Automated detection and classification of dental findings in radiographs using a computer vision (YOLO) model: a tool for enhancing dental education
TL;DR
A deep learning-based method using the state-of-the-art You Only Look Once 8 (YOLOv8) architecture of object detection for automating four frequent dental findings on pantomographs and the results are quite suggestive of the role generative artificial intelligence tools could foster in diversifying learning outcomes during AI-driven dentistry modules.
Abstract
There is considerable value to dental radiographs in the diagnosis and treatment planning of dentistry in modern practice, but the skill (and knowledge) required for a diagnostic interpretation is largely one of low skill but of high demands for knowledge. Very recently, modern artificial intelligence, especially computer vision, has opened up new possibilities for improving diagnostic accuracy and promoting educational activities. This article presents a deep learning-based method using the state-of-the-art You Only Look Once 8 (YOLOv8) architecture of object detection for automating four frequent dental findings on pantomographs: restorations, impacted teeth, implants, and caries. A standardized dataset was used in this case, even though several marking, cleaning, and enhancement (and thus, adjustments) processes have been applied to the dataset model. The configuration displayed remarkably high values of top alertness, positive guesses in every quadrant, and for every case to identify impacted teeth in the entire task. On top of its potential offerings, it may also provide a real-time and objective assessment of students’ skills in X-ray interpretation. The results are quite suggestive of the role generative artificial intelligence (AI) tools could foster in diversifying learning outcomes during AI-driven dentistry modules. There is, thus, a compelling case for further studies in this direction with more expansive and diverse datasets.