Enhanced Osteoporosis Detection Using Artificial Intelligence: A Deep Learning Approach to Panoramic Radiographs with...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Osteoporosis Research

Background:

• Osteoporosis affects 60% of women over 50, necessitating early detection beyond post-fracture assessments.
• Dual-energy X-ray absorptiometry (DXA) is the gold standard but often used late.
• Panoramic radiographs (PRs) are common dental tools, offering potential for early osteoporosis screening.

Purpose of the Study:

• To develop a robust AI application for accurate osteoporosis identification using PRs.
• To establish reliable and early diagnostic capabilities for osteoporosis.
• To address methodological concerns in previous AI-based PR analysis for osteoporosis.

Main Methods:

• A dataset of 250 PRs from osteoporosis and non-osteoporosis groups was utilized.
• A convolutional neural network (CNN) classifier was trained and validated on cropped mental foramen regions.
• The AI model was tested using comparative analysis (osteoporosis vs. matched non-osteoporosis) and (osteoporosis vs. younger non-osteoporosis).

Main Results:

• The AI model achieved an F1 score of 0.74 and an AUC of 0.8401 when comparing osteoporosis to age/gender-matched controls.
• For differentiating osteoporosis in younger patients, the model demonstrated 98% accuracy and an AUC of 0.9812.
• The study highlights the potential of deep learning in PRs for osteoporosis detection.

Conclusions:

• This research presents a proof-of-concept for AI-driven osteoporosis detection in PRs.
• The findings underscore the potential for early and reliable osteoporosis diagnostics via dental imaging.
• Methodological rigor is crucial for validating AI applications in medical diagnostics.