PETformer network enables ultra-low-dose total-body PET imaging without structural prior

Abstract Positron Emission Tomography (PET) is essential for non-invasive imaging of metabolic processes in healthcare applications. However, the use of radiolabeled tracers exposes patients to ionizing radiation, raising concerns about carcinogenic potential, and warranting efforts to minimize doses without sacrificing diagnostic quality. In this work, we present a novel neural network architecture, PETformer, designed for denoising ultra-low-dose PET images without requiring structural priors such as CT or MRI. The architecture utilizes a U-net backbone, synergistically combining multi-headed transposed attention (MDTA) blocks with kernel-basis attention (KBA) and channel attention (CA) mechanisms for both short- and long-range dependencies and enhanced feature extraction. PETformer is trained and validated on a dataset of 317 patients imaged on a total-body uEXPLORER PET/CT scanner. Quantitative evaluations using structural similarity index measure (SSIM) and liver SNR showed PETformer's significant superiority over other established denoising algorithms across different dose-reduction factors. Its ability to identify and recover intrinsic anatomical details from background noise with dose reductions as low as 2% and its capacity in maintaining high target-to-background ratios while preserving the integrity of uptake values of small lesions enables PET-only fast and accurate disease diagnosis. Furthermore, PETformer exhibits computational efficiency with only 37M trainable parameters, making it well-suited for commercial integration.