Validity guided clustering for brain tumor segmentation [treatment planning]
Robert P. Velthuizen · 2002
Clustering of MRI for unsupervised tumor volume segmentation often fails to identify pathological tissues. In this paper, a refinement of fuzzy c-means segmentation is proposed using a validity measure. The method is applied to three patient cases with different tumor types. A significant improvement compared to FCM is achieved for two tumor types, both in the detection of the tumor and the accuracy of the pixel assignments.