AI IN MEDICAL IMAGING
Vertebral Fractures Identified by Deep Learning Predict Incident Fractures
Authors: Mattias Lorentzon, Victor Wåhlstrand, Jennifer Alvén, Ida Häggström, Lisa Johansson
Publication: Osteoporosis International (DOI: 10.1007/s00198-026-08072-9)
Executive Impact: Pioneering AI in Osteoporosis Detection
This groundbreaking study introduces an explainable deep learning model (XVFA) that revolutionizes vertebral fracture assessment, offering superior scalability and objectivity compared to traditional methods, with significant implications for patient care and risk management.
Key Findings at a Glance
0
Women in Study
0
Years Follow-up
0
XVFA HR for Incident Fractures
0
XVFA Detection Sensitivity
Summary: XVFA is an AI-based method for identifying vertebral fractures on DXA images. In 423 women followed for 8 years, vertebral fractures identified by XVFA or manual assessment were associated with a twofold increased risk of incident fractures. XVFA predicted fracture risk comparably to manual assessment, supporting automated vertebral fracture detection.
Purpose: Vertebral fractures (VFs), identified by vertebral fracture assessment (VFA) using dual-energy X-ray absorptiometry (DXA), predict incident fractures independently of clinical risk factors (CRFs) and bone mineral density (BMD). Most VFs remain clinically unrecognized. This study evaluated whether VFs identified using a deep learning-based method on lateral DXA images predict incident fractures comparably to manual VFA.
Methods: Associations between prevalent VFs and incident fractures were investigated in 423 women from the population-based SUPERB study who were not included in development of the explainable deep learning model (XVFA). Vertebrae were classified by manual VFA and XVFA. Incident fractures were X-ray verified. Cox proportional hazards models assessed fracture risk adjusted for CRFs and femoral neck (FN) BMD.
Results: Manual VFA reading and XVFA were used on baseline lateral images and classified 4563 and 5532 vertebrae, respectively, with numerical differences partly reflecting image quality limitations. VFs were identified in 102 women by manual VFA and 187 by XVFA. During 8 years of follow-up, incident fractures occurred in 48% of women with manual VFA VFs and 43% with XVFA VFs, vs 20% and 16% of women without VFs. Women with VFs had a higher fracture risk whether identified manually (HR 2.04; 95% CI, 1.35–3.07) or by XVFA (HR 2.32; 95% CI, 1.55–3.48), compared with women without VFs. Results remained significant after adjustment for CRFs and FN BMD.
Conclusion: Automated XVFA predicted incident fractures similarly to manual assessment. These findings support the clinical utility of deep learning-based VF detection, which may enhance fracture risk assessment and management in routine practice.
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Explainable Deep Learning for VFA
XVFA, a novel deep learning method, uses neural networks for VFA in lateral spine DXA images, employing state-of-the-art landmark detection. It classifies identified vertebrae based on clinical GSQ compression criteria, providing explainability and uncertainty estimates. This approach allows tracing the model's decision back to specific vertebral measurements and clinically interpretable criteria, enhancing transparency. The model was trained on 11,605 annotated vertebrae from the SUPERB cohort.
Enterprise Process Flow
Deep Learning Model Development
→
Annotated Vertebrae (11,605 from SUPERB)
→
Landmark Detection & Localization
→
GSQ Compression Criteria Application
→
VF Classification & Uncertainty Estimation
→
Predict Incident Fractures
Quantifying AI's Predictive Power
XVFA-identified VFs were associated with a twofold increased risk of incident fractures (HR 2.32; 95% CI, 1.55–3.48), compared with women without VFs. This association remained significant after adjustment for CRFs and FN BMD. This predictive power was comparable to manual assessment (HR 2.04; 95% CI, 1.35–3.07).
2.32x
Times Increased Fracture Risk with XVFA-Identified VFs (vs. no VFs)
Manual vs. AI: Performance Comparison
XVFA predicted incident fractures similarly to manual assessment. During 8 years of follow-up, incident fractures occurred in 48% of women with manual VFA VFs and 43% with XVFA VFs, vs 20% and 16% of women without VFs respectively. Manual VFA identified VFs in 102 women (24%), while XVFA identified VFs in 187 women (44%). Manual VFA classified 4563 vertebrae, XVFA classified 5532 vertebrae, partly reflecting image quality limitations and XVFA not being explicitly trained to disregard poorly visible vertebrae, leading to more detections.
| Feature | Manual VFA | XVFA |
|---|---|---|
| VF Identification Method |
|
|
| Vertebrae Classified |
|
|
| Women with VFs Identified |
|
|
| Incident Fractures (8-year follow-up) |
|
|
| HR for Incident Fractures (vs. no VFs) |
|
|
| Clinical Utility |
|
|
Translating AI to Clinical Practice
The study supports the clinical utility of deep learning-based VF detection, which may enhance fracture risk assessment and management in routine practice. It provides a robust framework for translating deep learning tools into real-world skeletal health assessment, addressing the problem of clinically unrecognized VFs and the treatment gap among high-risk individuals.
Streamlining Osteoporosis Management with AI
Challenge: A significant proportion of vertebral fractures (VFs) remain clinically unrecognized, leading to delayed diagnosis and undertreatment of osteoporosis, increasing the risk of subsequent fractures.
Solution: Implementation of XVFA, a deep learning-based tool for automated VF detection on DXA images, provides a scalable and objective method for identifying prevalent VFs.
Outcome: Automated XVFA, with comparable predictive power to manual VFA, enhances fracture risk assessment, enabling earlier identification of high-risk patients and supporting timely intervention strategies to reduce incident fractures and improve patient outcomes.
Calculate Your Potential AI ROI
Estimate the efficiency gains and cost savings your enterprise could achieve by integrating advanced AI solutions for medical imaging analysis.
Estimated Annual Savings
$0
Annual Hours Reclaimed
0
Your AI Implementation Roadmap
A structured approach to integrate advanced AI solutions into your enterprise workflow, ensuring seamless adoption and measurable impact.
Phase 1: Discovery & Strategy
Initial consultation to understand your specific needs, assess existing infrastructure, and define clear objectives and a tailored AI strategy. This phase includes a detailed ROI projection.
Phase 2: Data Preparation & Model Training
Collecting and preparing relevant data, followed by custom AI model training or fine-tuning using proprietary datasets to ensure optimal performance and accuracy for your domain.
Phase 3: Integration & Deployment
Seamless integration of the AI solution into your existing systems and workflows, followed by rigorous testing and pilot deployment to validate functionality and user experience.
Phase 4: Optimization & Scaling
Continuous monitoring and iterative refinement of the AI models for ongoing optimization. Scaling the solution across your organization to maximize its impact and deliver sustained value.
Ready to Transform Your Enterprise with AI?
Don't be left behind. Leverage the power of explainable AI to drive efficiency, accuracy, and innovation in your operations. Schedule a personalized consultation with our experts today.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat