Age Group Classification From Dental Panoramic Radiographs Using Deep Learning Techniques

Abstract

Deep learning approach has been utilized in dental age determination from dental radiographs. In this study, several deep learning models were employed to predict age from a dataset of dental panoramic radiographs (OPGs). The dataset, collected from the Oral and Maxillofacial Radiology Clinics within the Faculty of Dentistry, Chiang Mai University, comprised 785 subjects with each subject having one OPG image ranging from 20 to 85 years old. We divided our dataset into four experimental cases: 2, 4, 6, and 12 classes. We utilized four deep learning models - SqueezeNet, VGG19, ResNet152, and EfficientNet - for training. In the experimental results, the efficiency of each model was compared in terms of accuracy. We found that VGG19 achieved the highest average degrees of accuracy at 92.95%, 80%, and 75.66% in the two-class, four-class, and six-class cases, respectively, while EfficientNet attained the highest degree of accuracy at 69.11% in the twelve-class case.