Fast region description method for content based medical image retrieval

A new approach to segmented image representation and description for content-based medical image searching is proposed. Using the method a table of image runs is dynamically constructed. The table is used by a grouping process running on the oversegmented image to reduce small regions. In the framework of the content-based querying the table is also applied to describe regions and their contours (inner borders). The shape descriptors like an area, a perimeter and a compactness are used to indicate basic region properties. Application of the method for CT and MRI images produce good results, which are presented and discussed in this paper.