Detection of external root resorption in periapical radiographs using YOLO-based deep learning model.

Area of Science:

• Dentistry
• Radiology
• Artificial Intelligence

Background:

• External root resorption is a destructive dental condition often asymptomatic, leading to significant tooth tissue loss and potential extraction.
• Early diagnosis is crucial but challenging due to the lack of symptoms.

Purpose of the Study:

• To develop and evaluate artificial intelligence (AI) algorithms for assisting in the diagnosis of external root resorption.
• To assess the efficacy of YOLOv5x-cls and YOLOv5x-seg models in detecting and segmenting external root resorption.

Main Methods:

• 110 extracted teeth were chemically demineralized and imaged using a radioconjugate phantom model.
• A dataset of 584 periapical radiographs was created.
• YOLOv5x-cls and YOLOv5x-seg models were employed for external root resorption detection and segmentation.

Main Results:

• The YOLOv5x-cls model achieved an F1 score of 1.0 for classifying external root resorption, indicating high accuracy.
• The YOLOv5x-seg model achieved an F1 score of 0.8593 for segmenting external root resorption, demonstrating effective performance.
• Classification model performance was superior to the segmentation model.

Conclusions:

• AI algorithms demonstrate high success rates in detecting external root resorption using a phantom model.
• This study presents an innovative and reliable AI-driven approach for radiological evaluation of external root resorption.