Automatic segmentation of the prostate into peripheral and transition zones is paramount in developing computer aided diagnosis systems for prostate cancer diagnosis, as cancer behaves differently in each zone. We propose a multi-atlas based segmentation (MAS) algorithm characterized by a new atlas selection strategy: the performance of a subset of atlases is evaluated considering how well that subset segments the image that is most similar to the target image. Comparison of our method with three other MAS algorithms on fifty-five patients shows a statistically significant improvement on the segmentation accuracy.
This abstract and the presentation materials are available to members only; a login is required.