Interactive 3D Segmentation of CT-scans of the Hip Joint, Using Active Learning to Select Delineation Planes

Abstract

This paper evaluates a method that improves segmentation e ciency by intelligently suggesting planes where correction is most valuable. An existing method is extended to work for segmentation of multiple bones simultaneously. This method is evaluated because in clinical practice it is often necessary that scans are segmented very accurate. When automatic segmentation algorithms are not accurate enough, a user needs to provide correction. The method described in this paper uses active learning to alleviate the user from picking a plane to provide correction on. We describe a series of experiments that compare the accuracy of the segmentation using the implemented method with a method that corrects random planes and with an upper limit. The implemented method performs better than random plane suggestion, however the accuracy does not reach the upper limit. Out of the two bones that were evaluated the more complex shaped bone takes longer to reach a high accuracy than the less complex bone.