An automatic rib segmentation method on X-ray radiographs

Abstract

In this paper, an automatic rib recognition method based on image processing and data mining is presented. Firstly, multiple template matching and graph based methods are used to detect rib center line; then, the support vector machine is used to build a rib relative position model and identify the error recognition results; finally, decision trees are employed to refine the center line recognition result. The JSRT database is employed to test our method. The result of rib recognition is over 92% for sensitivity and 98% for specificity.