Cloud-based serverless computing enables accelerated monte carlo simulations for nuclear medicine imaging.

Objective.This study investigates the potential of cloud-based serverless computing to accelerate Monte Carlo (MC) simulations for nuclear medicine imaging tasks. MC simulations can pose a high computational burden-even when executed on modern multi-core computing servers. Cloud computing allows simulation tasks to be highly parallelized and considerably accelerated.Approach.We investigate the computational performance of a cloud-based serverless MC simulation of radioactive decays for positron emission tomography imaging using Amazon Web Service (AWS) Lambda serverless computing platform for the first time in scientific literature. We provide a comparison of the computational performance of AWS to a modern on-premises multi-thread reconstruction server by measuring the execution times of the processes using between105and2·1010simulated decays. We deployed two popular MC simulation frameworks-SimSET and GATE-within the AWS computing environment. Containerized application images were used as a basis for an AWS Lambda function, and local (non-cloud) scripts were used to orchestrate the deployment of simulations. The task was broken down into smaller parallel runs, and launched on concurrently running AWS Lambda instances, and the results were postprocessed and downloaded via the Simple Storage Service.Main results.Our implementation of cloud-based MC simulations with SimSET outperforms local server-based computations by more than an order of magnitude. However, the GATE implementation creates more and larger output file sizes and reveals that the internet connection speed can become the primary bottleneck for data transfers. Simulating 109decays using SimSET is possible within 5 min and accrues computation costs of about $10 on AWS, whereas GATE would have to run in batches for more than 100 min at considerably higher costs.Significance.Adopting cloud-based serverless computing architecture in medical imaging research facilities can considerably improve processing times and overall workflow efficiency, with future research exploring additional enhancements through optimized configurations and computational methods.

Performance Characteristics of the NeuroEXPLORER, a Next-Generation Human Brain PET/CT Imager.

The collaboration of Yale, the University of California, Davis, and United Imaging Healthcare has successfully developed the NeuroEXPLORER, a dedicated human brain PET imager with high spatial resolution, high sensitivity, and a built-in 3-dimensional camera for markerless continuous motion tracking. It has high depth-of-interaction and time-of-flight resolutions, along with a 52.4-cm transverse field of view (FOV) and an extended axial FOV (49.5 cm) to enhance sensitivity. Here, we present the physical characterization, performance evaluation, and first human images of the NeuroEXPLORER. Methods: Measurements of spatial resolution, sensitivity, count rate performance, energy and timing resolution, and image quality were performed adhering to the National Electrical Manufacturers Association (NEMA) NU 2-2018 standard. The system's performance was demonstrated through imaging studies of the Hoffman 3-dimensional brain phantom and the mini-Derenzo phantom. Initial 18F-FDG images from a healthy volunteer are presented. Results: With filtered backprojection reconstruction, the radial and tangential spatial resolutions (full width at half maximum) averaged 1.64, 2.06, and 2.51 mm, with axial resolutions of 2.73, 2.89, and 2.93 mm for radial offsets of 1, 10, and 20 cm, respectively. The average time-of-flight resolution was 236 ps, and the energy resolution was 10.5%. NEMA sensitivities were 46.0 and 47.6 kcps/MBq at the center and 10-cm offset, respectively. A sensitivity of 11.8% was achieved at the FOV center. The peak noise-equivalent count rate was 1.31 Mcps at 58.0 kBq/mL, and the scatter fraction at 5.3 kBq/mL was 36.5%. The maximum count rate error at the peak noise-equivalent count rate was less than 5%. At 3 iterations, the NEMA image-quality contrast recovery coefficients varied from 74.5% (10-mm sphere) to 92.6% (37-mm sphere), and background variability ranged from 3.1% to 1.4% at a contrast of 4.0:1. An example human brain 18F-FDG image exhibited very high resolution, capturing intricate details in the cortex and subcortical structures. Conclusion: The NeuroEXPLORER offers high sensitivity and high spatial resolution. With its long axial length, it also enables high-quality spinal cord imaging and image-derived input functions from the carotid arteries. These performance enhancements will substantially broaden the range of human brain PET paradigms, protocols, and thereby clinical research applications.

Facial Anonymization and Privacy Concerns in Total-Body PET/CT.

Total-body PET/CT images can be rendered to produce images of a subject's face and body. In response to privacy and identifiability concerns when sharing data, we have developed and validated a workflow that obscures (defaces) a subject's face in 3-dimensional volumetric data. Methods: To validate our method, we measured facial identifiability before and after defacing images from 30 healthy subjects who were imaged with both [18F]FDG PET and CT at either 3 or 6 time points. Briefly, facial embeddings were calculated using Google's FaceNet, and an analysis of clustering was used to estimate identifiability. Results: Faces rendered from CT images were correctly matched to CT scans at other time points at a rate of 93%, which decreased to 6% after defacing. Faces rendered from PET images were correctly matched to PET images at other time points at a maximum rate of 64% and to CT images at a maximum rate of 50%, both of which decreased to 7% after defacing. We further demonstrated that defaced CT images can be used for attenuation correction during PET reconstruction, introducing a maximum bias of -3.3% in regions of the cerebral cortex nearest the face. Conclusion: We believe that the proposed method provides a baseline of anonymity and discretion when sharing image data online or between institutions and will help to facilitate collaboration and future regulatory compliance.

