On an approach to object recognition in X-ray medical images and interactive diagnostics process

The research behind this paper is dedicated to building a tool for batch-procesing X-ray medical images in medical establishments. This paper proposes the approach to object detection and edge finding suitable for X-ray medical images and producing object boundaries in vector representation as polygons. The approach consists of detection stage and edge refinement stage and exploits a class-specific trait of X-ray medical images. The detection is performed by minimization the distance function between contours. Edge refinement uses active contours approach with the new gradient map filtering procedure which improves the resulting boundary quality. The proposed method does not require a training set of images and a-priori data is represented by initial template which is obtained manually. The visual results and performance of this approach were quite satisfactory. Representation of the target boundary as a polyline helps perform distance measurements between specific points. Possible applications of this approach for interactive medical diagnosis process are discussed.