FUZZY FUSION SYSTEM FOR BRAIN MRI IMAGE SEGMENTATION

摘要

A fuzzy data fusion scheme is proposed in this paper to automatically segment different tumor areas of human brain from multispectral Magnetic Resonance Imaging (MRI) images. The information obtained from the segmentation allows then to help doctors to make the diagnostic. The proposed framework consists of four stages: a registration of the multispectral images, a fuzzification using a priori knowledge based fuzzy models, a fuzzy fusion and a post treatment by means of a fuzzy region growing. The comparison between results of the proposed method and hand-labels of a radiology expert shows that this automatic scheme is efficient.