, Geneva, Genève, Switzerland
3 Iran University of Medical Science, Rajaie Cardiovascular Medical and Research Center, Tehran, Iran (Islamic Republic of)
4 Kerman University of Medical Science, Department of Radiologic Sciences and Medical Physics, Kerman, Iran (Islamic Republic of)
5 Geneva University, Geneva University Neurocenter, Geneva, Genève, Switzerland

Content

To provide a comprehensive characterization of intra-tumor heterogeneity, this study propose multi-level multi-modality radiomic models derived from 18F-FDG PET and CT images by feature- and image-level fusion. Specifically, we developed fusion radiomic models to improve overall survival prediction of NSCLC patients. In this work, 168 NSCLC patients from two different institutions (86 for training and 82 for external validation) were included. TNM staging and histopathological grade was utilized to build a clinical model. By extracting 225 features from CT and PET images, radiomics analysis was used to build single-modality and multi-modality models where the fused images are constructed by 3D-wavelet transform fusion (WF) with 3 different CT weights (W) and 3 different band-pass filtering ratios (R). Two models were also developed using two feature-level strategies of feature concatenation (ConFea) and feature averaging (AvgFea). Among image-level fusion models, WF-W0.5-R1.5 (C-index=0.676) significantly outperformed CT and PET models (C-index=0.548, 0.619, respectively). Our results demonstrate that multi-modal radiomic models have the potential to improve prognosis by combining information from different tumor characteristics, including anatomical and metabolic captured by different imaging modalities.

AcknowledgmentThis work was supported by the Swiss National Science Foundation under grant SNSF 320030_176052 and Shahid Beheshti University of Medical Sciences.
Keywords: Radiomics, PET/CT, Fusion, Survival
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 248

Poster Number:
M-08-248

MRI Delta-radiomic features integrated with clinical parametersfor prediction of treatment response in colorectal cancer (#2337)

S. P. Shayesteh1, A. Salahshour2, M. Nazari3, G. Hajianfar4, S. Sandoughdaran5, A. Yaghobi Joybari5, F. Jozian5, S. H. Fatehi Feyzabad4, I. Shiri6, H. Zaidi6, 7

1 Alborz University of Medical Sciences, Department of Physiology, Pharmacology and medical physics, Tehran, Iran (Islamic Republic of)
2 Alborz University of Medical Sciences, Department of Radiology, Tehran, Iran (Islamic Republic of)
3 Shahid Beheshti University of Medical Sciences, Department of Biomedical Engineering and Medical, Tehran, Iran (Islamic Republic of)
4 Iran University of Medical Science, Rajaie Cardiovascular Medical and Research Center, Tehran, Iran (Islamic Republic of)
5 Shahid Beheshti University of Medical Sciences, Department of Radiation Oncology, Tehran, Iran (Islamic Republic of)
6 Geneva University Hospital, Division of Nuclear Medicine and Molecular Imaging, Geneva, Genève, Switzerland
7 Geneva University, Geneva University Neurocenter, Geneva, Genève, Switzerland

Content

Colorectal cancer (CRC) is the third and fourth most common cancer in the US and worldwide, respectively. Different responses to neoadjuvant chemo-radiation therapy (nCRT) in patients with locally advanced rectal cancer (LARC) led to failure of the current therapies. The prediction of response to radiation therapy is an important prognostic factor, however radiomic features, before and after treatment cannot describe all characteristics. Hence, radiomic features variation in different medical images throughout the treatment, called delta radiomics, has been examined as a predictive tool for prediction of response to treatment. The aim of this study was to determine the ability of pre/post-treatment and delta-radiomic features in response prediction in LARC patients. Different radiomics feature including shape, first order, GLDM, GLRLM, GLSZM and NGTDM were extracted from pre- and post-treatment MRI. The delta-radiomic features were defined as the relative net change of the feature after treatment for each patient. XGboost classifier and LASSO feature selector were used to find predictive performance of each feature based on the area under receiver operating characteristic (ROC) curves. The best results for response prediction in LARC patients were obtained using delta-radiomic features with AUC of 0.92 and 0.88 in training and validation dataset, respectively. Our results demonstrated that delta-radiomic features have the highest predictive performance in LARC patients.

Acknowledgment

This work was supported by the Alborz University of Medical Sciences and the Swiss National Science Foundation under grant SNSF 320030_176052.

Keywords: Radiomics, Machine Learning, MRI, colorectal cancer, Delta-Radiomics
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 251

Poster Number:
M-08-251

Joint Dose Minimization and Variance Optimization for Fluence-Modulated Proton CT (#1112)

J. Dickmann1, F. Kamp2, 3, R. W. Schulte4, K. Parodi1, G. Dedes1, G. Landry2, 3

1 Ludwig-Maximilians-Universität München, Department of Medical Physics, Fakultät für Physik, Garching bei München, Germany
2 University Hospital, LMU Munich, Department of Radiation Oncology, Munich, Germany
3 German Cancer Consortium (DKTK), Munich, Germany
4 Loma Linda University, Division of Biomendical Engineering Sciences, Loma Linda, California, United States of America

