Please note! All times in the online program are given in New York - America (GMT -04:00) times!
Dec 5, 2021, 9:23:27 AM
Dec 5, 2021, 11:23:27 PM
Click on an contribution to preview the abstract content.
Preliminary results of the performance characteristics for the new digital uMI Panorama PET/CT system (#1282)
Y. Liu1, Y. Sun1, Y. Wu1, L. He1, D. Hu1, Y. Dong1, S. An1, Y. Ding1, S. Chu1, D. Bi1
1 Shanghai United Imaging Healthcare, Shanghai, China
The uMI Panorama PET/CT system used SiPM- and ASIC-readout-based UDP-C21 detector platform. Due to the use of application specific integrated circuit (ASIC), high counting performance of the system has been significantly improved, which will increase the quantitative accuracy of applications such as dynamic imaging and myocardial perfusion imaging (MPI). The high-precision position decoding capability due to the 2.76 mm crystal size and nine-to-four coupling structure between crystal and SiPM allows for an ultra-high spatial resolution. This study evaluates the performance of the first uMI Panorama prototype according to the NEMA standard. The PET system shows a transaxial spatial resolution at Full Width at Half Maximum of 2.81 mm at 1 cm offset of the center, a sensitivity of 20 cps/kBq at the center of field-of-view, a peak NECR of 440.75 kcps at 36.95 kBq/cc, and a maximum relative count rate error among all slices less than 3% at the activity concentration values at or below NECR peak activity. The overall image contrast seen with the NEMA image quality (NEMA-IQ) phantom ranges from 78.3% to 86.9%, and the background variability ranges from 6.5% to 2.0%. Phantom imaging results indicate that the system has excellent lesion detectability, 1.6-mm rods of mini-Deluxe phantom can be clearly identified.
Keywords: PET-CT, NEMA performance, LYSO, uMI Panorama
The DE-SPECT System: A Clinical SPECT System based on 1-cm Thick CZT Imaging Spectrometers and Dynamic Compound-Eye Collimators for Broadband Multi-Isotope Imaging of the Lower Extremities (#1063)
E. M. Zannoni1, 2, S. Poopalasingam3, S. D. Metzler3, C. Liu4, A. J. Sinusas5, L. J. Meng2, 1
1 University of Illinois, Urbana-Champaign, Bioengineering, Urbana, Illinois, United States of America
In this work, we discuss the design and development of the Dynamic Extremity SPECT (DE-SPECT) system, a stationary and high-performance clinical SPECT system, dedicated to high performance imaging of peripheral vascular disease (PVD). PVD represents a large number of circulatory diseases, mostly prevalent in the lower extremities, and affecting more than 200 million people worldwide. Nowadays, PVD remains largely underdiagnosed and undertreated and requires the development of tailored diagnostic approaches and advanced imaging instrumentation.
We would like to acknowledge the National Institute of Biomedical Imaging and Bioengineering (NIBIB) as funding agency.
Keywords: CZT detector, peripheral vascular disease, SPECT imaging system, organ-dedicated clinical scanner
Optimization of a Hybrid PET (HyPET) Detector for Prostate Cancer Imaging (#188)
R. S. Miyaoka1, W. C. Hunter1, L. A. Pierce1, R. L. Harrison1, E. Lamprou2, J. Barrio2, C. Valladares2, A. J. Gonzalez2
1 University of Washington, Seattle, United States of America
The goal of this project is to develop hybrid PET (HyPET) detector technology that will enable moderate cost, high resolution, prostate specific PET imaging. The optimism for an imaging based prostate cancer (PCa) biomarker is being driven by the development of highly targeted PET tracers for PCa cells. Monte Carlo simulations and a tailored weighted least squares (TWLS) image reconstruction algorithm will be used to investigate the imaging performance of different hybrid PET detector systems. HyPET detectors will consist of a front detector optimized for coincidence timing performance (FFTOF) and a rear detector optimized for spatial resolution, depth of interaction (DOI) positioning and detection efficiency (BBDOI). The TWLS algorithm allows for different weighting parameters between FFTOF, BBDOI and mixed front-back data, allowing the FFTOF data to act as a prior for the rest of the data. Detector performance characteristics were modeled based upon experimentally measured performance results. Front detector designs included discrete crystal arrays composed of 3x3x5 mm^3 or 1x1x3(5) mm^3 LYSO crystals coupled to monolithic SiPM arrays. Back layer detector designs included monolithic crystals with 5, 7.5, 10 or 15 mm thick. The monolithic crystal detectors were characterized for spatial resolution, DOI resolution and timing resolution. A 28 cm diameter cylindrical phantom with a prostate-sized (i.e., 4x3 cm^2 ellipse by 2 cm tall) insert was simulated. In addition, a set of six 2 mm diameter spheres arranged in a triangular matrix was used to assess the image resolution and contrast to noise ratio between the spheres and background prostate uptake. The contrast ratios between sphere and prostate and prostate and background were 10 and 2. The contrast to noise ratios and the conspicuity of the 2 mm diameter spheres will be reported for different HyPET detector combinations and different levels of administered activity.
