Development of an automated YOLO v8-based deep learning teeth numbering model for digital orthopantomogram

Artificial Intelligence is a field of computer science which replicates human intelligence. The current study evaluates a deep learning model that automatically detects and numbers both primary and permanent teeth on panoramic images. This approach aims to streamline the time-consuming manual identification process, particularly during the mixed dentition stage, making treatment planning more efficient for dental professionals.

The dataset include 500 anonymized panoramic radiographs split into 320 for training and 180 for validation from the ages 3 to 60 years with 11,851 permanent and 2600 primary tooth labels, including third molars regardless of lesions, missing, restored teeth and implants. Poor diagnostic quality and completely edentulous images were excluded. An individual dataset of 1000 images were used for testing. The radiographs were sourced from the archive of the Department of Oral Medicine and Radiology, Government Dental College and Research Institute, Bangalore. The YOLO v8 deep CNN model architecture, following the FDI teeth numbering system was employed.

The recall and precision rate of the model was 0.954 and 1.000 respectively compared to the ground truth data. Further the F1 score was 0.92 indicating a good balance between the recall and precision.

The model has potential for practical application in automated dental radiograph analysis. It can aid clinicians in identifying and numbering the primary and permanent teeth on an orthopantomogram enhancing treatment planning and efficiency. It further expands the scope of artificial intelligence in the field of oral and maxillofacial diagnosis and radiology.

Not applicable.