Realization of a System Fuzzy-Possibiliste for the Detection of the Cerebral Tumors

This article concerns the problem of the segmentation of the pictures cerebral (imagery by magnetic resonance) MRI. It is precisely about making use of complementarities between several automatic classifiers (different methods or operators) and to increase thus the hardiness of the segmentation process. The combination is done in the objective to exploit to best the complementarities of the classifiers and to extract an useful and applicable information for the segmentation.

This approach uses the FCM algorithm of which (Fuzzy C-Means) the sum of the degrees of adherence of an individual given to all possible classes is equal to 1, and the algorithm PCM possibiliste that (Possibilistic C-means) consists in looking for partitions based on the idea of typicalité. In order to make the algorithm more robust facing imprecisions and to the ambiguous data that can influence considerably on the centers of classes, these last will be followed of the algorithm of growth of regions.