AI for Opportunistic Identification of BMD and Bone Microarchitecture: A Narrative Review.
Authors
Affiliations (3)
Affiliations (3)
- Department of Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, USA. [email protected].
- Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine, Munich, Germany. [email protected].
- Department of Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, USA.
Abstract
Artificial intelligence (AI) has rapidly transformed the potential for opportunistic assessment of bone mineral density (BMD) and bone microarchitecture from routinely acquired imaging. This narrative review synthesizes recent advances across radiography, computed tomography (CT), and magnetic resonance imaging (MRI), focusing on AI-driven methods for automatic extraction of bone quality markers. In radiography, convolutional and transformer-based architectures achieve near-DXA precision by integrating regional texture and clinical features. In CT, deep learning pipelines combining segmentation and end-to-end regression now enable scalable volumetric BMD and trabecular texture analysis, with correlations of approximately 0.8-0.9 versus reference techniques and near-perfect agreement with dedicated quantitative CT. MRI-based methods, leveraging fat-water signal ratios and textural radiomics, increasingly provide surrogate markers of bone integrity. Barriers to clinical translation include interoperability with existing systems, lack of reimbursement frameworks, and limited explainability of black-box models. Emerging foundation models (e.g., SAM, BiomedCLIP) and synthetic imaging networks promise cross-modality generalization. Future work must focus on regulatory validation, model drift surveillance, and equitable deployment to ensure AI-based opportunistic osteoporosis screening improves outcomes across diverse populations.