Detection of Fractured Endodontic Instruments in Periapical Radiographs: A Comparative Study of YOLOv8 and Mask R-CNN

Area of Science:

• Dental Radiology
• Artificial Intelligence in Medicine
• Medical Image Analysis

Background:

• Accurate localization of fractured endodontic instruments (FEIs) in periapical radiographs (PAs) is crucial but challenging.
• Existing methods for detecting FEIs and root canal treatments (RCTs) require improvement for clinical efficiency.

Purpose of the Study:

• To evaluate the performance of YOLOv8 and Mask R-CNN for detecting FEIs and RCTs in PAs.
• To compare the diagnostic capabilities of these AI models against experienced endodontists.

Main Methods:

• A dataset of 1050 annotated PAs was utilized for training and evaluation.
• YOLOv8 and Mask R-CNN models were assessed for accuracy, IoU, mAP50, and inference time.
• AI predictions were compared with human annotations, and observer agreement was analyzed.

Main Results:

• YOLOv8 achieved 97.40% accuracy and 98.9% mAP50 with a 14.6 ms inference time.
• Mask R-CNN achieved 98.21% accuracy and 95% mAP50 with an 88.7 ms inference time.
• No statistically significant difference was found between AI models and experienced endodontists' diagnostic performance.

Conclusions:

• Both YOLOv8 and Mask R-CNN demonstrate high diagnostic accuracy and reliability, comparable to endodontists.
• YOLOv8 is suitable for real-time applications due to its speed; Mask R-CNN excels in segmentation.
• AI integration in dental diagnostics shows promise for improving clinical outcomes.