Automatic Localization of Cephalometric Landmarks

Cephalometric analysis has an important role in diagnosis and treatment of dental disharmonies. In this paper, we propose an efficient, fast, and automatic method to localize cephalometric landmarks on digitized x-ray images. The proposed algorithm uses the information of the marked landmarks on a reference normal cephalometry image as the prior knowledge. In the first step of the proposed method, the image is automatically divided into several regions and three main control points are located on it. These are then matched to their corresponding points on the reference image to form an affine transform matrix that describes how other points on the reference image should be mapped to the target image. Each mapped point is used as an initial estimation of the landmark location on the target image. In the second step, the accurate location of each landmark is detected within a surrounding window positioned around the estimated initial point of the landmark using edge-detection, image histogram, and curve fitting schemes. The main idea in this approach is to use the similarities available in human anatomy to estimate the initial location of each landmark and then addresses the differences by applying the accurate landmark definition in a small region of interest (ROT). The experimental results show the efficiency of the proposed algorithm.