Segmentation of images IRM by Anatomical Atlas

This paper proposes a method for segmentation of brain MRI medical images based on the type of knowledge provided preconceptions as Anatomical Atlas. Our approach proceeds by registration of an atlas on a patient image picture of seeking repeatedly to maximize the correspondence between the features of Atlas model and segmented image. The procedure of our approach is essentially divided into two steps: a step of affine registration between the patient and the image atlas anatomiquel, thereby determining the geometric transformation leading to better MRI/Atlas alignment, and a step of resetting at levels nonlinear constrained by processing gray calculated in the first phase. The segmentation of tissues or structures is performed by a hierarchical procedure, where the contours spread atlas allows the segmentation of organs or structures in the patient image.