Abstract

This paper discusses the use of the state-of-the-art object detection models in the recognition of dental prob lems in panoramic X-ray images, with the particular attention paid to the most recent YOLOv8 architecture. A comparative study between YOLOv3, YOLOv5, YOLOv7, YOLOv8 and conventional detectors Faster R-CNN and SSD is also elaborated in our research. On a curated data-set of 126 panoramic dental radiographs with per-radiograph annotations of six different dental conditions, we trained each model on the data and the models were evaluated on precision, recall, and the mean Average Precision (mAP). YOLOv8 was proven to be better in detection accuracy, speed, and localization precision, which makes it a functional tool in enacting real-time dental diagnostics. The technology is the advance and the introduction of AI-based diagnostics will be included in the daily dental practice, enhancing early diagnosis and treatment strategy.

DOI: doi.org/10.63721/25JPAIR0104

To Read or Download the Article  PDF