Deep Learning based Teeth Detection and Numbering System using YOLO

In the field of Dental Medicine, the proper understanding and analysis of OPGs (Dental Orthopantomogramme) play a significant role in the diagnostic process. Accurate detection and numbering of teeth are crucial for applications such as enhancing dental diagnosis, treatment planning, and oral health assessment. This research incorporates a YOLOv8 algorithm with a custom curated dataset of $\mathrm{4 5 5 0}$ images to tackle this pivotal task. The dataset includes a wide variety of dental images with 32 classes based on the FDI numbering system. To ensure privacy protection, details like names, genders, ages, medical records, or timestamps are omitted. The evaluation metrics include precision, recall, and mean Average Precision (mAP). The results have provided valuable insights, with YOLOv8 exhibiting an impressive precision of 0.98, a recall of 0.97, and an mAP50 of 0.98, showcasing its robustness in teeth detection. These results demonstrate its potential for practical implementation, streamlining clinician tasks, saving time, and enhancing diagnostic accuracy. The potential to complement or even replace human assessment supports its role in the decision-making process in dental diagnostics.