Content

We present a joint dose minimization and variance optimization algorithm for fluence-modulated proton computed tomography that allows prescribing spatially inhomogeneous dose and image noise distributions. This is particularly meaningful if proton CT images are used for particle therapy treatment planning and online adaptation, where only the region-of-interest (ROI) encompassing the treatment beam path is relevant and imaging dose can be reduced elsewhere. This may allow for daily imaging at the treatment site with imaging doses that would not compromise the low dose to healthy tissue made possible by particle therapy. The method makes use of concepts of treatment planning, allowing for a flexible prescription of imaging dose and variance with corresponding penalties and constraints. We investigate a typical treatment scenario with two beams and optimize dynamic fluence maps using the proposed method. Resulting imaging doses and variances are calculated in a Monte Carlo simulation and result in a dose reduction of 30% outside of the ROI at equal variance inside the ROI. Increasing the magnitude of dose reduction inside a small volume around an organ-at-risk (OAR) brings the OAR dose down by 62% of that of a scan without fluence modulation, thus more than halving local imaging dose. This flexible optimization method may facilitate low-dose image guidance and online plan adaptation using proton CT.

AcknowledgmentDFG project #388731804.
Keywords: proton CT, fluence modulation, dose reduction, particle therapy, image noise
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 254

Poster Number:
M-08-254

Towards a clinical application of carbon-ion beam monitoring along depth using detection and tracking of prompt secondary charged particles (#1167)

L. Ghesquière-Diérickx1, 2, R. Félix-Bautista1, 4, T. Gehrke1, 3, J. Jakubek5, M. Winter6, 3, M. Ellerbrock6, 3, M. Martišíková1, 3

1 German Cancer Research Center (DKFZ), Department of Medical Physics in Radiation Oncology, Heidelberg, Germany
2 University of Heidelberg, Heidelberg Medical Faculty, Heidelberg, Germany
3 National Center for Research in Radiation Oncology (NCRO), Heidelberg Institute for Radiation Oncology (HIRO), Jeidelberg, Germany
4 University of Heidelberg, Department of Physics and Astronomy, Heidelberg, Germany
5 ADVACAM s.r.o, Prague, Czech Republic
6 Heidelberg Ion-Beam Therapy Center (HIT), Department of Radiation Oncology, Heidelberg, Germany

Content

The benefit of carbon-ion radiotherapy comes at the expense of a higher sensitivity to uncertainties compared to other standard photon radiotherapy. In-vivo monitoring methods of carbon-ion radiotherapy are, thus, of great importance to visualize the beam stopping position in depth and to monitor any interfractional changes in the patient. In this contribution, we focus on a monitoring method based on the detection and tracking of secondary charged nuclear fragments emerging from an irradiated target. More precisely, a correlation between the measured depth of the fragments’ origin and the planned primary carbon-ion's expected stopping position in the target is investigated.
To evaluate this monitoring method, irradiations were performed at the Ion-Beam Therapy Centre (HIT) Heidelberg, Germany, where a typical carbon-ion treatment fraction of 3 GyRBE of a head tumor was irradiated on an anthropomorphic head model. Emerging fragments were measured using 2 silicon TX3 detectors, placed in parallel, 12cm behind the head center, 30° with respect to the beam axis. Reconstructed fragment’s tracks were extrapolated in space back into the targeted head. Allowing to estimate the carbon-ion fragmentation location and, thus, to approximate the stopping positions of the beams in the head.
With our method, we found significant differences in the fragments’ distribution in depth for each carbon-ion beam energy. Interfractional changes created by placing a 1mm-thick PMMA slab placed upstream from the head model between irradiations were detected. Additionally, the method was also successfully applied in a first patient measurement at HIT, which validates the technical feasibility of the method’s first clinical application.

AcknowledgmentThe authors would like to thank the HIT facility for the beam-time for this work. We also thank the ADVACAM s.r.o. for lending us some AdvaPIX TPX3 modules and for their technical support.
Keywords: particle tracking, prompt secondary ions, carbon-ion beam radiotherapy, monitoring
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 257

Poster Number:
M-08-257

Proton Radiography for Water Equivalence (#1275)

M. S. Freeman1, J. C. Allison1, E. F. Aulwes1, B. Broder1, M. Espy1, P. Magnelind1, F. G. Mariam1, J. J. Medina1, W. Z. Meijer1, F. E. Merrill1, L. P. Neukirch1, T. Schurman1, R. B. Sidebottom1, A. Tainter1, Z. Tang1, F. R. Trouw1, D. Tupa1, J. L. Tybo1

1 Los Alamos National Laboratory, P-23: Neutron Science & Technology, Los Alamos, New Mexico, United States of America

Content

An accurate, 3D map of proton water equivalence path length is needed for proton therapy treatment planning. Typically, this is done using X-rays, with an accepted error of ±2%, in the conversion of electron density to proton stopping power. If instead, transmission radiography is acquired using protons, no assumptions need to be made in the estimation of proton stopping power, potentially leading to higher degrees of accuracy in the estimations of water equivalence. Here, the 800-MeV proton radiography system at the Los Alamos Neutron Science Center is used to provide transmission radiography of a 20-cm thick acrylic block with circular features of depth 0.25 - 5.00 mm, and tomography of a pediatric head phantom. The system was operated in two configurations, one optimized for spatial resolution, and the other for minimal dose deposition and increased sensitivity. In high-resolution mode, a sensitivity to a 1% change in water equivalence path length was observed, with a 300 µm spatial resolution. In low-dose mode, the spatial resolution was degraded due to chromatic blurring, while the radiography was able to detect changes in water equivalence path length of just 0.25%. This system could be deployed in a fixed beam configuration at a proton therapy center in order to provide real-time, beam's-eye-view estimates of water-equivalence path length. Such measurements, incorporated into the clinic, could lead to image guidance or adaptive proton therapy.

