Advances in Imaging-Based Fracture Risk Assessment for Unlocking Latent Skeletal Fragility

TL;DR

This paper reviews how imaging and deep learning can improve predicting bone fractures beyond traditional methods.

Contribution

The paper highlights novel imaging-derived methods and deep learning models for enhanced fracture risk assessment.

Findings

Opportunistic imaging analyses can detect hidden vertebral fractures and estimate bone density.

Deep learning models predict fracture risk directly from clinical images across modalities.

Imaging reveals additional skeletal fragility factors beyond traditional bone density measurements.

Abstract

This review summarizes recent advancements in imaging-based fracture risk assessment utilizing routinely acquired clinical images. We explore how imaging-derived methodologies and deep learning techniques can enhance conventional tools, such as dual energy X-ray absorptiometry (DXA)-derived bone mineral density and FRAX®, by capturing additional factors influencing skeletal fragility.

Recent studies indicate that opportunistic analyses of computed tomography, radiographs, DXA, and magnetic resonance imaging facilitate the estimation of bone density, the detection of previously unrecognized vertebral fractures, and the extraction of biomarkers associated with bone quality, muscle composition, and skeletal geometry. Additionally, recent research demonstrates that end-to-end deep learning models can directly predict future fracture risk from raw images across various imaging modalities.

Imaging-based approaches reveal that clinically relevant fracture risk information is embedded within routine clinical images beyond traditional measurements. These methods have the potential to mitigate gaps in fracture risk assessment and support scalable prevention strategies. Further research is necessary to enhance robustness and facilitate clinical integration.

Figures