Segmentation of Dynamic PET Images

The work presented in this section builds on our earlier research in which we applied the proposed clustering algorithm to tissue classification and segmentation of phantom data and a cohort of dynamic oncologic PET studies [94]. The study was motivated by our on-going work on a noninvasive modeling approach for quantification of FDG-PET studies where several ROIs of distinct kinetics are required [95,96]. Manual delineation of ROIs restrain the reproducibility of the proposed modeling technique, and therefore, some other semiautomated and automated methods have been investigated and clustering appears as a promising alternative to automatically segment ROI of distinct kinetics. The results indicated that the kinetic and physiological parameters obtained with cluster analysis are similar to those obtained with manual ROI delineation, as we will see in the later sections.

0 0

Post a comment