An Approach for Efficient Detection of Cephalometric Landmarks

In this paper, a method is developed for the automated identification of cephalometric landmarks in orthodontics. The process of soft tissue edge detection is divided into two steps: detecting the sub-images that contained the required landmarks using combination of the Histograms of Oriented Gradients (HOG) descriptor with the Support Vector Machine (SVM), then utilizing Thresholding and Mathematical Morphological (TMM) algorithm to trace soft tissue profile. In addition, the mandible's edge is detected by the Active contours without edges (Chan-Vese method). Finally, the landmarks of soft tissue profile and the mandible's edge are pinned based on analyzing the contour plot of these lines. The simulation results have high accuracy.