A process for the fully automatic segmentation of an organ, especially the renal parenchyma from volume data sets of medical imaging

The present invention relates to a method for segmentation of an organ, especially the renal parenchyma from volume dataset medical imaging. For this, a probability data set is generated on the basis of segmentation results of training dataset where each voxel of the volume data set is associated with a probability based on its Intensitatswerte to belong to the institution. Using a thresholding is a binary data set is generated, in which the body region is eroded gradually. In a splitting of the body portion into a plurality of sub-areas is in each case only one of the partial areas which includes with high probability art the organ selected for further erosion. This portion is recognized on the basis of geometric features of the organ, known from the training data sets. then the probability values ​​of voxels of the cleaved in the binary record partial areas are reduced in the probability data set, so that a correspondingly corrected probability data set is obtained. the organ from the probability data set is then segmented on the basis of this corrected probability data set. With the proposed method, the segmentation of organs from native MR data sets is made possible with high reliability.