Atlas Based Segmentation of the prostate in MR images (original) (raw)

The large variability and contrast differences between prostates make its segmentation difficult using traditional segmentation methods. In this paper we present an automatic method based on nonrigid registration of a set of prelabelled MR altas images. The algorithm consists of three stages. Firstly, the target image is nonrigidly registered with each atlas image, using mutual information as the similarity measure. Subsequently, the best registered image is selected comparing normalised mutual information measures after registration. Finally, the segmentation is obtained by deforming the corresponding atlas labelled image selected before using the transformation determined on the registration stage. The method is evaluated on 15 images using the Dice similarity coeficient and 95% Hausdorff distance, obtaining medians of 0.79 and 7.11 mm respectively.