Development of AI-Powered Oral Disease Detection Using YOLO Deep Learning Model

Early detection of dental issues, such as cracks and tooth decay, remains a challenge in many dental settings. Artificial intelligence (AI) offers promising solutions for automating and assisting dental professionals in early diagnosis of such conditions. This paper explores the development of a novel AI-powered solution using the YOLOv8 model for the detection of dental caries, cracks, and healthy teeth. By integrating an intraoral camera with a transilluminator, the proposed system provides a cost-effective, real-time screening tool that is accessible for everyday use. Two datasets were used for training the YOLOv8 extra-large models. Dataset 1 is comprised of high-resolution images captured in clinical settings using the DigitalDoc Model Iris Intraoral Camera, a device designed for routine dental checkups. Dataset 2 included lower-resolution images taken also in clinical settings but with the Daryou Intraoral Camera, a more affordable option also used in clinical settings. When trained on Dataset 1, the YOLOv8 model achieved an overall mAP (mean average precision) score of 82.7 %, with a precision of 81.5 %, and a recall of 81.4 %. For Dataset 2, the YOLOv8 model achieved an overall mAP score of 79.9 %, a precision of 81.0 %, and a recall of 79.7 %. The results for both Datasets are noteworthy, showing performance comparable to the results of other industry-leading studies. The results also show YOLOv8 model's potential for at-home use. This accessibility and affordable would allow for patients to conduct reliable prescreenings at home, helping them maintain oral hygiene and detect potential dental issues early on.