AI-derived bone mineral density from standard radiographs compared with DXA for fracture prediction in a 10-year real-world cohort study
Authors' conclusion
AI-derived BMD from routine radiographs closely correlates with DXA and demonstrates comparable fracture prediction.
Study context
- Population
- Adults aged ≥ 20 years with both lumbar DXA and radiographic examinations within six months, January 2014–December 2024mean follow-up 6.4 years
- Participants
- 540 adults
- Exposure
- AI-derived BMD from routine radiographs vs DXA-derived BMD
- Outcome
- Agreement of AI- and DXA-derived BMD
Methodological detailsreal-world, opportunistic screening, Pearson correlation, intraclass correlation coefficient (ICC) and Bland-Altman analysis, ROC analysis, Cohen's κ and logistic regression
- Real-world
- Opportunistic screening
- Pearson correlation, ICC and Bland-Altman analysis
- ROC analysis, Cohen's κ and logistic regression
Main finding
As reported in the abstract
Pearson r
0.943
ICC
0.934
Observational data: associations, not effects.
Further findings
| Outcome | AI-derived BMD | DXA-derived BMD |
|---|---|---|
| Prediction of vertebral fractures | ||
| AUC | 0.704 | 0.678 |
| Prediction of hip fractures | ||
| AUC | 0.716 | 0.678 |
Agreement of AI- and DXA-derived BMD
- Pearson r
- 0.943
- ICC
- 0.934
- Strong agreement
Osteoporosis diagnosis (T-score ≤ - 2.5) by AI-derived T-scores
- AUC
- 0.959
- κ
- 0.74
- Accuracy
- 90%
Prediction of vertebral fractures
- Comparable performance
Prediction of hip fractures
- Comparable performance
Association of AI-derived BMD with fracture risk
- Lower AI-derived BMD independently associated with higher fracture risk
Original publication
AI-derived bone mineral density from standard radiographs compared with DXA for fracture prediction in a 10-year real-world cohort study. Osteoporos Int. Published 18 June 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →