X-ray bone fracture segmentation by incorporating global shape model priors into geodesic active contours

Abstract

This paper describes a novel mathematical model of segmenting bone boundary of X-ray image that incorporates prior shape information into a geodesic active contour. The energy function of the model is minimized depending on the image gradient and the shape of the region of interest, so the boundary of the bones in the X-ray image can be captured either by higher magnitude image gradient or by the prior knowledge of the shape. This model improves on the level set method [Sethian, J.A., Level Set Methods and Fast Marching Methods, Cambridge Press, 1999] in which shape information is missing from its energy function. The evolving process of the proposed model stops when the prior shape and evolving curve are perfectly aligned. The prior shape adjusts its rotation, scale, and translation during the curve-evolving process. The algorithm is computation efficient. The model performs segmentation on images with noise background caused by casting material overlaying on the fractured bone in the X-ray image and produces good results. The experimental results of applying the on-the-patient's X-ray images are provided. The existence of the solution to the proposed minimization problem is also discussed.