Back
Science

AI-derived retinal age score predicts osteoporosis risk in two large cohorts

View source

Retinal ‘Age’ Scan Shows Promise as Low-Cost Osteoporosis Screening Tool

A new study published in PLOS Digital Health suggests that a simple, non-invasive retinal scan could help identify individuals at risk for osteoporosis.

The research details an AI-derived score called RetiAGE, which estimates the biological age of the retina. The model was trained on a vast dataset of 129,236 retinal images and then tested on two large cohorts: the Singaporean PIONEER group (n=1,965) and the UK Biobank (n=43,938).

Key Findings:

  • An inverse relationship was observed: Higher retinal age was significantly linked to lower bone mineral density.
  • Predictive power: For every standard deviation increase in retinal age, the risk of developing osteoporosis over a mean follow-up of 12.2 years rose by 12% (hazard ratio 1.12).
  • Improved screening: Adding the RetiAGE score to the standard Osteoporosis Self-assessment Tool substantially improved its predictive accuracy, boosting the C-index from 0.585 to 0.635.

"Retinal imaging could serve as a low-cost, non-invasive opportunistic screening tool for osteoporosis," the authors suggest.

However, they also caution that the model may require calibration for different populations and clinical workflows before widespread implementation.