Rapid assessment of parametric PET images with computer observers

The theoretical formulae based only on computable first and second-order reconstruction statistics have recently been proposed in literature as a relatively rapid method to evaluate relevant figures of merit. In this paper, we extend the theoretical assessment approach from static to dynamic PET images. We consider a detection task of a SEK/BKE and variable signal location. The extended theoretical tools can accommodate different forms of spatial-temporal processing as well as model observer. The quality of single parametric image resulted from two forms of popular temporal processing is assessed and compared. Our theoretical results reveal that the FOM of applying computer observers on a single static image is inferior to that of on the parametric image obtained by Patlak analysis of multiple frame dynamic PET under certain condition.