Performance evaluation of Compton probe and SPECT in lesion detection

In this paper, we evaluate the performance of the previously proposed Compton imaging probe in comparison to SPECT for lesion detection using a non-prewhitening matched filter observer model. Specifically, multiple realizations of tumor present and tumor absent data were generated for both systems with Monte Carlo simulations for ¹¹¹In ProstaScint distribution. The penalized list-mode ML-EM and FBP methods were used for image reconstruction for the probe and SPECT respectively. The NPWMF observer was then applied to the reconstructed images and AUC was calculated to measure quantitatively the performance of two systems. Using simulation that represents a realistic prostate imaging scenario, we demonstrate a significant improvement of the Compton probe over SPECT for lesion detection in terms of both visual observation and quantitative measure of lesion detectability. For the tasks of detecting 8 mm and 5 mm tumors, the Compton probe demonstrates an SNR improvement of 3-3.5× over SPECT.