Automatic bone boundary detection in hand radiographs by using modified level set method and diffusion filter

Abstract

RA (Rheumatoid Arthritis) is a chronic inflammatory joint disease characterized by a distinctive pattern of bone and joint destruction. To give an RA diagnosis, a hand radiograph is taken and analyzed. Hand bone radiograph analysis starts with the detection of the boundary of bones. It is, however, an extremely exhausting and time consuming task for radiologists, not only because of the complexity, but also because of the precision required for a correct diagnosis. Automatic bone boundary detection is thus required. The Level Set Method has been widely used in boundary detection. However, the convergence and stability of the level set are strongly affected by the speed function and the parameters of the level set, which often leads to either a complete breakdown or a premature termination of the curve evolution process, resulting in unsatisfactory results. In this paper, we propose a modified speed function of the level set for bone boundary detection in hand radiographs. And in order to preserve the boundary of an image and to reduce noise, we further apply diffusion filter to substitute Gaussian Filter in the standard Level Set Method. Evaluating the experiments using a particular set of hand bones radiographs, the proposed method worked well for almost all of the images that we used.