IEEE 2021 NSS MIC

Please note! All times in the online program are given in New York - America (GMT -04:00) times!

New York - America ()
Jan 26, 2022, 3:17:23 PM
Your time ()
n/a
Tokyo - Asia ()
Jan 27, 2022, 5:17:23 AM
Our exhibitors and sponsors – click on name to visit booth:

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

MIC Mini-Oral I: Student Paper Award Poster Competition

Session chair: Abbaszadeh , Shiva (University of California, Santa Cruz, USA); Akamatsu , Go (National Institutes for Quantum and Radiological Science and Technology (QST), Chiba, Japan)
 
Shortcut: MO-01
Date: Wednesday, 20 October, 2021, 11:40 AM - 2:00 PM
Room: MIC - 1
Session type: MIC Session

Contents

Click on an contribution to preview the abstract content.

11:40 AM MO-01-01

A temperature-dependent gain compensation technique for positron emission tomography detectors based on silicon photomultiplier (#324)

H. S. Shim1, 2, H. Park4, J. S. Lee3

1 Seoul National University, Interdisciplinary Program of Bioengineering, Seoul, Republic of Korea
2 Seoul National University, Integrated Major in Innovative Medical Science, Seoul, Republic of Korea
3 Seoul National University, Department of Biomedical Sciences, Seoul, Republic of Korea
4 Seoul National University, Department of Nuclear Medicine, Seoul, Republic of Korea

Abstract

In this study, we suggest a simple gain compensation technique for SiPM-based PET detectors using a temperature sensor that automatically controls the bias voltage of SiPM depending on the ambient temperature. The output of the temperature sensor of which the temperature coefficient can be controlled with the input voltage is used as one end of the bias voltage. By adjusting the temperature coefficient, the proposed gain compensation method can be applied to various SiPMs with different breakdown voltage. For the proof of concept, the proposed method was evaluated on two scintillation detector setups. Applying the proposed method to a single-channel SiPM (ASD-NUV3S-P; AdvanSiD, Italy) coupled with a 3 × 3 × 20 mm3 LGSO crystal, the 511-keV photopeak position in the energy histogram changed by only 1.52% per 10℃ while it changed by 13.27% per 10℃ without gain compensation. On 4 × 4 array MPPC (S14161-3050HS-04; Hamamatsu, Japan) coupled with 3.12 × 3.12 × 15 mm3 4 × 4 LSO array, the photopeak changes with and without gain compensation were 2.42% and 17.06% per 10℃, respectively. The energy resolution degradation of SiPM-based scintillation detectors operating at varying temperatures between 10 °C and 30 °C was negligible when the proposed gain compensation method was applied.

Keywords: PET, PET detector, gain compensation, temperature, temperature sensor
11:50 AM MO-01-02

Monte-Carlo Modelling of a WristPET Scanner for Non-Invasive Measurement of the Arterial Input Function (#572)

M. I. Akerele1, S. A. Nehmeh1

1 Weill Cornell Medical College, Department of Radiology, New York, New York, United States of America

Abstract

Kinetic analysis of PET studies requires measurement of patient-specific arterial input function (PSAIF), an invasive and risky procedure. The aim of this study is to assess the feasibility to image-derive the AIF using a miniature wrist PET device (wristPET) using Monte Carlo (MC) simulation. The scanner hosted 11 rings, each with 80 detector modules. Each module is an array of 5×5 LSO crystals (1×1×20mm3). The scanner was simulated with axial FOV’s of 15mm, 25mm, and 55mm to study the effect of system sensitivity on the accuracy of image-derived input function (IDIF). The wrist was simulated using a cylindrical phantom with a 3mm diameter line source positioned at 2.52 mm off-center to mimic the radial artery. Dynamic PET data were simulated using as input the Time-Activity-Curves of AIF and normal tissue background deduced from a [11C]-DPA-713 PET brain study.  The PET images were reconstructed with CASTOR using MLEM algorithm with 4 subsets and up to 100 iterations. Logan VT kinetic modeling was performed for eight brain segments, using the PSAIF and the wrsitPET IDIF.  We compared the IDIF and PSAIF using area under curve (AUC) and distribution volume (VT). Variability in VT was assessed for selected brain regions using Bland-Altman analysis. The results show that for all FOVs, the AUC_ratio and the % difference in VT consistently improves with increasing number of iterations. The 55mm FOV performed the best, mostly due to the increased system sensitivity compared to the 15mm and 25mm axial FOV’s, with less deviation from the line of unity (1 for the AUC ratio, and 0 for % difference). Thirty iterations allowed reproducing the PSAIF-based VT values, for all the three axial FOV’s, to within the limits of agreements (38%) that were previously determined from a test-retest [11C]-DPA-713 brain study.   Our preliminary results show the feasibility to accurately recover the arterial input function from the wrist radial artery using a dedicated miniature wristPET.

