The Diagnostic Value of Image-Based Machine Learning for Osteoporosis: Systematic Review and Meta-Analysis.

Area of Science:

• Medical imaging analysis
• Artificial intelligence in healthcare
• Osteoporosis diagnosis

Background:

• Osteoporosis (OP) poses a significant health challenge for aging populations.
• Machine learning (ML) and deep learning (DL) models enhance diagnostic accuracy and efficiency in medical imaging.
• A systematic assessment of different medical imaging modalities for OP diagnosis using ML is lacking.

Purpose of the Study:

• To systematically review and elucidate the role of deep learning (DL) models in osteoporosis (OP) detection across various medical imaging modalities.
• To provide a comprehensive overview of the diagnostic performance of ML-based OP detection methods.

Main Methods:

• Systematic literature search of PubMed, Embase, Cochrane Library, and Web of Science for ML-based OP diagnosis studies.
• Quality assessment of included studies using the Quality Assessment of Diagnostic Accuracy Studies-2 tool.
• Meta-analysis of sensitivity and specificity using a bivariate mixed-effects model, stratified by imaging modality (X-ray, CT, MRI) and subgroup analyses.

Main Results:

• Included 60 studies with 66,195 participants; 22 used X-ray, 37 used CT, and 3 used MRI.
• X-ray models showed high pooled sensitivity (SEN) and specificity (SPC): appendicular skeleton (0.97, 0.90), mandible (0.94, 0.80), lumbar spine (0.87, 0.82).
• CT models demonstrated strong performance: hip joint (0.87, 0.92), thoracic spine (0.91, 0.94), lumbar spine (0.91, 0.92).

Conclusions:

• ML, particularly DL with X-ray and CT, demonstrates high diagnostic accuracy for osteoporosis.
• Limited studies utilized MRI, and a lack of external validation presents interpretative limitations.
• Future research should focus on developing AI tools with broader applicability and improved diagnostic precision for osteoporosis.