As the result of developments in computer technology in recent years, the calculation, simulation and analysis of the stress environment for the artificial hip prosthesis and the femur have been made possible. In order to perform an exact calculation, three-dimensional modeling with the hip prosthesis and femur is required. The modeling of a femur is done by reconstructing the femoral outline data from CT images using three-dimensional CAD. The outline extraction work is performed by the experience and subjectivity of the operator, and the reliability of the data and working efficiency is low. In this paper, algorithms that extract the femur outline from CT images automatically are proposed. A Snake-based outline extraction algorithm is used. Although this method is capable of extracting the cortical bone outline of femoral diaphysis with sufficient accuracy, there was high incorrect detection at the trochanter. The shape of the trochanter is complicated, and it is difficult to process using a simple algorithm. Therefore, we proposed a method of visually expressing the possibility that a certain pixel in an image is high or low. This method is capable of detecting portions of images with high differentiations in value (contrast changes rapidly). It can also detect outlines at the diaphysis and trochanter. The problems of this technique are that the pixels judged as possibly being a part of the outline appear in the portion that is not a true outline, and that the pixels judged as possibly being on the outline may not be continuous. When each method was used independently, neither was capable of performing a positive outline extraction. However, an outline sampling sufficiently high certainty was achieved when using the two methods simultaneously.