Keywords: IDIF, Monte-Carlo simulation, kinetic analysis, wrist artery
12:00 PM MO-01-03

Brain-dedicated Prism-PET Scanner Performance Evaluation: A Preliminary GATE Simulation Study (#878)

X. Cao1, Z. Wang2, X. Zeng1, A. LaBella3, W. Zhao3, A. Goldan3

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

Abstract

The state-of-the-art PET scanner models that emerged in recent years, Biograph Vision and Explorer, have just demonstrated with significant and comprehensive performance enhancement on many aspects compared to the previous generation of scanners, specifically on spatial resolution and sensitivity gains by the introduction of excellent time-of-flight (TOF) and long axial field-of-view (FOV) coverage, respectively. However, the next line of state-of-the-art PET scanners has focused on large FOV geometries, whereas more practical, cost-effective organ dedicated geometries haven’t been fully developed. We introduce our brain-dedicated Prism-PET scanner with a conformal decagon geometry design to achieve ultra-high spatial resolution and high sensitivity simultaneously and compare its preliminary simulation performance with the Biograph Vision and the Explorer. Geant4 application for tomographic emission (GATE) was used in this work for all scanners' performance evaluation of spatial resolution, count rate, and sensitivity. We see a great potential of conformal Prism-PET brain scanner for further development, making the high performance and relatively low-cost PET scanner possible for future research and clinical use.

Keywords: Prism-PET, Decagon, DOI, TOF, Sensitivity
12:10 PM MO-01-04

Performance of detector modules for a second-generation RF-penetrable brain TOF-PET insert for simultaneous PET/MRI (#785)

Q. Dong1, 2, S. Sajedi2, Z. Adams2, C. - M. Chang2, I. Sacco2, R. Watkins2, C. Levin2, 3

1 Stanford University, Department of Electrical Engineering, Stanford, California, United States of America
2 Stanford University, Department of Radiology, Molecular Imaging Program, Stanford, California, United States of America
3 Stanford University, Department of Bioengineering, Stanford, California, United States of America

Abstract

Multimodality imaging is a very promising field. The tremendous success of integrating positron emission tomography (PET) with computed tomography (CT) has demonstrated the beneficial role it can play in clinical diagnosis. Compared to CT, MRI is similar in spatial resolution and offers better soft-tissue contrast, zero ionizing radiation exposure, simultaneous imaging with PET, and functional information with functional MRI (fMRI) and MR spectroscopy (MRS). Currently, clinical permanently-integrated PET/MRI systems are available. However, the high cost impedes the widespread availability of PET/MR imaging. An alternative approach is to develop a portable PET insert that leverages the existing MRI units and is thus more affordable than the commercial permanently-integrated PET/MRI solution. We are developing a cost-efficient RF-penetrable time of flight (TOF)-PET insert for simultaneous PET/MRI, which is portable, MR-compatible, and TOF-capable. In this paper, we evaluate the performance of two fully assembled detector modules comprising 128 x 6 x 2 LYSO crystal elements coupled to matching SiPMs for our RF-penetrable TOF-PET insert design. The global coincidence time resolution, energy resolution, coincidence count rate, and detector temperature achieved over 10 1-minute datasets were 238.0 ± 0.4  ps  FWHM,  11.5 ± 0.02% FWHM,  25.1 ± 0.2  kcps,  and  22.6 ± 1.0◦C, respectively. These results demonstrate the excellent ToF capability and the performance stability for a scalable detector module.

