Towards Improved Identification of Vertebral Fractures in Routine Computed Tomography (CT) Scans: Development and External Validation of a Machine Learning Algorithm
Authors' conclusion
We demonstrated that a machine learning algorithm trained for VF detection achieved strong performance on an external validation set.
Study context
- Population
- Patients aged ≥50 years with routine abdominal/chest CT scans; training set from Universitair Ziekenhuis (UZ) Brussel (1011 scans), external validation set from Hospital of Holbaek (2000 subjects)validation set
- Participants
- 2,000 patients
- Comparison
- Ensemble machine learning model vs reference standard readings
- Outcome
- Identification of subjects with ≥1 moderate or severe VF (SQ2-3) in the external validation set
Methodological detailsmachine learning algorithm development, external validation, ensemble of four models
- Machine learning algorithm development
- External validation
- Ensemble of four models
Performance
As reported in the abstract
Identification of subjects with ≥1 moderate or severe VF
| Test or model | AUROC | Accuracy | Kappa | Sensitivity | Specificity |
|---|---|---|---|---|---|
| Ensemble model | 0.88 (95% CI, 0.85-0.90) | 0.92 (95% CI, 0.91-0.93) | 0.72 (95% CI, 0.67-0.76) | 0.81 (95% CI, 0.76-0.85) | 0.95 (95% CI, 0.93-0.96) |
Ensemble model
AUROC0.88 (95% CI, 0.85-0.90)
Accuracy0.92 (95% CI, 0.91-0.93)
Kappa0.72 (95% CI, 0.67-0.76)
Sensitivity0.81 (95% CI, 0.76-0.85)
Specificity0.95 (95% CI, 0.93-0.96)
Further findings
VF prevalence in validation set per reference standard
- Scans with ≥1 VF (SQ2-3)
- 15.3%
- Fractured vertebrae (SQ2-3) of evaluable vertebrae
- 663 of 24,930 (2.7%)
Original publication
Towards Improved Identification of Vertebral Fractures in Routine Computed Tomography (CT) Scans: Development and External Validation of a Machine Learning Algorithm. J Bone Miner Res. Published 15 October 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →