Can MRI scans and machine learning predict osteoporosis and abnormal bone density?

Osteoporosis weakens bones and raises fracture risk. Doctors usually diagnose it with DEXA scans or QCT. But MRI scans are common for back pain, and researchers are now using machine learning (a type of artificial intelligence) to analyze those MRI images and detect bone loss. Studies show this approach can predict osteoporosis and abnormal bone density with good accuracy, sometimes over 95%.

What the research says

A 2024 study developed machine learning models using routine lumbar MRI scans from 160 patients. The models analyzed radiomic features (detailed image patterns) from T1-weighted and T2-weighted MRI sequences. Using QCT as the reference standard, the best model (K-nearest neighbor on T1WI) achieved an AUC of 0.821 for predicting osteoporosis, meaning it correctly identified most cases ⁵. Another 2024 study used deep learning on MRI and CT scans from 220 patients. Their unimodal MRI model reached 96.54% balanced accuracy, and a multimodal model combining MRI and CT hit 98.90% accuracy ⁹. These results show MRI-based machine learning can be as reliable as CT-based methods.

MRI-based scores also predict future fractures. A meta-analysis of 29 studies (7364 patients) found that the vertebral bone quality (VBQ) score from MRI was significantly higher in patients who later had fractures after vertebral augmentation. The VBQ score had a sensitivity of 0.85 and specificity of 0.66 for predicting subsequent fractures ². This suggests MRI can identify high-risk patients even without a DEXA scan.

Machine learning is not limited to MRI. CT-based quantitative CT (QCT) parameters, such as cortical bone thickness and volume at the L1 vertebra, also predict osteoporosis with good accuracy ⁸. A narrative review notes that AI tools for osteoporosis have evolved from simple radiomics to end-to-end deep learning, improving both image segmentation and fracture risk prediction ¹⁰¹¹. However, the evidence for MRI-based prediction is still growing, and most studies are retrospective, meaning they look back at existing data rather than following patients forward.

What to ask your doctor

• Could an MRI scan with machine learning analysis help assess my bone health instead of a DEXA or CT scan?
• If I already have a lumbar MRI for back pain, can it be re-analyzed to check for osteoporosis?
• How does the accuracy of MRI-based bone density prediction compare to standard DEXA or QCT in your practice?
• Are there any risks or added costs to using MRI for osteoporosis screening?
• Should I consider a follow-up DEXA scan even if my MRI-based prediction looks normal?