An improved interactive segmentation method for extracting the edge features of femur digital radiographs

摘要

By comparing the advantages and disadvantages of two interactive image segmentation algorithms: level set and live wire, we propose a improved multi-step realization method of interactive image segmentation, which could help the operators to extract the important anatomical structure features from the femur digital radiographs (DR) images more accurately. Firstly, a preprocessing step including median filtering and image enhancement was made to eliminate the noise during the DR imaging; Secondly, with the advantages of level set such as simple operation and fast convergence rate, the coarse outline contour extraction was realized; Finally, with the advantages of live-wire such as repeated local operation and high precision, the fine contour extraction of special anatomic areas, the profile of fracture edge and the overlapping area was realized. So, all the interesting anatomical structure features of DR images were obtained. In this paper, our method was applied to the complete femur DR images and artificial fracture femur DR images. The segmentation result shows that our method has a good performance in accuracy and efficiency.