Filling of incomplete sinograms from sparse PET detector configurations using a residual U‐Net

Abstract Background Long axial field‐of‐view PET scanners are becoming increasingly available worldwide for clinical and research nuclear medicine examinations, providing an increased field‐of‐view and sensitivity compared to traditional PET scanners. However, a significant cost is associated with manufacturing the densely packed photodetectors required for the extended‐coverage systems. Despite improved performance allowing ultralow dose or ultrafast scans, the financial barrier remains, limiting clinical utilisation. Purpose To mitigate the cost limitations, alternative sparse system configurations with strategically placed inter‐detector gaps have been proposed, allowing an extended field‐of‐view PET design with detector costs similar to a standard PET system, albeit at the expense of image quality. Methods To address the challenges posed by sparse detector configurations, particularly the heavy undersampling of PET measurements, we propose a deep sinogram restoration network to fill in the missing sinogram data. The network, a modified Residual U‐Net, is trained end‐to‐end using standard clinical PET scans from a GE Signa PET/MR. The training involves simulating the removal of 50% of the detectors in chessboard patterns of varying sizes, leading to incomplete sinograms with significant count losses (thus retaining only 25% of all lines of response). Results The model successfully recovers missing counts in incomplete sinograms, with a mean absolute error consistently below two events per pixel for typical injected radioactivity, outperforming 2D interpolation of incomplete sinograms based on mean absolute error and structural similarity in both sinogram and reconstructed image domain. Notably, the predicted sinograms exhibit a smoothing effect, leading to reconstructed images lacking sharpness in finer details. Despite these limitations, the model demonstrates a substantial capacity for compensating for the undersampling caused by the sparse detector configuration. Conclusions This proof‐of‐concept study suggests that sparse detector configurations, combined with deep learning techniques, offer a viable alternative to conventional PET scanner designs. This approach supports the development of cost‐effective, total body PET scanners, allowing a significant step forward in medical imaging technology.