Ventricles segmentation and matching for content - based medical image retrieval

In this paper, we propose a new methodology for the segmentation and retrieval system for Computed Tomography (CT) brain images. For the segmentation part, two segmentation techniques are considered which are modified FCM with population-diameter independent (PDI) and expectation-maximization (EM) segmentation. However, only one of the techniques is selected based on the average intensity in order to obtain the more proper results. The ultimate goal of the segmentation is to acquire the ventricles. For the retrieval part, features are extracted from the ventricles and images are retrieved based on the similarities of the ventricles. From the obtained experimental results, the proposed methodology is feasible and attains satisfactory results.