Organ delineation using factor analysis on the Genisys preclinical PET system

Abstract

The Genisys preclinical PET imaging system suffers from relatively poor spatial resolution and reconstruction artifacts in the sagittal and transaxial planes due to limited angular sampling. This prevents reliable quantification and delineation of organs and tumors in close proximity. The use of factor analysis (FA) is proposed as a method to mitigate this effect by separation of structures into "factor" images. Two studies are performed The first study involved the application of FA on a synthetically created dynamic image in order to create factor curves. These factor curves were then compared to the synthetic curves used to create the synthetic image The second study applied FA to a dynamic image of tumor-bearing mouse The resulting factor images were assessed in order to determine how well the tumor was separated from other structures. The results from the first study displayed a very strong agreement between the synthetic curves and the factor curves. The second study displayed a prominent tumor presence within the two most significant factors. The results show promise for FA, but further research into optimal conditions must be performed