Preliminary Study of Pseudo-PET Image Synthesis of Glucose Metabolism from Early-Phase PET Images of an Uncorrelated Radiotracer

Prabath Hetti Mudiyanselage, Ruwan B. Tennakoon, John Thangarajah, Robert Ware, Jason H. Callahan, Lucy Vivash

Abstract: Positron Emission Tomography (PET) is widely used in modern medicine to identify diseases such as cancer, neurological disorders, and heart conditions to observe biochemical functional processes. Despite its widespread usage, the requirement for different radiotracers to observe separate biological processes imposes resource constraints and patient inconvenience. Recent clinical studies suggest the utility of using a single radiotracer to examine more than one distinct biological process.In this paper, we present a preliminary study of a novel framework for generating pseudo-equilibrium PET images of glucose metabolism from early-phase images of an uncorrelated radiotracer (i.e., a radiotracer targeting a different biological process). The proposed framework further includes a quantitative analysis of the synthesized images for noise and structural similarity, as well as a radiomics analysis. The proposed framework achieved an R-squared score of 0.961 and SSIM of 0.913 when synthesized images were evaluated with the target images. The results show a significant improvement over the baseline of averaged input (R-squared = 0.689, SSIM = 0.801), which was also confirmed by the radiomics analysis.