Acknowledgment

This work was supported by the US Department of Energy through the Los Alamos National Laboratory. Los Alamos National Laboratory is operated by Triad National Security, LLC, for the National Nuclear Security Administration of U.S. Department of Energy (Contract No. 89233218CNA000001). Research presented in this article was supported by the Laboratory Directed Research and Development program of Los Alamos National Laboratory under project number 20180238ER.

Keywords: proton radiography, water equivalence, therapy guidance
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 260

Poster Number:
M-08-260

Proton Radiography for a Small-Animal Irradiation Platform based on a Miniaturized Timepix Detector (#1420)

M. Würl1, K. Schnürle1, J. Bortfeldt1, C. Oancea2, C. Granja2, E. Verroi3, F. Tommasino3, 4, K. Parodi1

1 Ludwig-Maximilians-Universität München, Department of Medical Physics, Faculty of Physics, Garching, Bavaria, Germany
2 ADVACAM, Prague, Czech Republic
3 National Institute for Nuclear Physics (INFN), Trento Institute for Fundamental Physics and Applications (TIFPA), Povo, Italy
4 University of Trento, Department of Physics, Povo, Italy

Content

Pre-treatment proton radiography and computed tomography can improve precision of small animal proton irradiation. A compact imaging setup for small-animal proton radiography, based on a miniaturized Timepix detector is presented along with results from proof-of-concept experiments. The MiniPIX detector was placed behind a µ-CT calibration phantom with 10 inserts of tissue-mimicking material. The intensity of the 70 MeV proton beam was adjusted such that clusters of individual protons on the detector could be resolved. Cuts on various cluster properties were used to suppress bad events. The energy deposition of the remaining clusters was converted to water-equivalent thickness (WET) of the traversed material using a conversion curve based on Monte Carlo simulations and measured clusters of protons after traversing PMMA slabs of known thickness. Despite a systematic underestimation of up to 3%, retrieved WET values are in good agreement with ground truth. Spatial resolution is ranging from 0.3 to 0.7 mm for phantom-detector-distances of 1 to 5 cm. Applicability to living samples is currently limited by the relatively long acquisition time of up to 20 minutes. This obstacle can however be overcome with the latest detector model, allowing to handle higher particle rates and thus requiring shorter irradiation times.

AcknowledgmentThe authors acknowledge financial support from the European Research Council (grant agreement number 725539).
Keywords: proton radiography, proton radiotherapy, small animal irradiation, Timepix
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 262

Poster Number:
M-08-262

Dictionary based MLEM-algorithm for real-time proton range verification from PET data: the virtue of contrasts (#1436)

V. Valladolid-Onecha1, P. Galve1, A. Espinosa1, S. M. España1, 2, D. Sánchez Parcerisa1, 3, P. Ibáñez1, L. M. Fraile1, 2, J. M. Udías1, 2

1 Complutense University of Madrid, Faculty of Physical science, Nuclear Physics Group and IPARCOS, EMFTEL, CEI Moncloa, Madrid, Spain
2 Health Research Institute of the Hospital Clínico San Carlos (IdSSC), Madrid, Spain
3 Sedecal Molecular Imaging, Madrid, Spain

Content

Proton therapy holds potential for unsurpassed
conformality of dose. But this potential will not be fully realized
until a complete control of the uncertainties involved in the
determination of the Bragg peak position is mastered. In-vivo
dose verification techniques are being pursued to tackle this
problem. Nuclear activation produced by protons in the patient
can be tracked to obtain dose maps in the irradiated body. One of
these techniques uses detected PET activity to reconstruct the
absorbed dose. In this work we propose a fast and accurate dose
reconstruction algorithm based in Monte Carlo (MC) simulation
from activation maps. Prior to irradiation, we precalculate the
dose deposition and activity distribution produced on the patient
by each individual pencil beam (PB) in the irradiation plan.
During or right after the irradiation, with an MLEM algorithm,
we calculate the linear combination of the pre-calculated PB
activities that best fits the observed PET activation data. This
method can incorporate the complete physics in the
precalculation phase, employing for instance MC packages for
protontherapy and PET, making it possible dose reconstruction
in real time since the matching linear combination of PBs is
obtained within seconds in a common GPU. As an example, we
apply our method to estimate the effect of activity-enhancing
contrast agent 18O in a simulated phantom irradiated with a
proton beam.

