Active shape model based segmentation of bone structures in hip radiographs

This paper presents a novel method for segmentation of bone structures in anterior-posterior (AP) radiographs based on active shape models. A priori global knowledge of the geometric structure of each hip is captured by a statistical deformable template integrating a set of admissible deformations. Then, to represent the image structure expected at each shape point, two statistical models are built: the first model contains knowledge about the edges extracted from the radiograph. This information is summarized in shape context histograms. The second model describes the local image structure around each model point. After gathering these two types of data over the training set, an independent component analysis allows us to derive two data representations for each contour part of the shape. The search is performed over a median pyramid using the shape context model. The obtained segmentation is refined at the original radiograph resolution using the more local model. A leave-one-out test was used to evaluate the performance of the proposed method and to compare it with other conventional methods. The results demonstrate that the method is very robust and precise, and that it can be useful in the context of preoperative planning of hip surgery.