Deep residual-convolutional neural networks for event positioning in a monolithic annular PET scanner

Abstract

PET scanners based on monolithic pieces of scintillator can potentially produce superior performance characteristics (high spatial resolution and detection sensitivity, for example) compared to conventional PET scanners. Consequently, we initiated development of a preclinical PET system based on a single 7.2 cm long annulus of LYSO, called AnnPET. While this system could facilitate creation of high-quality images, its unique geometry results in optics that can complicate estimation of event positioning in the detector. To address this challenge, we evaluated deep-residual convolutional neural networks (DR-CNN) to estimate the three-dimensional position of annihilation photon interactions. Monte Carlo simulations of the AnnPET scanner were used to replicate the physics, including optics, of the scanner. It was determined that a ten-layer-DR-CNN was most suited to application with AnnPET. The errors between known event positions, and those estimated by this network and those calculated with the commonly used center-of-mass algorithm (COM) were used to assess performance. The mean absolute errors (MAE) for the ten-layer-DR-CNN-based event positions were 0.54 mm, 0.42 mm and 0.45 mm along the x (axial)-, y (transaxial)- and z- (depth-of-interaction) axes, respectively. For COM estimates, the MAEs were 1.22 mm, 1.04 mm and 2.79 mm in the x-, y- and z-directions, respectively. Reconstruction of the network-estimated data with the 3D-FBP algorithm (5 mm source offset) yielded spatial resolutions (full-width-at-half-maximum (FWHM)) of 0.8 mm (radial), 0.7 mm (tangential) and 0.71 mm (axial). Reconstruction of the COM-derived data yielded spatial resolutions (FWHM) of 1.15 mm (radial), 0.96 mm (tangential) and 1.14 mm (axial). These findings demonstrated that use of a ten-layer-DR-CNN with a PET scanner based on a monolithic annulus of scintillator has the potential to produce excellent performance compared to standard analytical methods.

A High-Resolution Portable Gamma-Camera for Estimation of Absorbed Dose in Molecular Radiotherapy

“…Compared to commercial gamma-cameras, our detector is around 3 times better in term of intrinsic spatial resolution for an equivalent scintillator thickness. Some research teams have investigated the use of neural networks for the same task in PET systems [19], [39], [40]. Our experimental results, obtained at a lower gamma energy, are similar or better than what can be found in the literature, even compared to neural networks trained and tested on simulation data.…”

Evaluation of monolithic crystal detector with dual-ended readout utilizing multiplexing method

“…Adding a layer of horizontal Polaroid between the crystal and SiPM improves system spatial resolution (Version 2020). A PET system constructed using a single annulus of LYSO with a 72 mm axial length has been explored (Stolin et al 2017, Jaliparthi et al 2021. Additionally, deep neural networks have been employed to enhance system spatial resolution (Sanaat and Zaidi 2020), and a whole-body clinical scanner has been constructed using thick crystals and Philips Digital Photon Counting (PDPC) (Mikhaylova et al 2017).…”

A deep neural network for positioning and inter-crystal scatter identification in multiplexed PET detectors: a simulation study

“…More recently, positioning algorithms based on deep learning methods, such as deep neural networks (DNNs) and convolutional neural networks (CNNs), have been reported for both monolithic PET detectors (Iborra et al 2019, Sanaat and Zaidi 2020, Jaliparthi et al 2021, Carra et al 2022 and light-sharing pixelated PET detectors (Labella et al 2020, Lee and Lee 2021, Petersen et al 2024. Contrary to other machine learning techniques (e.g.…”