Feature Extraction of X-ray Fracture Image and Fracture Classification

摘要

X-ray images are the essential aiding means all along in clinical diagnosis of fracture. So the processing and analysis of X-ray fracture images is particularly important. Extracting the features of X-ray images is a very important process in classifying fracture images according to the principle of AO classification of fractures. A proposed algorithm is used in this paper. First, use marker-controlled watershed transform based on gradient and homotopy modification to segment X-ray fracture images. Then the features consisted of region number, region area, region centroid and protuberant polygon of fracture image are extracted by marker processing and regionprops function. Next we use Hough transform to detect and extract lines in the protuberant polygon of X-ray fracture image. The lines are consisted of fracture line and parallel lines of centerline. Through the parallel lines of centerline, we obtain centerline over centroid and perpendicular line of centerline over centroid. Finally compute the angle between fracture line and perpendicular line of centerline. This angle can be used to classify femur backbone fracture.