Median and interquartile ranges (IQR) for Structural Fragility Score-Artificial Intelligence (SFS-AI). Fracture Risk Assessment score (FRAX) with bone mineral density (BMD) and femoral neck BMD in women remaining fracture-free and women having Any Fragility Fractures or Major Fragility Fractures during 5 years follow-up.

Performance of the Structural Fragility Score-Artificial Intelligence (SFS-AI) using data in women of any age and women aged 65 years and over

Left two panels: Receiver Operator Characteristic (ROC) curves for Structural Fragility Score Artificial Intelligence (SFS-AI), Fracture Risk Assessment Score (FRAX) with bone mineral density (BMD) and BMD as a continuous trait predicting for ‘All Fragility Fractures’ and ‘Majority Fragility Fractures’ for women of any age. Area under the Curves (AUCs) with 95% Confidence Intervals (CI) were significant (*p < 0.05) for SFS-AI only. Right two panels: Sensitivity and specificity of SFS-AI, FRAX with BMD and BMD as categorical traits.

Left panels: Advancing age is associated with a higher Structural Fragility Score-Artificial Intelligence (SFS-AI) in women having ‘All Fragility Fractures’ or Major Fragility Fractures (closed circles) and in women remaining fracture-free (open circles). The images of the distal radius and ulna with the heat map illustrate regions commonly encountered in women having fractures.

Left panels. The Structural Fragility Score-Artificial Intelligence (SFS-AI) was associated directly with cortical porosity, FRAX with BMD and negatively with trabecular density and BMD. Right diagram. Of the 47% of explained variance in the SFS-AI, most was attributed to trabecular density, cortical porosity, age and the FRAX with BMD. The contribution of BMD was not significant. The remaining 53 percent remained unexplained.