AcknowledgmentWork supported by the Spanish Government (FPA2015-65035-P, RTC-2015-3772-1),
Comunidad de Madrid (B2017/BMD-3888 PRONTO-CM),
European Regional Funds and the European Union’s Horizon 2020 research and
innovation programme under the Marie Sklodowska-Curie grant agreement No 793576
(CAPPERAM). This is a contribution for the Moncloa Campus of International
Excellence, “Grupo de Física Nuclear-UCM”, Ref. 910059. Part of the calculations of
this work were performed in the “Clúster de Cálculo para Técnicas Físicas”, funded in
part by UCM and in part by EU Regional Funds.
Keywords: Proton Therapy, Range Verification, PET, Dose Reconstruction, MLEM algorithm.
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 266

Poster Number:
M-08-266

Assembly and initial characterization of MACACO III Compton telescope (#1549)

L. Barrientos1, J. Roser1, A. Esteban1, M. Borja-Lloret1, J. V. Casaña1, C. Lacasta1, E. Muñoz1, A. Ros1, C. Senra1, C. Solaz1, R. Viegas1, G. Llosá1

1 IFIC (CSIC-U. Valencia), Paterna (Valencia), Spain

Content

The IRIS group at IFIC-Valencia has developed a third version of the MACACO (Medical Applications CompAct COmpton camera) prototype for hadron therapy treatment monitoring, with the aim of improving the performance with respect to previous versions. The development includes the use of new LaBr3 crystals and photodetectors, improved readout electronics and more accurate image reconstruction codes. Silicon photomultipliers (SiPMs) with different micro pixel size (25 and 50 µm) have been considered and tested. The detector linearity has been improved and an energy resolution of the 5.3 % FWHM at 511 keV has been achieved. Three detectors have been arranged in Compton camera coincidence mode and images of a 22Na point-like source have been reconstructed selecting two and three-interaction events. The full width half maximum (FWHM) of the reconstructed images was 2.7 mm at 1275 keV, better than the one obtained with previous versions of the prototype in the same configuration. The experimental data have been reproduced with Monte Carlo simulations using a Compton camera module (CCMod) in GATE v8.2. Results of the selected hardware, detector performance and system characterization are presented.

AcknowledgmentThis work has received funding from the Spanish Ministerio de Ciencia e Innovación (FPA2017-85611-R and IFIC's Center of Excellence Severo Ochoa SEV-2014-0398) and from European Commission H2020 ENSAR2-MediNet (project number 654002). Group members are supported by Ramón y Cajal, UVEG Atracció de Talent and Generalitat Valenciana SEJI and predoctoral contracts.
Keywords: Compton camera, hadron therapy, LaBr3, silicon photomultipliers
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 269

Poster Number:
M-08-269

A new strategy for measurement of dose profiles and LET in therapeutic proton beam (#1710)

A. Ruciński1, J. Baran1, C. Granja2, C. Oancea2, M. Pawlik-Niedźwiecka1, M. Rydygier1, A. Schiavi3, P. Stasica4, J. Gajewski1

1 Institute of Nuclear Physics PAN, Proton Radiotherapy Group, Krakow, Poland
2 Advacam, Prague, Czech Republic
3 Sapienza University of Rome, Rome, Italy
4 AGH University of Science and Technology, Krakow, Poland

Content

The commissioning and validation of proton therapy treatment planning system (TPS) requires accurate measurements of lateral and longitudinal dose profiles in air and water. These measurements are time consuming and often inaccurate enough for high precision dose modeling in patients. Moreover, the increasing interest in clinical application of models accounting for variable radiobiological effectiveness (RBE) of proton beams demands experimental methods for single-particle sensitive validation of linear energy transfer (LET), the key physical component of phenomenological RBE models. We propose a new strategy for experimental characterization and validation of dose profiles and LET spectra in clinical proton beams.
We measured LET of single particles depositing energy in 0.3 μm thick pixelated silicon sensor (active area of 14x14 mm, 256x256 pixels) of MiniPix TimePix detector. The detector was placed in a waterproof PMMA holder and was used for calibration and measurements of lateral dose profiles in air as well as measurements of lateral dose profiles and LET spectra in water phantom, both performed with proton pencil beams at 100, 150, and 200 MeV. We compared the experimental results to Monte Carlo simulations of the experimental setup performed in GATE/Geant4 simulation toolkit and TPS calculations and data acquired during the facility commissioning.
We performed more than 300 measurements obtaining very high precision lateral dose profile determination in air and water. The comparison of calibration measurements to simulations, which show a very good agreement, indicate that the results of the measurements of LET spectra performed with TimePix in water are precise and allow for validation of the LET component of variable RBE models available in research versions of Monte Carlo based TPS. The moderate difference between simulated and measured LET spectra and their impact on the RBE modeling in proton therapy, will be discussed.