AcknowledgmentThis work was supported in part by NIH Grants 3R01EB01946504 and 3R01EB01946504S2.
Keywords: TOF-PET, simultaneous PET/MRI, RF-penetrable, PET insert, MR-compatible
12:20 PM MO-01-05

A Prototype Conformal Brain PET Scanner with Depth-Encoding Prism-PET Modules (#728)

X. Zeng1, Z. Wang2, X. Cao1, A. Labella3, W. Zhao3, A. Goldan3

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

Abstract

We have developed a conformal positron emission tomography (PET) scanner prototype with a decagon geometry for the human brain imaging based on our single-ended readout prism-PET depth-encoding detector modules. The scanner consists of 40 prism PET detector modules arranged in a short diameter of 29.1 cm and a long diameter of 38.5 cm decagon-shaped ring with the axial field of view (FOV) of 26.55 mm. The Prism-PET detector module consists of a 16 × 16 array of 1.5 × 1.5 × 20 mm3 lutetiumyttrium oxyorthosillicate (LYSO) scintillator crystals with 1.6 mm crystal pitch coupled 4-to-1 on one end to an 8 × 8 multipixel photon counters (MPPC) readout arrays with 3.2 × 3.2 mm2 silicon photomultipliers (SiPM) pixels and to a prismatoid light guide on the opposite end. An Ultra-Micro Hot Spot Phantom Study demonstrates the high spatial resolution of the scanner. The average spatial resolution measured with the 18F point source is 1.67 mm full width at half maximum (FWHM) in the transaxial direction using filtered back projection (FBP). These results indicate that our conformal Prism-PET brain scanner has high potential to provide ultra-high resolution at a low cost for human brain studies.

Keywords: Brain-dedicated PET, Prism-PET, Decagon geometry, High-resolution
12:30 PM MO-01-06

Reduction of the low energy threshold through new modular data acquisition electronics for cross-strip cadmium zinc telluride (CZT) based PET system (#712)

Y. Wang1, S. Abbaszadeh1

1 University of California, Santa Cruz, Eletrical and Computer Engineering, Santa Cruz, California, United States of America

Abstract

New modular data acquisition electronics for cross-strip cadmium zinc telluride (CZT) detector readout were developed to lower the electronic noise and increase the applied voltage bias capabilities compared to previous designs.  The modular data acquisition electronics will be utilized to develop a two-panel head and neck dedicated positron emission tomography (PET) system. Each panel will consist of 150 CZT crystals (4x4x0.5 cm3) covering an area of 20x15 cm2 in an edge-on configuration to achieve high detector efficiency at 511 keV. This paper presents the first characterization results of CZT detectors based on the new readout electronics system. Ge-68 and Cs-137 were used as the point source for measuring the energy spectra. Three individual CZT detectors (117 anode channels) were tested together with a full data acquisition chain. The mean FWHM energy resolution across all 117 anode channels is 8.36% ± 0.52% at 511 keV without any correction and the lower keV threshold of the energy spectra resides between 0 keV to 50 keV. For reference, the previous small animal system had a mean FWHM energy resolution of 10.78% ± 1.45% at 511 keV before correction and a lower keV threshold of 100 keV to 150 keV.

AcknowledgmentThe authors acknowledge the support from the National Institute of Biomedical Imaging and Bioengineering of the National Institutes of Health under Award NumberR01EB028091.
Keywords: Positron emission tomography, Cadmium Zinc Telluride, head and neck, energy resolution, lower energy threshold
12:40 PM MO-01-07

Federated Learning-based Deep Learning Model for PET Attenuation and Scatter Correction: A Multi-Center Study (#833)

I. Shiri1, A. Vafaei Sadr2, A. Sanaat1, S. Ferdowsi3, H. Arabi1, H. Zaidi1

1 Geneva University Hospital, Division of Nuclear Medicine and Molecular Imaging, Geneva, Genève, Switzerland
2 Universite de Geneve, Departement de Physique Theorique, Geneva, Genève, Switzerland
3 University of Applied Sciences and Arts of Western Switzerland, Geneva, Genève, Switzerland

