UNSUPERVISED MEASUREMENT OF BRAIN-TUMOR VOLUME ON MR-IMAGES

We examined unsupervised methods of segmentation of MR images of the b rain for measuring tumor volume in response to treatment. Two clusteri ng methods were used: fuzzy c-means and a nonfuzzy clustering algorith m, Results were compared with volume segmentations by two supervised m ethods, k-nearest neighbors and region growing, and all results were c ompared with manual labelings, Results of individual segmentations are presented as well as comparisons on the application of the different methods with 10 data sets of patients with brain tumors. Unsupervised segmentation is preferred for measuring tumor volumes in response to t reatment, as it eliminates operator dependency and may be adequate for delineation of the target volume in radiation therapy. Some obstacles need to be overcome, in particular regarding the detection of anatomi cally relevant tissue classes. This study shows that these improvement s are possible.