Keywords: Proton therapy, LET, Treatment planning, Monte Carlo simulations, Radiation therapy
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 272

Poster Number:
M-08-272

Simulation Study of Dose Estimation via Compton-based Prompt Gamma imaging during Proton Therapy: A Deconvolution Approach based on Evolutionary Algorithm (#1865)

J. Zhao1, 2, Z. Yao1, 2, Y. Xiao1, 2

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

Content

Dose monitoring and range verification are the key concerns connected with the therapeutic effect in proton therapy. Different methods allowing for a verification of the proton range have been developed, while the quantification of deposited dose remains challenging due to the uncertainty of treatment. In recent years, more attention has been paid to dose verification during the treatment, and predicting the dose in proton therapy with information from prompt gamma (PG) emission is considered highly feasible.
In this work, we present a deconvolution approach to predict dose distributions from prompt gamma images obtained by Compton camera. Based on the assumption that the one-dimensional PG distribution can be derived from the convolution of the depth dose distribution and a filter kernel, an approximation function of Bragg curve and evolutionary algorithm (EA) are introduced to achieve the inverse process and estimate the depth dose distribution from a measured PG distribution. Filter kernels used in the approach are obtained by the same evolutionary algorithm with predictions of the PG and depth dose distributions and arranged as a library.
The applicability of this approach is demonstrated for the monoenergetic proton irradiation of homogeneous phantoms made of different tissue materials, using Monte-Carlo simulated data. The distal falloff of depth dose distribution is reconstructed within less than 1 mm deviation, showing its potential for real-time dose estimation in future.

Acknowledgment

This work was supported by Beijing Municipal Nature Science Foundation (No. 7191005).

Keywords: Dose Estimation, Prompt Gamma imaging, Deconvolution Approach, Evolutionary Algorithm, Proton Therapy
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 275

Poster Number:
M-08-275

An advanced simulation and reconstruction framework for a novel in-beam PET scanner for pre-clinical proton irradiation (#1948)

G. Lovatti1, M. Nitta1, M. Safari1, C. Gianoli1, M. Pinto1, A. Zoglauer2, H. G. Kang3, T. Yamaya3, P. Thirolf1, G. Dedes1, K. Parodi1

1 Ludwig-Maximilians-Universität München, Department of Medical Physics, München, Germany
2 University of California Berkeley, Space Sciences Laboratory,, Berkeley, United States of America
3 National Institute of Radiological Sciences, Chiba, Japan

Content

Within the project "Small animal proton Irradiator for Research in Molecular Image-guided radiation-Oncology" (SIRMIO) we have designed an in-beam PET scanner for pre-clinical application. The system is based on a novel spherical geometry, and in order to fully exploit its potential we are developing an integrated computational framework for simulation, image reconstruction and range verification. The software comprises a full Monte Carlo engine to simulate the proton treatment with related detector response, and an image reconstruction tool for simulated and experimental data. The platform is designed to integrate robust analytical reconstruction algorithms and new statistical approaches based on deep learning.
The core of the framework is based on MEGAlib (The Medium Energy Gamma-ray Astronomy software library). The physical simulation is based on GEANT4. The machine learning method for the event classification is implemented with the  ROOT based Toolkit for Multivariate Data Analysis (TMVA). The first prototype of the SIRMIO irradiation platform foresees a fixed beam, thus requiring the movement of the mouse for scanned beam delivery. Hence, we have extended the MEGAlib image reconstruction algorithm based on maximum-likelihood expectation-maximization (ML-EM) to correct for geometrical efficiency and attenuation taking into account the mouse motion. The goal is to be able to discriminate proton range shifts of $\sim 0.5$ mm. Moreover, we are augmenting the image reconstruction framework with a new approach based on machine learning, which aims at using all photon events collected during irradiation (dominated by prompt gamma) to retrieve on-the-fly the range of the beam, to complement the PET information.

AcknowledgmentThis work is funded by the European Research Council (ERC) under the EU's Horizon 2020 (725539), and BaCaTec (Bavaria California Technology Center
Keywords: in-beam PET scanner, deep learning, proton beam image, Image-guided radiation-Oncology, small animal
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 278

Poster Number:
M-08-278

This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 281

Poster Number:
M-08-281

Deep Learning-based Automated Delineation of Head andNeck Malignant Lesions from PET images (#2292)

H. Arabi1, I. Shiri Lord1, E. Janebi2, M. Becker3, H. Zaidi1, 4

1 Geneva University Hospital, Division of Nuclear Medicine and Molecular Imaging, Geneva, Switzerland
2 Tehran University of Medical Sciences, Research Center fo nuclear medicine, Tehran, Iran (Islamic Republic of)
3 Geneva University, Division of Radiology, Geneva, Switzerland
4 Geneva University, Geneva Neuroscience Center, Geneva, Switzerland

