Deformable model for serial ultrasound images segmentation: application to computer assisted hip athroplasty

In this paper, we present a segmentation method for ultrasound images acquired serially for intraoperative data acquisition in computer assisted total hip athroplasty application. To extract the bone surface from ultrasound image, we propose a method based on model deformable (snake) with integration of local intensity variable around the evolved contour. We also proposed an adaptive additional force calculated from image gradient vector flow in narrow band around the contour. In this serial image segmentation, we utilized the segmentation result of previous image as an input of next image segmentation. Finally, we perform a post treatment of the final contour to select exclusively the real points on the bone surface. This point selection is based on intensity criteria, so that we only keep the points having strong intensity amongst the points on the final contour. Validated on a fresh cadaver and a health subject, we found that the precision of this method is applicable for the input of registration method in computer assisted orthopedic surgery. Despite of its possibility to be applied in off line manner, the needed time calculation is acceptable for intraoperative application.