Self-normalization for a 1 mm3resolution clinical PET system using deep learning

Objective.This work proposes, for the first time, an image-based end-to-end self-normalization framework for positron emission tomography (PET) using conditional generative adversarial networks (cGANs).Approach.We evaluated different approaches by exploring each of the following three methodologies. First, we used images that were either unnormalized or corrected for geometric factors, which encompass all time-invariant factors, as input data types. Second, we set the input tensor shape as either a single axial slice (2D) or three contiguous axial slices (2.5D). Third, we chose either Pix2Pix or polarized self-attention (PSA) Pix2Pix, which we developed for this work, as a deep learning network. The targets for all approaches were the axial slices of images normalized using the direct normalization method. We performed Monte Carlo simulations of ten voxelized phantoms with the SimSET simulation tool and produced 26,000 pairs of axial image slices for training and testing.Main results.The results showed that 2.5D PSA Pix2Pix trained with geometric-factors-corrected input images achieved the best performance among all the methods we tested. All approaches improved general image quality figures of merit peak signal to noise ratio (PSNR) and structural similarity index (SSIM) from ∼15 % to ∼55 %, and 2.5D PSA Pix2Pix showed the highest PSNR (28.074) and SSIM (0.921). Lesion detectability, measured with region of interest (ROI) PSNR, SSIM, normalized contrast recovery coefficient, and contrast-to-noise ratio, was generally improved for all approaches, and 2.5D PSA Pix2Pix trained with geometric-factors-corrected input images achieved the highest ROI PSNR (28.920) and SSIM (0.973).Significance.This study demonstrates the potential of an image-based end-to-end self-normalization framework using cGANs for improving PET image quality and lesion detectability without the need for separate normalization scans.

Optical Simulation and Experimental Assessment with Time-Walk Correction of TOF-PET Detectors with Multi-Ended Readouts

As a commonly used solution, the multi-ended readout can measure the depth-of-interaction (DOI) for positron emission tomography (PET) detectors. In the present study, the effects of the multi-ended readout design were investigated using the leading-edge discriminator (LED) triggers on the timing performance of time-of-flight (TOF) PET detectors. At the very first, the photon transmission model of the four detectors, namely, single-ended readout, dual-ended readout, side dual-ended readout, and triple-ended readout, was established in Tracepro. The optical simulation revealed that the light output of the multi-ended readout was higher. Meanwhile, the readout circuit could be triggered earlier. Especially, in the triple-ended readout, the light output at 0.5 ns was observed to be nearly twice that of the single-ended readout after the first scintillating photon was generated. Subsequently, a reference detector was applied to test the multi-ended readout detectors that were constructed from a 6 × 6 × 25 mm³ LYSO crystal. Each module is composed of a crystal coupled with multiple SiPMs. Accordingly, its timing performance was improved by approximately 10% after the compensation of fourth-order polynomial fitting. Finally, the compensated full-width-at-half-maximum (FWHM) coincidence timing resolutions (CTR) of the dual-ended readout, side dual-ended readout, and triple-ended readout were 216.9 ps, 231.0 ps, and 203.6 ps, respectively.

Application-specific nuclear medicalin vivoimaging devices

Nuclear medical imaging devices, such as those enabling photon emission imaging (gamma camera, single photon emission computed tomography, or positron emission imaging), that are typically used in today's clinics are optimized for assessing large portions of the human body, and are classified as whole-body imaging systems. These systems have known limitations for organ imaging, therefore application-specific devices have been designed, constructed and evaluated. These devices, given their compact nature and superior technical characteristics, such as their higher detection sensitivity and spatial resolution for organ imaging compared to whole-body imaging systems, have shown promise for niche applications. Several of these devices have further been integrated with complementary anatomical imaging devices. The objectives of this review article are to (1) provide an overview of such application-specific nuclear imaging devices that were developed over the past two decades (in the twenty-first century), with emphasis on brain, cardiac, breast, and prostate imaging; and (2) discuss the rationale, advantages and challenges associated with the translation of these devices for routine clinical imaging. Finally, a perspective on the future prospects for application-specific devices is provided, which is that sustained effort is required both to overcome design limitations which impact their utility (where these exist) and to collect the data required to define their clinical value.

Intercrystal scatter studies for a 1 mm3 resolution clinical PET system prototype

Positron emission tomography (PET) systems designed with multiplexed readout do not usually have the capability to resolve individual intercrystal scatter (ICS) interactions, leading to interaction mispositioning that degrades spatial resolution and contrast. A 3D position sensitive scintillation detector capable of individual ICS readout has been designed and incorporated into a 1 mm³ resolution clinical PET system used for locoregional imaging. Incorporating ICS events increases photon sensitivity by 51.5% compared to using only photoelectric events. A Compton scatter angle error minimization algorithm is used to estimate the first ICS interaction location for accurate line-of-response pairing of coincident photons. An optimal scatter angle error threshold of 15 degrees is used to discard ICS events with a high mismatch between energy-derived and position-derived intercrystal scatter angles. Finally, positioning rather than rejecting ICS events boosts peak contrast to noise ratio by 8.1%, and allows for an equivalent dose reduction of 12% while maintaining equivalent image quality.