Evaluation of Resolution Properties of Median Priors in List-Mode Reconstruction Methods for PET

In list-mode acquisition, data is collected as a list of events with different attributes recorded for each event. List-mode reconstruction methods are used in this case. In this paper, we have contributed towards an assessment of resolution properties of list-mode image reconstruction methods, through an investigation of impulse responses, obtained using Local Impulse Response (LIR) function in Penalized-Likelihood (PL-EM) reconstruction methods with our proposed median root based (MPs) priors and have compared our results with standard quadratic priors (QPs). We have shown that better resolution can be recovered with these priors, whereas standard priors still produce a kind of over-smoothing. We have also presented results for Combine Priors (CPs), a hybrid of standard Quadratic Priors (QPs) and Median Based Priors (MPs), which can be used for better user controlled resolution characteristics, and noise resolution trade-off. Results show that our proposed priors are less sensitive to the parameter value which is a great advantage in PL-EM methods.