Content

The accurate delineation of the gross tumor volume (GTV) is critical for treatment planning in radiation oncology. This task is very challenging due to the irregular and diverse shapes of malignant lesions. Manual delineation of the GTVs on PET images is not only time-consuming but also suffers from inter- and intra-observer variability. In this work, we develop deep learning-based approaches for automated GTV delineation on PET images of the head and neck region. To this end, V-Net, a fully convolutional neural network for volumetric image segmentation, and HighResNet, a 20-layer residual convolutional neural network, algorithms were adopted. 18F-FDG-PET/CT images of 510 patients presenting with head and neck cancer on which manually defined (reference) GTVs were utilized for training, evaluation and testing of these algorithms. The input of these networks (in both training or evaluation phases) were the 12 cm ×12 cm ×12 cm sub-volumes of PET images containing the whole volume of the tumors. These networks were trained to generate a binary mask representing the GTV on the input PET sub-volume. Standard segmentation metrics, including Dice similarity and precision were used for performance assessment of these algorithms. HighResNet achieved automated GTV delineation with a Dice index of 0.87±0.04 compared to 0.86±0.06 achieved by the V-Net. Despite the close performance of these two approaches, HighResNet exhibited less variability among different subjects reflected in the smaller standard deviation and significantly higher precision index (0.87±0.07 vs 0.80±0.10). Deep learning techniques, in particular HighResNet algorithm, exhibited promising performance for automated GTV delineation on head and neck PET images. Incorporation of the anatomical information, particularly MRI, may result in higher segmentation accuracy or less variability among the different subjects.

AcknowledgmentThis work was supported by the Swiss National Science foundation under grant SNFN 320030_176052.
Keywords: PET, Segmentation, deep learning, head and neck, PET/CT
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 284

Poster Number:
M-08-284

PET-18F-FDG pharmacokinetic modeling without blood sampling in arteries with atherosclerosis (#1361)

M. S. Al-enezi1, M. Bentourkia1

1 Université de Sherbrooke, Nuclear Medicine and Radiobiology, Sherbrooke, Québec, Canada

Content

Pharmacokinetic modeling of 18F-FDG in the arteries is necessary in comparison to the standard uptake value (SUV) and tissue-to-blood ratio (TBR) for two main reasons: the small thickness of the artery wall (around 2.3 mm) and the domination by activity emanating from blood. Consequently, the artery image is subject to partial volume effect (PVE) and it is not appropriate to use an input function IF determined from an artery. In the present work, we used a modified pharmacokinetic model derived from the classical two-tissue compartment model of 18F-FDG. In the classical model, the artery time-activity curve (TAC) is fitted with the model convolved with the IF to calculate the rate constants together with tissue blood volume (TBV) which is a fraction of IF (TBV=k5xIF, k5<1). In the modified model, the images are decomposed with factor analysis (FA) to provide blood image. The same region of interest of the artery is used in the blood image to determine TBV. The model is therefore rewritten to convolve the model with . Such model, called without blood sampling (WOBS) has several advantages mainly, since the same region of interest (ROI) is used for both TBV and the artery, PVE cancels out. We applied these two models and SUV to artery images in patients with atherosclerosis, with and without medication. For the classical model, artery TAC and IF, determined on the sagittal view of the aorta, were corrected for PVE. The metabolic rate of glucose (MRG) was calculated in 710 artery image slice with both models. We also compared MRG as a function of the ratio of CT calcification area (RCA) and Agatston calcification scores (ACS). TAC fits with WOBS were more accurate in almost all artery TACs and MRG had globally less variation with WOBS. MRG values in non-medication group were statistically significantly different in both models. WOBS and the classical model also detected the same differences between RCA and ACS.

Keywords: Pharmacokinetic modeling, Input function, PET imaging, Partial volume effect, 18F-FDG
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 287

Poster Number:
M-08-287

A single dual-tracer PET imaging acquisition to provide information on tumor heterogeneities (#1937)

B. Le Crom1, A. Bousse2, M. Chérel1, N. Costes3, S. Gouard1, S. Marionneau-Lambot1, T. Merlin2, D. Visvikis2, S. Stute1, 4, T. Carlier1, 4

1 INSERM, CRCINA, Nantes, France
2 INSERM, LaTIM, Brest, France
3 INSERM, CERMEP, Lyon, France
4 University Hospital, Nuclear Medicine, Nantes, France

Content

In PET imaging, using different tracers may provide complementary information on tumor heterogeneities. The use of a single PET acquisition with dual-tracer injection prevents bias between sequential acquisitions. As already shown in the literature, it is possible to separate two PET signals from equal half-life isotope tracers based on their pharmacokinetics. With the goal of building a generic framework for reconstructing separated images of each tracer from dual-tracer PET acquisitions, we use the spectral model without any assumptions about kinetics. Using 1D simulations of FLT+FDG, we evaluated the ability of the model to separate and extract parameters of interest for each tracer with respect to the delay between injections. Hopefully, results from preclinical acquisitions on mice will be presented.

AcknowledgmentThis study has been supported in part by the French National Agency for Research called "Investissements d’Avenir" IRON Labex ANR-11-LABX-0018-01, INCa-DGOS-Inserm-12558 (SIRIC ILIAD), Cancéropôle Grand Ouest and France Life Imaging network ANR-11-INBS-0006.
Keywords: PET Imaging, Dual-tracer, Tracer kinetic modeling, Spectral model, kinetic parameters
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 290

Poster Number:
M-08-290

Feasibility of Population-Based Input Function for Kinetic Analysis of [11C]-DPA-713 (#2113)

M. I. Akerele1, S. A. Zein1, S. Pandya1, A. Nikolopoulou1, S. A. Gauthier1, C. Henchcliffe1, P. D. Mozley1, R. Ashish1, N. A. Karakatsanis1, A. Gupta1, J. W. Babich1, 2, S. A. Nehmeh1

