Method for the fully automatic segmentation of an organ, in particular of the renal parenchyma, of volume data sets of the medical imaging

摘要

A process for the segmentation of an organ, in particular of the renal parenchyma, in volume data sets of the medical imaging, wherein the– on the basis of segmentation results from training records– a region of interest within a recorded volume data set is fixed, of the organ, and with the highest probability contains– a probability data set is generated, in which each voxel of the volume data set based on its intensity values and the segmentation results from the training records a probability is assigned to the organ to include,– in the region of interest to the probability data set a threshold value method is used, by means of which a organ area is separated from the background, in order to obtain a binary data record, in the voxel of the background and voxel outside the region of interest have a different value in the form of a voxel of the organ area,– Performing a stepwise erosion of the organ area in the binary data set to a predeterminable number of steps, in the case of the– at each step, starting from an outer boundary of the organ area a layer of the organ area is removed,– in the case of a splitting of the organ area into a plurality of partial areas in the case of a step in each case only one of the partial areas, of the with the highest probability of the member, is selected for the further erosion or eroding,– wherein the partial region, the with the highest probability of the member contains, in each case on the basis of geometrical features in at least one layer of the binary data set is selected, which are known from the training records,– in the probability data set the probability values of voxels of the partial regions of the substrips in the binary data record be reduced, so that a corrected probability data set is obtained, and– on the basis of the corrected probability data set a segmentation technique is used, in order to the member from the probability data set from the volume data set and thus also be segmented.