Abstract

Recently, deep learning algorithms were applied in PET images for Scatter correction (SC) and attenuation correction (AC).  For building a generalizable and reproducible deep learning model a huge large dataset is needed to tune millions of model parameters. Because of the sensitivity (protection regulation) of medical images and tight regulation, gathering a huge dataset for deep learning model training is the main challenge in the health care system. In this study non-attenuation corrected and CT based attenuation corrected 18F-FDG PET images of 300 patients were enrolled  in this study. Data  consisted of 50 patients from 6 different centers which scanners, image acquisition, and reconstruction vary across the different centers. In this study, we  used a deep residual network as the main architecture. . For model evaluation, voxel-wise, mean absolute error (MAE),  absolute relative error (ARE%) and structural similarity index (SSIM) were calculated between ground truth CT based attenuation/scatter corrected and predicted PET images. We implemented server aggregate federated learning workflow  which included step1-3: (1) central global model distributed through several different departments and then (2) models will be trained in each center separately and finally (3) local trained models returned to central server and model aggregated as central global models. Steps 1-3 repeated until the model is fully trained and converged. Quantitative analysis of results  was showed MAE: 0.43±0.01,  ARE%: 15.0±8.8 and  SSIM:0. 90±0.09 in the test set. In this study we built a deep learning based AC/SC model for PET images using six different centers without sharing the dataset.  Federated learning algorithms provide this opportunity to build a model using multicenter data set without sharing data.

AcknowledgmentThis work was supported by the Swiss National Science Foundation under grant SNRF 320030_176052; the Swiss Cancer Research Foundation under Grant KFS-3855-02-2016
Keywords: Federated Learning, Distributed Learning, Deep Learning, Privacy, PET
12:50 PM MO-01-08

Optimization of FPGA-based Gradient Tree Boosting Models for Position Estimation in PET Detectors (#769)

K. Krueger1, F. Mueller1, P. Gebhardt1, B. Weissler1, 2, D. Schug1, 2, V. Schulz1, 2

1 RWTH Aachen University, Physics of Molecular Imaging Systems, Aachen, Germany
2 Hyperion Hybrid Imaging Systems GmbH, Aachen, Germany

Abstract

A key challenge in PET is deducing the interaction position of a gamma interaction in the scintillator of depth-of-interaction (DOI) capable detectors from the measured light distribution. Gradient tree boosting (GTB) has been introduced as a method for positioning in planar and DOI direction in radiation detectors. GTB is a supervised machine learning technique based on building ensembles of independent predictive binary decision trees. Its computational efficiency and low memory requirement make GTB feasible for an integration into the FPGA of a PET system architecture for real-time positioning of gamma interactions. Here, we investigate the influence of different input feature sets on the positioning performance and logic resource (LUT) consumption of a previously presented FPGA-based GTB implementation. GTB models were trained using data acquired with a pixelated high-resolution LYSO scintillator coupled to a sensor array consisting of 16 digital SiPMs (DPC-3200-22, Philips Digital Photon Counting) with 64 photon channels. From these data, GTB models with varying model parameters were trained for five different feature sets including combinations of the 64 raw photon counts, row and column sums, center of gravity, total photon sum and the main channel. All models as well as the calculation of features from the raw data were implemented on a XC7K410T Kintex-7 FPGA. Using row and column sums instead of raw data as well as adding the center of gravity as a feature improves PA. LUT consumption and can be reduced by about 40 % when reducing the number of input features by using row and column sums instead of raw photon counts. A consumption of about 2500 LUTs already provides a positioning performance of 98.63 % of the best achieved performance (at a cost of about 45000 LUTs), which shows the feasibility of implementing GTB models offering real-time positioning for PET systems inside an FPGA.

Keywords: Gradient Tree Boosting, PET Gamma Positioning, Machine Learning, FPGA, PET
1:00 PM MO-01-09

Inter-frame motion correction for whole-body parametric imaging using long short-term memory in a deep convolutional framework (#178)

X. Guo1, B. Zhou1, D. Pigg2, B. Spottiswoode2, M. E. Casey2, C. Liu1, 3, N. C. Dvornek1, 3

1 Yale University, Department of Biomedical Engineering, New Haven, Connecticut, United States of America
2 Siemens Medical Solutions USA, Inc., Knoxville, Tennessee, United States of America
3 Yale University, Department of Radiology and Biomedical Imaging, New Haven, Connecticut, United States of America

