Segmentation of brain structures using PET-CT images

Abstract

The accurate segmentation of PET-only brain images is challenging because of the low spatial resolution and high noise level in PET data. PET/CT has now replaced PET and offers the opportunity to improve segmentation through the high resolution, lower noise CT data. This paper pioneers the research of PET-CT brain image segmentation, which takes advantage of the full information available from the combined scan. In the proposed approach, the contrast stretched CT image is utilized to delineate cerebrospinal fluid (CSF) from other brain tissues. Gray matter is separated from white matter by applying the fuzzy clustering of spatial patterns (FCSP) algorithm to the joint PET-CT image. We compared our approach to a widely used PET segmentation method in the SPM toolbox for simulation and patient data. Our results prove that the incorporation of anatomical information in CT images substantially improves the accuracy of brain structure delineation.