Relating18F-FDG image signal-to-noise ratio to time-of-flight noise-equivalent count rate in total-body PET using the uEXPLORER scanner.

Objective.This work assessed the relationship between image signal-to-noise ratio (SNR) and total-body noise-equivalent count rate (NECR)-for both non-time-of-flight (TOF) NECR and TOF-NECR-in a long uniform water cylinder and 14 healthy human subjects using the uEXPLORER total-body PET/CT scanner.Approach.A TOF-NEC expression was modified for list-mode PET data, and both the non-TOF NECR and TOF-NECR were compared using datasets from a long uniform water cylinder and 14 human subjects scanned up to 12 h after radiotracer injection.Main results.The TOF-NECR for the uniform water cylinder was found to be linearly proportional to the TOF-reconstructed image SNR2in the range of radioactivity concentrations studied, but not for non-TOF NECR as indicated by the reducedR2value. The results suggest that the use of TOF-NECR to estimate the count rate performance of TOF-enabled PET systems may be more appropriate for predicting the SNR of TOF-reconstructed images.Significance.Image quality in PET is commonly characterized by image SNR and, correspondingly, the NECR. While the use of NECR for predicting image quality in conventional PET systems is well-studied, the relationship between SNR and NECR has not been examined in detail in long axial field-of-view total-body PET systems, especially for human subjects. Furthermore, the current NEMA NU 2-2018 standard does not account for count rate performance gains due to TOF in the NECR evaluation. The relationship between image SNR and total-body NECR in long axial FOV PET was assessed for the first time using the uEXPLORER total-body PET/CT scanner.

Study of Cerenkov Light Emission in the Semiconductors TlBr and TlCl for TOF-PET.

Thallium bromide (TlBr) and thallium chloride (TlCl) are semiconductor materials with high transparency to visible light, high index of refraction, and high detection efficiency for gamma rays and annihilation photons. This manuscript reports on measurements of the light intensity and timing response of Cerenkov light emitted in one 3 mm × 3 mm × 5 mm slab of each of these materials operated in coincidence with a lutetium fine silicate (LFS) crystal with dimensions of 3 mm × 3 mm × 20 mm. A 22Na radioactive source was used. The measured average number of detected photons per event was 1.5 photons for TlBr and 2.8 photons for TlCl when these materials were coupled to a silicon photomultiplier. Simulation predicts these results with an overestimation of 12%. The best coincidence time resolution (CTR) for events in TlBr and TlCl were 329 ± 9 ps and 316 ± 9 ps, respectively, when events with 4 photons and >7 photons were selected. Simulation showed the CTR degraded from 120 ps to 405 ps in TlCl, and from 160 ps to 700 ps in TlBr when the first or second Cerenkov photon were selected. Results of this work show TlCl has a stronger Cerenkov light emission compared to TlBr and a greater potential to obtain the best timing measurements. Results also stress the importance of improving detection efficiency and transport of light to capture the first Cerenkov photon in timing measurements.

Characterization of four readout circuits for an MR compatible, preclinical PET detector.

We are building a high sensitivity preclinical PET/MRI insert using a highly multiplexed light sharing PET module. Each module incorporates four 19 × 19 arrays of 1 × 1 × 20 mm3 LYSO crystals with dual-ended DOI encoding readout, requiring 32 readout channels for positioning and eight channels for timing. These constraints necessitate compact, robust electronics for digitization. We have characterized four linearized time-over-threshold circuits based on these detector requirements. The four circuits allow for high channel density and can digitize signals from highly multiplexed light sharing detectors. Each circuit digitizes one channel of a multiplexed SiPM array, yielding a binary output that interfaces directly with an FPGA. Using the optimal circuit, we have characterized the performance of a pair of PET modules. The four circuits were characterized based on linearity of the 22Na photopeak positions and energy resolution at 511 keV, as well as separation of elements in a 10 × 10 array of 1.2 mm LYSO crystals coupled with a specular reflector. Practical measures of performance were comparable to those obtained with a DRS evaluation board, which served as a reference acquisition system. The ratio of the 22Na photopeak positions was 2.0 for each circuit and the reference system, implying 20% saturation due to the SiPM. PET energy resolution of the optimal circuit was 11.8% FWHM for a single crystal versus 12.6% for the reference system, and crystals were equally well separated in all cases. PCBs implementing the optimal readout circuit were fabricated and used to construct two complete detector modules. Crystals in each of the four blocks in the module were well resolved, with a mean energy resolution of 24.4 ± 4.7%. Two modules operating in coincidence showed a single detector timing resolution of 3.0 ns, which is appropriate for preclinical applications.

Optimization of a depth of interaction encoding PET block detector for a PET/MRI insert.