Abstract

Inter-frame body motion in whole-body dynamic PET has a serious impact on parametric imaging. Traditional non-rigid registration motion correction methods are generally time-consuming and computationally expensive. Deep learning approaches offer the promise of fast inference with high accuracy, but have yet been investigated in the whole-body scope or with consideration for tracer distribution changes.  In this work, we aimed to develop a deep learning framework to correct inter-frame body motion. The displacement estimation network is a convolutional neural network with a combined long short-term memory layer, extracting dynamic temporal features in addition to spatial information, and the following spatial transformation layer warps the moving frames with the displacement field output. Under a 9-fold cross-validation with 27 subjects, our proposed framework further enhanced qualitative and quantitative spatial alignment between parametric Ki and Vb images, significantly reduced fitting error in parametric estimation compared with both traditional and deep learning baselines, and was around 60 times faster than the traditional registration baseline.

Acknowledgment

This work was supported by NIH grant R01 CA224140 and a research contract from Siemens Medical Solutions USA, Inc.

Keywords: convolutional network, long-short term memory, motion correction, whole-body parametric imaging
1:10 PM MO-01-10

PSD Neutron Discrimination for Dose Monitoring Applications in Particle Therapy (#1224)

L. Buonanno1, 2, I. D'Adda1, 2, D. Di Vita1, 2, A. Caracciolo1, L. Malentacca1, M. Carminati1, 2, C. Fiorini1, 2

1 Politecnico di Milano, Dipartimento di Elettronica Informazione e Bioingegneria, Milano, Italy
2 Istituto Nazionale di Fisica Nucleare, Sezione di Milano, Milano, Italy

Abstract

We present an analytical and experimental study aimed to evaluate Pulse Shape Discrimination (PSD) for neutron rejection in medical imaging applications, with particular focus on dose monitoring in hadrontherapy. This topic is of particular interest for carbon ion beam therapy, where the neutron background fluctuations challenge the prompt gammas profile identification associated to the Bragg peak. The first integrated circuit solution we present aims at avoiding dense waveform sampling, operating the PSD analysis with two signal samples. Analytical modeling and simulations show that the readout noise doesn't worsen the measured PSD fluctuations between 100 keV and 10 MeV. Experimental characterization of a custom miniaturized DAQ module with PSD capability, collecting the signal from a 2"x2" cylindrical CLYC crystal coupled to an 8x8 array of NUV-HD SiPM, will be presented. The second integrated circuit solution involves typical temporal structures of the delivered beam. Rejection of interaction generating the scintillation light outside a short gate correlated with prompt gamma photons time of arrival allows to largely reduce the neutron background. We will also present at the conference the characterization of this time-gated solution, using a custom laser-based scintillation crystal emulator.

Keywords: clyc, psd, sipm, asic, particle therapy
1:20 PM MO-01-11

Timepix3 based Single Layer X ray Fluorescence Compton Camera (#683)

C. Wu1, L. Li1, M. Zen1, J. Wen1, Y. Zhang1

1 Tsinghua University,Beijing,China / Department of Engineering Physics, Beijing, China

Abstract

XFCT (X-ray fluorescence CT) is a new type of medical imaging system which can present molecular and functional information in organisms. XF (X-ray fluorescence) imaging has the advantages of low cost, easy storage and transportation of contrast agents and low internal radiation. However, the mechanical collimation limits the detection efficiency of the system. Compton camera is an imaging system composed of double-layer detectors, which can obtain the direction of incident rays without a collimator. So combining the advantages of Compton camera system and XF photon imaging is a solution worth exploring. Timepix3 is one of the most advanced high resolution photon counting detectors, which has well time resolution and can obtain the depth position where particle deposit energy. In this work, We use Si-Timepix3 single-layer detector to build up a Compton camera system, in order to detect the 43keV XF photons from Gd solution excited by X-ray. We establish the first real experimental platform to verify the feasible of XFCC(X-ray Fluorescence Compton Camera) system and obtain the reconstruction images.

AcknowledgmentThis work was partially funded by grants from NNSFC 11775124, the National Key Research and Development Program of China 2018YFC0115502.
Keywords: X-ray Fluorescence, Compton Camera, medical imaging, Timepix3 detector

Our exhibitors and sponsors – click on name to visit booth: