Leveraging topological data analysis to estimate bone strength from micro-CT as a surrogate for advanced imaging

TL;DR

This paper introduces a novel application of topological data analysis to high-resolution bone images, improving bone strength prediction and potentially enhancing osteoporosis risk assessment.

Contribution

The study demonstrates that topological features derived from persistent homology outperform traditional morphometrics in predicting bone strength from micro-CT images.

Findings

Topological features improve bone strength prediction accuracy.

Internal voids are highly predictive of biomechanical strength.

TDA-based models outperform traditional morphometric models.

Abstract

Accurate bone strength prediction is essential for assessing fracture risk, particularly in aging populations and individuals with osteoporosis. Bone imaging has evolved from X-rays and DXA to clinical computed tomography (CT), and now to advanced modalities such as high-resolution peripheral quantitative CT and synchrotron radiation CT, which offer unprecedented resolution of bone microarchitecture. However, analytical methods have not kept pace with these imaging advances. This study applied topological data analysis (TDA) to extract biomechanically relevant features from high-resolution bone images, offering a new framework for bone strength prediction. We extracted topological features, specifically those derived from persistent homology, and combined them with standard bone morphometric descriptors to train machine learning models for apparent strength prediction. Models based solely on topological features outperformed those using traditional morphometrics, highlighting TDA's ability to capture biomechanically relevant structure. In particular, internal voids, often dismissed as imaging noise, proved to be the most predictive. While limited by dataset size and class imbalance, these results suggest that TDA offers a promising approach for advancing osteoporosis risk assessment.