Can CT Radiomics Detect Osteoporosis?

Radiomics analysis of lumbar spine computed tomography is compared with established imaging-based measures of bone density.

By
Kathryn Wighton
February 24, 2026
3 min

Conexiant

Overview

A machine-learning radiomics model using lumbar spine CT outperformed traditional methods in detecting osteoporosis. The study involved 166 patients and demonstrated high sensitivity and specificity for the radiomics model compared to vertebral bone quality scoring and Hounsfield unit measurements.

Background

Osteoporosis is a significant global health issue characterized by reduced bone density and increased fracture risk, particularly in the aging population. Accurate detection of osteoporosis is crucial for preventing fragility fractures, which can lead to severe morbidity and increased healthcare costs. Traditional screening methods, such as dual-energy X-ray absorptiometry (DXA), are standard, but advancements in imaging techniques like CT radiomics may enhance diagnostic accuracy.

Data Highlights

{'table': {'headers': ['Model', 'AUC (Training)', 'AUC (Test)', 'Sensitivity', 'Specificity', 'Negative Predictive Value'], 'rows': [['XGBoost', '0.89', '0.91', '89%', '82%', '94%']]}}

Key Findings

The radiomics model achieved an AUC of 0.91 in the test cohort.
210 out of 664 vertebrae were classified as osteoporotic, representing 32% prevalence.
Feature selection reduced the initial 851 radiomic features to 30 candidates, ultimately identifying nine key features.
The XGBoost model outperformed traditional vertebral bone quality scoring and Hounsfield unit measurements.
Decision curve analysis indicated greater net clinical benefit for the radiomics model across probability thresholds.

Clinical Implications

The radiomics-XGBoost model may serve as a valuable adjunctive screening tool for osteoporosis, particularly in patients undergoing CT imaging for other reasons. However, it should not replace DXA, which remains the standard for osteoporosis diagnosis.

Conclusion

The study highlights the potential of machine-learning radiomics in enhancing osteoporosis detection through CT imaging, warranting further validation in multicenter studies to confirm its generalizability.