A Public dynamic PET brain database for lesion detection and quantification

Abstract

Purpose: Dynamic PET imaging gives insight into various metabolic processes. Considerable work has been carried out with the aim of facilitate visual inspection and interpretation of the studies. However, comparison between different algorithms and methods is an arduous task. Source codes are rarely distributed, and if available, there is no single, unified public database to test them on. Therefore, turns out mandatory for any new research team, to invest considerable time and effort in acquiring or creating their own database before diving headlong into research. This paper presents a dynamic PET brain database. Methods: The ASIM and STIR software packages are used to simulate and reconstruct phantoms based on a pharmacokinetic model activity. Results: There are 18 phantoms in total, each with 5 implanted regions, and with varying lesion activities. Conclusions: Phantoms are made available to the community. We hope this makes intercomparisons between algorithms easier, as well as accelerate the research process in the area