Periodontitis bone loss detection in panoramic radiographs using modified YOLOv7

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Periodontitis is a prevalent dental disease leading to tooth loss.
• Diagnosing periodontitis-related bone loss from radiographs is challenging, requiring expertise and time.
• Automated diagnosis systems are needed to improve efficiency and accuracy.

Purpose of the Study:

• To develop and evaluate a deep learning model for automated diagnosis of bone loss in periodontitis.
• To address the limitations of manual radiographic analysis.

Main Methods:

• A modified You Only Look Once (YOLO)v2 model, termed YOLOv7-M, was developed.
• YOLOv7-M incorporates a focus module and feature fusion for enhanced performance.
• The model was trained and evaluated on a dental radiograph dataset.

Main Results:

• YOLOv7-M achieved high performance metrics: F1-score (92.5%), precision (91.7%), recall (87.1%), and mAP (91.0%).
• The model demonstrated superior accuracy and speed compared to YOLOv5 and YOLOv7.
• YOLOv7-M outperformed other state-of-the-art object detectors in statistical evaluations.

Conclusions:

• The proposed YOLOv7-M model shows significant potential for automated periodontitis diagnosis.
• This technology can aid clinicians in early detection and treatment, improving patient outcomes.
• Automated analysis of dental radiographs can enhance the management of periodontitis.