Fuzzy Clustering Methods for the Segmentation of Multimodal Medical Images
Francesco Masulli, Andrea Beatriz Schenone, Anna Maria Massone · Studies in fuzziness and soft computing · 2000
Multimodal medical imaging (MMI) volumes can be derived by spatial correlating intensity distributions from a number of different diagnostic volumes with complementary information. An unsupervised approach to MMI volumes segmentation is recommended by many authors. Due to complexity of the data structure, this kind of segmentation is a very challenging task, whose main step is clustering in a multidimensional feature space. The partial volume effect originated by the relatively low resolution of sensors produces borders not strictly defined between tissues. Therefore memberships of voxels in boundary regions are intrinsically fuzzy and computer assisted unsupervised fuzzy clustering methods turn out to be particularly suited to handle the segmentation problem. In this paper a number of clustering methods (HCM, FCM, MEP-FC, PNFCM) have been applied to this task and results have been compared.