Automatic landmarking of cephalograms

This paper presents an algorithm for automatically locating certain characteristic anatomical points called landmarks on cephalograms (skull X-rays). These landmarks are used by orthodontists in diagnosis and treatment planning. The algorithm uses digital image processing and feature recognition techniques to locate the landmarks. A resolution pyramid of the digitized cephalogram is first created. The algorithm works on the smaller, lower resolution images to locate features of interest and moves to the bigger, higher resolution images if greater location accuracy is required. Prefiltering using the median filter, contrast enhancement using histogram equalization, and edge enhancement using different gradient operators are performed on the images. The algorithm uses anatomical knowledge of the human facial structure to search for features containing the landmarks. The accuracy of the algorithm in locating the landmarks is compared with values obtained from human experts. At present the algorithm attempts to locate 10 landmarks of 27 needed for a complete analysis. All 10 landmarks have been successfully located on five cephalograms of varying quality.