Semantic association mining on spatial patterns in medical images

Abstract

The advancement in the field of image mining owes the credibility to the discovery of significant image patterns from the archive. An important pattern of an image is the spatial displacement of objects in an image, and the term is coined as spatial relationship feature. In this paper, we incorporate the concept of spatial relationship into medical images. The spatial relationship of structures in the medical image is expressed in terms of fuzzy set theory, thus making the scenario closer to the semantics of the images. The fuzzy spatial relationships existing between the structures in the medical images are mined to identify relevant spatial patterns, so that characteristic knowledge generated is highly authentic for the medical domain under consideration. From the spatial patterns generated spatial association rules are deduced that can steer as an aid to medical diagnosis or rather new diagnosis rules. © 2011 Springer-Verlag.