Automatic brain tumor detection and segmentation for MRI using covariance and geodesic distance

In this paper, we present a new approach that allows the detection and segmentation of brain tumors automatically. The approach is based on covariance and geodesic distance. The detection of central coordinates of abnormal tissues is based on the covariance method. These coordinates are used to segment the brain tumor area using geodesic distance for T1 and T2 weighted magnetic resonance images (MRI). The ultimate objective is to retrieve the attributes of the tumor observed on the image to use them in the step of segmentation and classification. The present methods are tested on images of T1 and T2 weighted MR and have shown a better performance in the analysis of biomedical images.