Keywords: PET detectors, hybrid, prostate cancer, time of flight, depth of interaction
Design of an ultra-low-dose, stationary, tomographic Molecular Breast Imaging system. (#1078)
B. F. Hutton1, K. Erlandsson1, A. Wirth2, I. Baistow2, K. Thielemans1, A. Cherlin2
1 University College London, Institute of Nuclear Medicine, London, United Kingdom
Molecular Breast Imaging (MBI) has been shown to have high sensitivity in detection of cancer, even in patients with dense breasts where conventional mammography has issues. However the technique has limited acceptance due to the relatively high radiation dose and long imaging time. Improved lesion detection can be achieved using tomography, however this normally involves detector motion and complex mechanics. Our goal is to develop a low-dose stationary tomographic MBI system with similar or better sensitivity for lesion detection to conventional planar MBI. The proposed system utilises state-of-the-art cadmium zinc telluride (CZT) detectors based on 2mm pixels, with sub-pixelisation and depth of interaction (DOI) capability, combined with densely packed multi-pinhole collimators. Use of closely-spaced pinholes improves efficiency and angular sampling, but results in significant multiplexing. De-multiplexing algorithms have been developed that take advantage of the DOI acquisition to achieve tomographic reconstruction using two opposing planar detectors which apply mild compression to the breast. Simulation studies of multiple lesions with clinically realistic contrast have been used to demonstrate the feasibility of the design and to characterize the expected performance. Reconstruction without de-multiplexing resulted in significant artefacts. De-multiplexing without DOI had limited success but with DOI resulted in artefact-free images, with good contrast and axial plane definition. Lesion detectability was preserved even with reduction of acquisition time (or radiation dose) by a factor of 5. Further optimisation has potential for even greater dose reduction. A prototype system is currently being constructed to validate these findings.
AcknowledgmentINM staff are partly funded by the NIHR University College London Hospitals Biomedical Research Centre. Kromek are supported by an Innovate UK grant (104296).
Keywords: molecular breast imaging, CZT, multi-pinhole collimator, multiplex, low-dose imaging
First bench-top breast-dedicated PET prototypewith a dual round-edge detector arrangement (#388)
G. Akamatsu1, E. Yoshida1, H. Tashima1, S. Ito2, M. Takahashi1, T. Yamaya1
1 National Institutes for Quantum and Radiological Science and Technology (QST), Chiba, Japan
Breast-dedicated PET systems are categorized into two geometries: a ring-shaped detector arrangement and a dual flat-panel (FP) detector arrangement. Although there are some advantages in the FP arrangement, PET images are blurred due to the limited angular coverage. To compensate for this issue, we have proposed a dual round-edge (RE) detector arrangement, in which detectors at both edge positions are tilted toward the center of the field-of-view. In this study, for a proof-of-concept of the RE arrangement, we developed the first bench-top prototype. Our original crosshair light-sharing (CLS) depth-of-interaction (DOI) detector was used; it consisted of a 14×14 array of 1.45×1.45×15 mm3 GFAG scintillator crystals coupled with an 8×8 array of silicon photomultipliers each with a photosensitive area of 3.0×3.0 mm2. The data acquisition system was the TOFPET2 ASIC (PETsys Electronics). We examined the conventional FP arrangement and the proposed RE arrangement using 32 CLS detectors. A 22Na point source and a multi-rod phantom (rod diameters: 1.6, 2.2, 3.0, 4.0, 5.0 and 6.0 mm) were used for spatial resolution measurements. The RE arrangement showed better spatial resolution compared with the FP arrangement (64% improvement in the vertical direction). For the multi-rod phantom, in the coronal plane (parallel to the detector surface), the 1.6 mm rods were clearly resolved in both arrangements. However, in the axial plane (perpendicular to the detector surface), the RE arrangement resolved the 2.2 mm rods that had not been resolved by the FP arrangement. We demonstrated that the proposed RE detector arrangement showed better spatial resolution than the conventional FP arrangement.
