A method for automatic extraction of striatal structures for dose-finding studies in PET

Abstract

The extraction of regions of interest (ROIs) from positron emission tomography (PET) neuroreceptor studies enables the computation of regional binding potential (BP) value of the radioligand in a specific brain structure. When a subject is scanned during baseline (no medication) and intervention (medication) we can compute the receptor occupancy value (a function of BP values). This is important in dose-finding in drug development. In the present study, we have extended an earlier presented automatic method to extract the left and right caudate and putamen for receptor occupancy studies. This is a difficult task even with the baseline images. The task becomes even harder with the intervention images owing to considerably reduced uptake of the tracer. We compared the receptor occupancy values obtained by the introduced automatic method (pairwise dual surface minimization - inner surface) and the traditional, manual ROI drawing and found them to be close to each other. The reproducibility, reliability and validity investigations of BP values showed that the introduced method performed at least as good as manual segmentation. These results support the use of automatic segmentation in receptor occupancy type drug development studies.