Improving Content Based Retrieval of Magnetic Resonance Images by Applying Graph Based Segmentation

Medical content based retrieval systems are continuously researched and improved. Image segmentation techniques offer a possibility for improvements in the retrieval process. According to this, the aim of the paper is to examine a graph based segmentation technique when it is applied to content based retrieval systems for magnetic resonance images. For this purpose, an evaluation of seven descriptors in both cases: applied to the whole images and on the segmented images is performed. The examination was performed using dataset of hierarchically organized magnetic resonance images. According to the obtained results, we conclude that from the explored descriptors, when Edge histogram descriptor, Region-based shape descriptor and Wavelet transformations are used for feature extraction, the examined segmentation technique leads to improvements in the retrieval effectiveness.