Use of neural network algorithms for the automated arrangement of cephalometric markers on lateral cefalograms

The lateral cephalogram analysis is an important method of a research in orthodontics and maxillofacial surgery that allows to obtain necessary diagnostic information on a structure of brain and facial skull parts for treatment planning. There are numerous techniques of the analysis based on processing of cephalometric points (markers) on the lateral cephalogram. Such approaches take a considerable time for the doctor to arrange cephalometric points. Modern digital technologies with use of artificial intelligence allow to improve this method of a research and to significantly simplify doctor’s work. Purpose. In this paper, we propose a neural network and training strategy that enables to place cephalometric points on the lateral cephalogram with high precision. Materials and methods. The research used 80 lateral cephalograms of the head. Results. The developed method handles the cephalograms regardless of the source of the image, while error percentage below 2% of the size of the images. The offered approach demands 2-3 times less time than a traditional "manual" method of arrangement of cephalometric points, depending on quantity of points and complexity of the cephalometric analysis. © 2018 Russian Electronic Journal of Radiology. All Rights Reserved.