AI-Powered Identification of Osteoporosis in Dental Panoramic Radiographs: Addressing Methodological Flaws in Current...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Osteoporosis Research

Background:

• Osteoporosis affects 60% of women over 50, with diagnosis often delayed until fractures occur.
• Current diagnostic methods like DXA scans are not always used for early detection.
• Panoramic radiographs (PRs) show potential for osteoporosis screening, but require improved methodologies.

Purpose of the Study:

• To develop a robust artificial intelligence (AI) application for accurate osteoporosis identification in PRs.
• To address methodological flaws in previous studies using PRs for osteoporosis detection.
• To create a reliable AI tool for early osteoporosis screening via dental radiographs.

Main Methods:

• Utilized 348 PRs for model development, 58 for validation, and 51 for testing.
• Employed the YOLOv8 object detection model to identify regions of interest in PRs.
• Processed extracted regions using the EfficientNet classification model for osteoporosis detection.

Main Results:

• The AI model achieved an overall sensitivity of 0.83 and an F1-score of 0.53 for osteoporosis detection.
• The Area Under the Curve (AUC) for the model was 0.76.
• Detection sensitivity varied by region, with the mental foramen region showing the highest (0.80).

Conclusions:

• This research demonstrates a proof-of-concept AI algorithm for identifying osteoporosis in PRs.
• Deep learning holds significant potential for early osteoporosis detection using dental radiographs.
• Rigorous methodological evaluation is crucial for validating AI algorithm performance in medical diagnostics.