A method for the knowledge based image segmentation using the form of models

摘要

A method for the segmentation of an object of interest in an image of a patient with such an object is described. Each of a plurality of training forms is deformed, in order to superimpose a reference shape, wherein a parameter θ_i a measure for the thickness of the deformation is, which is necessary in order to effect the superposition. A vector of the parameters θ_i for each of the training forms by minimizing a cost function together with an estimate of the degree of uncertainty for each of the resulting vectors of parameters θ_i is obtained, whereby such an uncertainty in the form of a covariance matrix σ_i is quantified. A statistical model the form of a is shown, with the sum of cores, with a means θ_i and covariance σ_i generates. The desired object of interest in the image of the patient is identified, in that the reference shape on the image is arranged and deformed, in order to superimpose the image obtained, wherein a parameter θ is a measure for the thickness of the deformation is, which is necessary in order to effect the superposition. An uncertainty in the form of a covariance matrix σ is quantified and an energy function e = e_shape + E_image is calculated, the probability of the current form in the statistical model and the fitting into the image e_image to obtain.