Vertebral fractures identified on lateral DXA images by deep learning predict incident fractures in older women
Authors' conclusion
Automated XVFA predicted incident fractures similarly to manual assessment.
Study context
- Population
- Women from the population-based SUPERB study not included in development of the XVFA model
- Participants
- 423 women
- Exposure
- Women with vertebral fractures (VFs) vs women without VFs, identified by manual vertebral fracture assessment (VFA) or by XVFA (deep learning) on lateral DXA images
- Manual VFA VFsn = 102
- XVFA VFsn = 187
- Outcome
- X-ray verified incident fractures
- Follow-up
- 8 years
Methodological detailspopulation-based, Cox proportional hazards models adjusted for clinical risk factors (CRFs) and femoral neck (FN) bone mineral density (BMD)
- Population-based
- Cox proportional hazards models adjusted for CRFs and FN BMD
Main finding
As reported in the abstract
Incident fractures
Manual VFA VFs
48%
XVFA VFs
43%
No VFs
20% (manual VFA) and 16% (XVFA)
Each group vs no VFsHazard ratio (95% CI)
Manual VFA VFs
2.04(1.35–3.07)
XVFA VFs
2.32(1.55–3.48)
Observational data: associations, not effects.
Further findings
| Outcome | Manual VFA VFs | XVFA VFs |
|---|---|---|
| Vertebrae classified on baseline lateral images | 4563 vertebrae | 5532 vertebrae |
- Vertebrae classified on baseline lateral images: counts by method (manual VFA, XVFA); numerical differences partly reflecting image quality limitations
Original publication
Vertebral fractures identified on lateral DXA images by deep learning predict incident fractures in older women. Osteoporos Int. Published 12 May 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →