Early Detection of Interproximal Caries using YOLOv8 Convolutional Neural Network on Bitewing Radiographs

Early childhood caries is a common oral health issue in both developed and developing countries. Identifying caries lesions early can help in implementing minimally invasive treatment plans, preserving the remaining tooth structure. Interproximal lesions are typically found between the adjacent surfaces of two teeth. The first step is visual inspection, followed by examination using dental probes to detect the dental caries. Usually, radiographic examination is done for early detection of dental caries. This process is time-consuming. In recent years, AI has been used to detect early caries from X-ray images. In dentistry, proximal caries is difficult to identify from periapical and panoramic radiographs, so Bitewing Radiographs (BWR) are used to detect the interproximal caries. Since the X-ray source is parallel to the sensor, Bitewing Radiographs has least distortion and help in the detection of early proximal caries. Several studies have shown that Bitewing Radiographs have high sensitivity in detecting interproximal caries occurring in existing restorations compared to visual clinical detection. The main objective of this study is to detect dental caries and classify them into various classes enamel, dentin and restorations. In this study, a dataset was collected from Mendeley data and Kaggle. The caries classification utilized a Deep Convolutional Neural Network. A YOLOv8 algorithm was developed based on the severity of the lesions observed in bitewing digital radiographs, enabling convenient and reliable detection of these lesions. The performance of these classifications is evaluated. This study aimed to evaluate the performance of the deep neural network for detecting interproximal caries lesions on Bitewing Radiographs (BWR). Accurate detection and diagnosis of dental caries are crucial for reducing the cost of oral health management and improving the likelihood of preserving natural teeth, leading to better patient outcomes.