Improving the Results of Automatic Landmark Detection in Cephalometry with Extraction of Bony Structures

This paper introduces a new method for automatic landmark detection in cephalometry. In first step, some feature points of bony structures are extracted to model the size, rotation, and translation of skull, we propose two different methods for bony structure discrimination in cephalograms. The first method is using bit slices of a gray level image to create a layered version of the same image and the second method is to make use of a Smallest Univalue Segment Assimilating Nucleus (SUSAN) edge detector and discriminate the pixels with enough thickness as bony structures. Then a Learning Vector Quantization (LVQ) neural network is used to classify images according to their geometrical specifications. Using LVQ for every new image, the possible coordinates of landmarks are estimated. Then a modified ASM is applied to locate the exact location of landmarks. On average the first method can discriminate feature points of bony structures in 78% of cephalograms and the second method can do it in 94% of them. As a final result 24 percent of the 16 landmarks are within 1 mm of correct coordinates, 63 percent within 2 mm, and 95 percent within 5mm, which shows a distinct improvement on other proposed methods.

Keywords:

Cephalometry , Craniofacial Landmarks , Active Shape Model , SUSAN Edge Detector , Learning Vector Quantization