In this paper, we present the various stages for lesion recognition in brain images. We firstly apply a filtering based on geodesic reconstruction operator for increasing the quality of image. After, we use an unsupervised segmentation genetic algorithm for detecting the abnormal zones with respect of theirs morphological characteristics because they define the nature of illness (cyst, tumour, malignant, benign, …). The obtained segmented images are analyzed for computing the characteristics of illness which are necessary for the recognition stage for deducing a decision about the type of illness. So, we give also the various algorithms used for computing the morphological characteristics of lesions (size, shape, position, texture, …). Since we obtain a decision about the malignity or benignity of the lesion and a quantitative information for helping the doctors to locate the sick part.
