Deep Learning for Opportunistic Vertebral Fracture Detection on Routine Thoraco-abdominal Computed Tomography: A
Kalab Yigermal Gete1, Zulaika Nassir Idris2, Belete Achamyelew Ayele3
1School of Medicine, College of Medicine and Health Sciences, Bahir Dar University, Bahir Dar, Ethiopia (K.Y.G.); EPIC Health Systems, Addis Ababa, Ethiopia (K.Y.G.).
Academic Radiology
|June 29, 2026
Summary
Deep learning (DL) shows promise for detecting vertebral fractures (VFs) on routine CT scans, with high specificity and moderate-to-high sensitivity. Further prospective studies are needed before widespread clinical deployment.
Area of Science:
- Radiology
- Artificial Intelligence
- Medical Imaging
Background:
- Vertebral fractures (VFs) are common and often missed on routine computed tomography (CT) scans.
- Deep learning (DL) presents a potential solution for opportunistic case-finding of VFs.
Purpose of the Study:
- To systematically review the diagnostic accuracy of deep learning (DL) algorithms for detecting vertebral fractures (VFs) on routine CT scans.
- To assess the current evidence base for DL in VF detection and provide recommendations for future research and deployment.
Main Methods:
- Systematic review following PRISMA-DTA guidelines, searching MEDLINE, Embase, and Web of Science.
- Quality assessment using QUADAS-2 and descriptive assessment of AI reporting.
- Bivariate random-effects/HSROC modeling and sensitivity analyses.
Main Results:
- Seven retrospective studies (N=11,615) evaluated DL for detecting grade 2-3 VFs.
- Pooled sensitivity was 0.83 (95% CI: 0.73-0.90) and specificity was 0.92 (95% CI: 0.90-0.94).
- Evidence is early, heterogeneous, with low-to-moderate risk of bias and incomplete AI reporting.
Conclusions:
- DL demonstrates high specificity and moderate-to-high sensitivity for patient-level VF detection on CT.
- Current evidence is limited, supporting prospective evaluation of DL as an aid, not for routine standalone use.
- Further research should focus on prospective validation and integration into clinical workflows.
