A process for the segmentation of objects in a three-dimensional image data set with the help of a deformable model

摘要

Method for segmenting a target structure of a body made of a three-dimensional image data set, wherein an initial multilayer deformable network model (mldm) to the boundaries (boi) is adapted to the target structure:• the initial multilayer deformable network model at least two layers(L1, l2) and each layer has a network structure of interconnected network points (m1 - 1, m1 - 2, m1 - 3, m2 - 1, m2 - 2, m2 - 3), each network point of a layer with at least one other network point of the same layer by layer compound (i1 - 1, i1 - 2, i1 - 3, i2 - 1, i2 - 2, i2 - 3) is connected,• a first layer (l1) forms an outer surface of the model,• at least one inner layer (l2) is within the outer surface of the model is,• each of the network points of the first layer (m1 - 1, m1 - 2, m1 - 3) with a corresponding network point of an adjacent inner layer (m2 - 1, m2 - 2, m2 - 3) by means of a low connection (d12 - 1, d12 - 2, d12 - 3) is connected,• in the case of each further inner layer (l3) of each network point of each inner layer (m2 - 1, m2 - 2, m2 - 3) with a corresponding network point of each adjacent inner layer (m3 - 1, m3 - 2, m3 - 3) by means of a low connection (d23 - 1, d23 - 2, d23 - 3) is connected, andthe initial multilayer deformable network model to the boundaries (boi) of a target structure is adapted by means of at least the following steps:a) receiving a three-dimensional image data set of a target structure of a body with a pick-up device,b) identifying boundary points (a) of the target structure, which potentially lie on the limit of the target structure, by means of– Evaluation of the three-dimensional data set with an evaluation means of an image processing system, wherein gray value profiles of the three-dimensional data set along of the normal vectors of the first layer be sampled, wherein the starting point to the scanning in each case the position of a network point of the first layer ..