1 Department of Radiology, Weill Cornell Medicine, New York, New York, United States of America
2 Citigroup Biomedical Imaging Center, Weill Cornell Medicine, New York, New York, United States of America

Content

Quantitative PET studies of neurodegenerative diseases typically require the measurement of arterial input function (AIF), an invasive and risky procedure. The aim of this study was to assess the accuracy of population-based input function (PBIF) for [11C]DPA-713 PET kinetic analysis. The final goal is to possibly eliminate the need for AIF. Eighteen subjects from two [11C]-DPA-713 PET protocols in Multiple Sclerosis and Parkinson Disease were included in this study. Each subject underwent 90min dynamic PET imaging on a Siemens Biograph mCTTM scanner. Five healthy subjects underwent a Test/Retest within the same day to assess the reproducibility of the kinetic parameters.
Kinetic modeling was carried out using a 2-tissue-4-compartment model with PBIF, and again with the patient-specific AIF (gold standard). Using the leave-one-out cross validation method, we generated a PBIF for each subject from the remaining 17 subjects after normalizing the AIFs by three techniques: (a) patient weight×injected dose (b) Area Under AIF Curve (AUC), and (c) weight×AUC. The variability in the total distribution volume (VT) and non-displaceable binding potential (BPND) due to the use of PBIF was assessed for selected brain regions using Bland-Altman analysis, and for each of the three normalization approaches.
For the test-retest study, the thalamus exhibited the largest variability (up to 55%), as measured by the Bland-Altman limit of agreement (LOA), among the brain structures. Also, % relative difference between PBIF and AIF is significantly different across the normalization techniques, with the normalization by weight×AUC yielding the least % relative difference. For the Bland-Altman analysis, the mean % difference for VT lies within ±2% and the 95% LOA lies within ±40%. For the BPND, the mean difference lies within ±4%. and the corresponding 95% LOA is ±80%. This work is still in progress, and the feasibility of PBIF-based kinetic modelling will depend on the complete results.

Keywords: PBIF, Multiple Sclerosis, Parkinson’s Disease, [11C]DPA-713
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 293

Poster Number:
M-08-293

Improved image quality with optimal energy window in a compact brain-dedicated TOF-PET (#1238)

G. Akamatsu1, E. Yoshida1, H. Tashima1, Y. Iwao1, H. Wakizaka1, T. Maeda2, M. Takahashi1, T. Yamashita3, T. Yamaya1

1 National Institute of Radiological Sciences (NIRS-QST), Chiba, Japan
2 QST hospital, Chiba, Japan
3 ATOX Co. Ltd., Tokyo, Japan

Content

To make a high-sensitivity brain PET system with a limited number of detectors, we have proposed and developed a helmet-type TOF-PET prototype for which the coincidence timing resolution (CTR) was 245 ps. The energy resolution of 12.6% was obtained with the energy window of 420-590 keV. For such a compact system, the scattered coincidence counts increase due to the increased solid angle coverage. Our aim in this study was to investigate an optimal energy window of this prototype in order to obtain better count rate performance and better image quality. CTR, noise-equivalent count rate with TOF gain (TOF-NECR), and image quality were evaluated with 16 different patterns of the energy windows (range, 400-700 keV). A brain-sized image quality (IQ) phantom and a hemispherical Hoffman phantom were used for image quality evaluation. The highest TOF-NECR was obtained with the energy window of 440-590 keV, and the CTR was improved to 236.1 ps. Comparing the image quality between the previous (420-590 keV) and the optimized (440-590 keV) energy windows, the background variability of the IQ phantom image was reduced by 5.4%, and the contrast in the Hoffman phantom image was increased by 14.5%. In conclusion, we significantly enhanced imaging performance of the helmet-type TOF-PET prototype by using the optimized energy window. We found that the lower energy window setting was a key factor to get a higher TOF-NECR and better image quality in compact TOF-PET systems.

Keywords: PET, energy window, time-of-flight (TOF)
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 296

Poster Number:
M-08-296

Impact of De-noising and Contrast Enhancement on Segmenting Small PET Features with Low Contrast (#1362)

J. - C. (. Cheng1, 2, C. W. J. Bevington2, V. Sossi2

1 The University of British Columbia, Pacific Parkinson's Research Centre, Vancouver, British Columbia, Canada
2 The University of British Columbia, Physics and Astronomy, Vancouver, British Columbia, Canada

Content

We describe the impact of an advanced PET image reconstruction method, which achieves model/prior-free de-noising while enhancing image contrast, on segmenting small PET unique features (< 5 mm in diameter) with low contrast (< 1.7 : 1 contrast ratio). Results from brain imaging simulations showed that the model/prior-free de-noised reconstruction improved the segmentation accuracy by 3-4 times as compared to standard reconstruction with post filter. A further 3-4 times improvement in segmentation accuracy was achieved by contrast enhancement with model/prior-free de-noising as compared to model/prior-free de-noising alone. Contrast enhancement with de-noising was also observed to capture small PET features with low contrast more reproducibly than de-noising alone. Similar trends were observed for data obtained from human studies as well.

