Improving Vertebral Fracture Prediction via Multitask Deep Learning of Bone and Muscle CT

Overview

This study developed a multitask deep learning model using opportunistic CT scans to simultaneously detect vertebral fractures and predict future fracture risk by analyzing bone and muscle features. The model was externally validated on independent cohorts, demonstrating improved fracture risk stratification beyond traditional methods.

Background

Fragility fractures are increasing globally with aging populations, posing significant healthcare challenges. Dual-energy X-ray absorptiometry (DXA) is the standard for osteoporosis diagnosis but is underutilized, leaving many at risk untreated. Opportunistic CT scans, increasingly performed for other indications, offer a cost-effective opportunity to assess fracture risk by analyzing bone mineral density and muscle quality. Incorporating muscle imaging alongside bone assessment may enhance fracture risk prediction beyond clinical tools like FRAX, which do not capture detailed imaging features.

Data Highlights

Key Findings

• A multitask deep learning model was developed to detect vertebral fractures and predict future fracture risk using abdominal CT scans.
• The model incorporated both bone and muscle imaging features, leveraging shared information to improve prediction accuracy.
• External validation on independent cohorts from two hospitals confirmed the model's robustness and generalizability.
• The approach outperformed traditional fracture risk assessment tools by integrating imaging-based structural and compositional data not captured by FRAX or BMD alone.
• Patients aged 50-80 years undergoing routine abdominal CT scans can be opportunistically screened for fracture risk without additional imaging or cost.

Clinical Implications

This multitask deep learning framework enables clinicians to identify patients at high risk for vertebral fractures using existing CT imaging, facilitating earlier intervention and personalized management. Incorporating muscle quality alongside bone assessment provides a more comprehensive risk profile, potentially improving prevention strategies. Opportunistic use of CT scans could enhance fracture risk screening in populations less likely to undergo DXA.

Conclusion

Multitask deep learning analysis of bone and muscle via opportunistic CT scans offers a promising tool for improved vertebral fracture detection and risk prediction. This approach may complement existing clinical assessments and support proactive fracture prevention in aging populations.

References

1. Global Aging and Fragility Fractures Impact -- References [1,2,3]
2. DXA as Gold Standard and Treatment Gaps -- Reference [4,5]
3. Opportunistic CT for Fracture Risk Assessment -- References [6,7,8,9]
4. Role of Muscle in Fracture Risk -- Reference [10]
5. FRAX Tool for Fracture Risk Estimation -- Reference [11]