Fast automated segmentation of femoral heads in fluoroscopic X-ray images

In fluoroscopic tracking for fractured femur bone fixation, a precise identification of the femoral head provides valuable guidance for positioning the implant instruments such as nails and screws. Noise, occlusions and weak edges challenge the task of automatically segmenting the femoral head. In this paper, a fast and fully automated method to precisely delineate the femoral head in fluoroscopic X-ray images is presented. The proposed method comprises two stages: First, a candidate circle detection stage using a set of curved Gabor filters and a Gabor-based Hough transform is applied to estimate a few candidate circles approximating the femoral head. Second, a fine circle determination stage extracts the target circle from the candidates by analyzing the anatomical features of the femoral head and its spatial relation to the acetabulum. The validity and robustness of the method were tested on a set of 1184 fluoroscopic images from different vendors.