Simultaneous segmentation and registration for medical image

Although segmentation and registration are usually considered separately in medical image analysis, they can obviously benefit a great deal from each other. In this paper, we propose a novel scheme of simultaneously solving for segmentation and registration. This is achieved by a maximum a posteriori (MAP) model. The key idea is to introduce an additional hidden Markov random vector field into the model. Both rigid and non-rigid registration have been incorporated. We have used a B-spline based free-form deformation for non-rigid registration case. The method has been applied to the segmentation and registration of brain MR images. © Springer-Verlag Berlin Heidelberg 2004.