Hippocampus segmentation through gradient based reliability maps for local blending of ACM energy terms
This paper presents a novel 3D segmentation framework for structures with spatially varying boundary properties, such as the hippocampus (HC). The proposed method is based on Active Contour Models (ACMs) built on top of the multi-atlas concept. We propose the incorporation of an Adaptive Gradient Distibution on the Boundary map (AGDB) into the ACM framework. AGDB, by being adapted to the evolving contour, constantly redefines, at a voxel level and at each contour evolution, the degree of contribution of the image information and the prior information to the energy minimization. The proposed segmentation scheme was tested for HC segmentation using the publicly available IBSR database.