Improvement of CNR in low-count PET scans using tissue-scattered data as initial estimate in non-TOF MLEM reconstruction

Abstract

We have proposed a method to improve contrast-to-noise ratios (CNRs) of lesions in low-count PET scans using the rejected single scattered (inside tissue) data. Indeed, with the advent of high-quality detectors regarding their energy and time resolution, and list mode data acquisition, adding single scattered events by estimating their original line of response has become more and more feasible. The low-count PET scan data were simulated using GATE software considering Jaszczak type object, and images were reconstructed using STIR. A deconvolution based method to create images from rejected single scatter data has been proposed. These scatter images were, in turn, used as an initial estimate in non-TOF MLEM reconstruction. Three types of human-sized PET scanners—ideal, state-of-art, and future generation—with different timing and energy resolutions were considered. We found a significant improvement in trade-off between the contrast recovery coefficient (CRC) and coefficient of variation (CoV) for the non-TOF MLEM reconstructed images while comparing with other less computational ways of creating initial estimates. CNR improvements were found for all lesions. The present work demonstrated the beneficial use of tissue-scattered TOF PET events which constitutes a high percentage of data in PET imaging, and rejected in general. Our approach has a potential and scope for further study.