A Convolutional Neural Network for Dental Panoramic Radiograph Classification

Abstract

Radiographs are X-rays of the craniofacial area, used for orthodontic diagnosis and treatment planning. Analysis of radiographs is a manual process. Public medical centers in developing countries such as South Africa experience a bottleneck in the analysis of these radiographs, due to excessive numbers of patients and severe shortages in orthodontic radiologists that serve at these public medical centers. Access to dental diagnostics is therefore becoming an ever-increasing problem in rural communities. This paper reports on the first phase of a framework to automate the analysis of panoramic radiographs, which are X-rays of the frontal croniofacial area. This first phase automates the process to predict whether a captured panoramic radiograph is workable or not, i.e. whether automated analysis of the X-ray can proceed. A is trained on a large set of panoramic radiographs, and results show that the prediction accuracy of this is very good in identifying workable radiographs.