Deep-Learning-Based Detection of Vertebral Fracture and Osteoporosis Using Lateral Spine X-Ray Radiography
Authors' conclusion
The proposed method could detect prevalent VFs and osteoporosis, and it improved referral of individuals at high risk of fracture to DXA testing more than clinical indications alone.
Study context
- Population
- Patients aged ≥50 years who underwent lateral spine radiography, Severance Hospital, Korea, 2007–2018 (26,299 X-rays; VF prevalence 18.6%, osteoporosis prevalence 40.3%); external test set from another hospital (Yongin)derivation cohort
- Participants
- 9,276 patients
- Comparison
- Deep-learning scores (VERTE-X pVF, VERTE-X osteo) vs clinical models for detecting prevalent VF and osteoporosis on lateral spine radiography
- Outcome
- AUROC for detection of prevalent VF and osteoporosis
Methodological detailsdeep learning, derivation cohort, internal test set (20% hold-out), external test set
- Deep learning
- Derivation cohort
- Internal test set (20% hold-out)
- External test set
Performance
As reported in the abstract
- Internal test set, VF: VERTE-X pVF vs clinical model AUROC
- 0.93 vs 0.78
- Internal test set, osteoporosis: VERTE-X osteo vs clinical model AUROC
- 0.85 vs 0.79
- External test set, VF: VERTE-X pVF vs clinical model AUROC
- 0.92 vs 0.79
- External test set, osteoporosis: VERTE-X osteo vs clinical model AUROC
- 0.83 vs 0.65
- P value (all comparisons)
- p < 0.01 for all
Further findings
Net reclassification improvement (NRI) for referral of individuals with osteoporosis to DXA testing (deep-learning scores added to clinical indications)
- Internal test set NRI
- 0.10
- External test set NRI
- 0.14
- P value (all)
- P < 0.001 for all
Original publication
Deep-Learning-Based Detection of Vertebral Fracture and Osteoporosis Using Lateral Spine X-Ray Radiography. J Bone Miner Res. Published 10 April 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →