In ASM [30], edge points, at which the voxel values rapidly change, are detected, and the model surface is deformed so that points on the surface are located at the edges. Edges, ridges, or other local image features specific to the points on the target boundaries must be detected in given images to enable accurate model-to-image registration. Some details of fundamental operations for the image feature detection will be described in Sect. 2.2.1.

Statistical Inference

Given a new image, one can register the SSM to the image by inferring the values of the parameters of the SSM: One can deform the model by changing the values of the parameters, and one can register the model by finding appropriate values of the parameters that describe the shape of the target organ in the given image. One of the most fundamental approaches for the inference of the parameter values is to construct an objective function of the parameters in a statistical framework and to compute the optimal parameter values that minimize (or maximize) the objective function.