AcknowledgmentThis research is supported by AMED under Grant Number JP20he2202004.
Keywords: PET, breast, round-edge, CLS, DOI
Pre-training and Transfer Learning for Training Set Reduction and Improving Predictions of Automated Clinical Assessment of PET Image Quality (#1330)
J. B. Hopson1, R. Neji3, J. T. Dunn2, V. Kersemans3, C. McGinnity2, A. J. Reader1, A. Hammers1, 2
1 King's College London, School of Biomedical Engineering and Imaging Sciences, London, United Kingdom
Quality scoring of simultaneous positron emission tomography - magnetic resonance (PET-MR) images by clinicians is time-consuming and not scalable. However, clinical image quality assessments ensure that images, especially low-count PET reconstructions, are of sufficient quality. A major challenge is the paucity of clinical readings. We hypothesise that exploiting readily available quantitative information or using available pre-trained networks to predict clinical readings, could reduce the number of clinically assessed datasets required for training convolutional neural networks (CNNs). In this study, a CNN was pre-trained to predict injected dose (in MBq) of patches extracted from real patient datasets, which had been reconstructed using 100% of the available data and different thinned datasets down to 0.5% of available data. Using transfer learning with five separate patients, the CNN was used to predict three clinically scored quality metrics: global quality rating, pattern recognition and diagnostic confidence, based on a four-point scale (0-3). A VGG16 network pre-trained using ImageNet was also explored. This preliminary study shows that pre-training improves test performance compared to a model without pre-training. Pre-training reduces the overall mean absolute error across all three metrics for three unseen patients from 0.81 to 0.54 for dose inference and to 0.51 for VGG16. This proof-of-concept study shows it is possible to exploit accessible quantitative labels or pre-trained networks to predict scarce clinical metrics. Future work will include increasing the clinical dataset number and optimising model architectures.
This work is supported by the Wellcome EPSRC Centre for Medical Engineering at King’s College London (WT 203148/Z/16/Z) and the Department of Health via the NIHR comprehensive Biomedical Research Centre award to GSTT NHS Foundation Trust in partnership with King’s College London and King’s College Hospital NHS Foundation Trust
Keywords: Deep Learning, Image Quality Assessment, PET imaging, Transfer Learning, PET Reconstruction
3D Printing of Germanium-68 PET Phantoms (#347)
L. Meier1, T. Läppchen1, A. Rominger1, M. Hentschel1, G. A. Prenosil1
1 Inselspital, Bern University Hospital and University of Bern, Department of Nuclear Medicine, Bern, Bern, Switzerland
We aimed to print a solid-state germanium-68 phantom for positron emission tomography (PET) with precise control of the activity concentration in the phantom material, without cold walls, and with safe handling characteristics. To this end, we developed a novel method for incorporating aqueous germanium-68 into hydrophobic stereo-lithography monomers. With this method, we additively manufactured an 8 ml sphere with near perfect source tightness and homogeneity. A clinical PET scan showed homogenous activity distribution and no fill-related artefacts as seen in a fillable phantom sphere. The energetically favorable process provides complexation and phase transfer of germanium-68 into the hydrophobic building material, ensures homogenous isotope distribution in the rigid polymer matrix and prevents leaching in aqueous environment. Long-lived solid-state phantoms produced with this method will provide consistent and reproducible results in scanner validation, in quality assurance, and in multicenter clinical studies.
Keywords: 3D-printing, germanium-68, polymer, quantitative imaging, solid state PET/CT phantom