Preclinical positron emission tomography, combined with magnetic resonance imaging (PET/MRI), is increasingly used as a tool to simultaneously characterize functional processes in vivo. Many emerging preclinical applications, however, are limited by PET detection sensitivity, especially when generating short imaging frames for quantitative studies. One such application is dynamic multifunctional imaging, which probes multiple aspects of a biological process, using relationships between the datasets to quantify interactions. These studies have limited accuracy due to the relatively low sensitivity of modern preclinical PET/MRI systems. The goal of this project is to develop a preclinical PET/MRI insert with detection sensitivity above 15% (250-750 keV) to improve quantitation in dynamic PET imaging. To achieve this sensitivity, we have developed a detector module incorporating a 2 cm thick crystal block, which will be arranged into a system with 8 cm axial FOV, targeting mice and rats. To maintain homogenous spatial resolution, the detector will incorporate dual-ended depth-of-interaction (DOI) encoding with silicon photomultiplier (SiPM) based photodetector arrays. The specific aim of this work is to identify a detector configuration with adequate performance for the proposed system. We have optimized the SiPM array geometry and tested two crystal array materials with pitch ranging from 0.8 to 1.2 mm and various surface treatments and reflectors. From these configurations, we have identified the best balance between crystal separation, energy resolution, and DOI resolution. The final detector module uses two rectangular SiPM arrays with 5 × 6 and 5 × 4 elements. The photodetector arrays are coupled to a 19 × 19 array of 1 mm pitch LYSO crystals with polished surfaces and a diffuse reflector. The prototype design has 14.3% ± 2.9% energy resolution, 3.57 ± 0.88 mm DOI resolution, and resolves all elements in the crystal array, giving it sufficient performance to serve as the basis for the proposed high sensitivity PET/MRI insert.

Open-field mouse brain PET: design optimisation and detector characterisation.

'Open-field' PET, in which an animal is free to move within an enclosed space during imaging, is a very promising advance for neuroscientific research. It provides a key advantage over conventional imaging under anesthesia by enabling functional changes in the brain to be correlated with an animal's behavioural response to environmental or pharmacologic stimuli. Previously we have demonstrated the feasibility of open-field imaging of rats using motion compensation techniques applied to a commercially available PET scanner. However, this approach of 'retro-fitting' motion compensation techniques to an existing system is limited by the inherent geometric and performance constraints of the system. The goal of this project is to develop a purpose-built PET scanner with geometry, motion tracking and imaging performance tailored and optimised for open-field imaging of the mouse brain. The design concept is a rail-based sliding tomograph which moves according to the animal's motion. Our specific aim in this work was to evaluate candidate scanner designs and characterise the performance of a depth-of-interaction detector module for the open-field system. We performed Monte Carlo simulations to estimate and compare the sensitivity and spatial resolution performance of four scanner geometries: a ring, parallel plate, and two box variants. Each system was based on a detector block consisting of a 23 × 23 array of 0.785 × 0.785 × 20 mm3 LSO crystals (overall dim. 19.6 × 19.6 × 20 mm). We found that a DoI resolution capability of 3 mm was necessary to achieve approximately uniform sub-millimetre spatial resolution throughout the FoV for all scanners except the parallel-plate geometry. With this DoI performance, the sensitivity advantage afforded by the box geometry with overlapping panels (16% peak absolute sensitivity, a 36% improvement over the ring design) suggests this unconventional design is best suited for imaging the mouse brain. We also built and characterised the block detector modelled in the simulations, including a dual-ended readout based on 6 × 6 arrays of through-silicon-via silicon photomultipliers (active area 84%) for DoI estimation. Identification of individual crystals in the flood map was excellent, energy resolution varied from 12.4% ± 0.6% near the centre to 24.4% ± 3.4% at the ends of the crystal, and the average DoI resolution was 2.8 mm ± 0.35 mm near the central depth (10 mm) and 3.5 mm ± 1.0 mm near the ends. Timing resolution was 1.4 ± 0.14 ns. Therefore, the DoI detector module meets the target specifications for the application and will be used as the basis for a prototype open-field mouse PET scanner.

Rat embryonic hippocampus and induced pluripotent stem cell derived cultured neurons recover from laser-induced subaxotomy.

Axonal injury and stress have long been thought to play a pathogenic role in a variety of neurodegenerative diseases. However, a model for studying single-cell axonal injury in mammalian cells and the processes of repair has not been established. The purpose of this study was to examine the response of neuronal growth cones to laser-induced axonal damage in cultures of embryonic rat hippocampal neurons and induced pluripotent stem cell (iPSC) derived human neurons. A 532-nm pulsed [Formula: see text] picosecond laser was focused to a diffraction limited spot at a precise location on an axon using a laser energy/power that did not rupture the cell membrane (subaxotomy). Subsequent time series images were taken to follow axonal recovery and growth cone dynamics. After laser subaxotomy, axons thinned at the damage site and initiated a dynamic cytoskeletal remodeling process to restore axonal thickness. The growth cone was observed to play a role in the repair process in both hippocampal and iPSC-derived neurons. Immunofluorescence staining confirmed structural tubulin damage and revealed initial phases of actin-based cytoskeletal remodeling at the damage site. The results of this study indicate that there is a repeatable and cross-species repair response of axons and growth cones after laser-induced damage.