Acknowledgment

This work was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) grant.

Keywords: PET segmentation, De-noising, Contrast enhancement
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 299

Poster Number:
M-08-299

A study of kernel characteristics in HYPR post-processing denoising (#1642)

C. W. J. Bevington1, J. - C. (. Cheng1, 2, V. Sossi1

1 University of British Columbia, Department of Physics and Astronomy, Vancouver, British Columbia, Canada
2 University of British Columbia, Pacific Parkinson's Research Centre, Vancouver, British Columbia, Canada

Content

Denoising methods in PET typically exchange a reduction in noise for a small induced bias and loss of resolution. HighlY constrained backPRojection (HYPR) is a denoising operator for dynamic (4D) datasets that overcomes the issue of resolution loss; it uses a 3D composite to return high frequency spatial features lost during filtering. However, mismatches in contrast between the composite and individual dynamic images will introduce biases that propagate into the temporal domain. In this work, we use simulations to compare the effect of different filtering kernel type (boxcar or Gaussian) and size on these biases and denoising efficacy, as well as introduce a new version of the HYPR operator that adds a deconvolution step after filtering to minimize the bias introduced by the composite. We find that boxcar and Gaussian kernel types perform similarly, provided that the central moments of the are matched. Larger kernel sizes increase noise reduction, but also increase bias and amplify errors in contrast. Our proposed HYPR operator largely removes these bias and contrast errors, with a small increase in noise.

AcknowledgmentThis work was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) grant.
Keywords: Denoising, Dynamic Imaging, Iterative Deconvolution
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 302

Poster Number:
M-08-302

Flood Histogram Quality Metric for Light Sharing Depth-Encoding PET Modules (#1973)

A. LaBella1, X. Cao2, W. Zhao3, A. H. Goldan3

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

Content

Spatial performance of single-ended readout depth-encoding PET modules with multicrystal scintillator arrays that are n-to-1 coupled to readout pixels relies on the ability to identify the crystal in which each gamma ray is absorbed based on light sharing patterns. Energy weighted average method is the most popular method for performing crystal identification in such detector modules. However, quantitative metrics that characterize flood histogram quality haven’t yet been developed for this practical, cost-effective detector module configuration. In this work, we introduce a flood histogram quality metric that determines how well-separated the crystal identification clusters are when coupling multiple crystals to the same readout pixel. We compare the flood histogram quality between 4-to-1 coupled modules with a standard uniform glass light guide and our newly developed prismatoid light guide array, which is used in Prism-PET detector module configurations. Both modules consisted of 16 x 16 arrays of 1.4 x 1.4 x 20 mm3 LYSO crystals coupled to 3.2 x 3.2 mm2 SiPM pixels. The Prism-PET module exhibited 40% better flood histogram quality than the uniform light guide module. Crystal clusters acquired at 5 different depths in both modules demonstrated how Prism-PET increases the depth-dependence of crystal contours, thus enhancing crystal separation. Our flood histogram quality metric is a quantitative measure that helps characterize high resolution single-ended readout modules with n-to-1 crystal-to-readout coupling.

AcknowledgmentWe would like to acknowledge our colleagues at PETsys Electronics SA
Keywords: Centroiding, DOI, Prism-PET, Light Guide, Light Sharing
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.
2:00 PM Poster panel: 305

Poster Number:
M-08-305

Harmonization of PET image reconstruction parameters on the Siemens ECAT HR+ PET and the GE Discovery MI PET/CT scanners (#2221)

C. Lois1, 3, K. S. Grogg1, 3, J. C. Price2, 3, G. El Fakhri1, 3, K. A. Johnson1, 3

1 Massachusetts General Hospital, Gordon Center for Medical Imaging, Boston, Massachusetts, United States of America
2 Massachusetts General Hospital, Athinoula A. Martinos Center for Biomedical Imaging, Boston, Massachusetts, United States of America
3 Harvard Medical School, Boston, Massachusetts, United States of America

Content

The quantitative performance of PET scanners varies with detector hardware and image reconstruction methods. These differences produce systematic inter-scanner variability and hamper the ability to compare and combine data between scanners. In this work, we aim at identifying optimal harmonized reconstruction parameters that achieve the most comparable quantitative performance between data acquired using a GE-DMI and a Siemens HR+ scanner. We found that GE-DMI reconstruction parameters of OSEM 34 subsets 2 iterations (34s2i), 6 mm gaussian filter, no time of flight, no point spread function modeling and 1.5 minute frames achieve the most comparable quantitative values to the legacy HR+ data with OSEM 16s4i 5 mm gaussian smoothing and 5 minute frames.

Keywords: Harmonization, PET, quantification
This presentation is available for registered delegates only. Please login to the conference or contact abstract@eventclass.org.

Our exhibitors and sponsors:

Please support our exhibitors by visiting their virtual booths between Monday 9:00 AM and Friday 6:00 PM. Click on the company logo in the scrolling list above to the website now and to the exhibitor booth during the conference.
Their booths will be open for viewing equipment 24 hours a day and for live discussions during conference hours for the whole conference week.