A multimodal retinal aging clock for biological age prediction and systemic health assessment via OCT and fundus imaging.

Prediction performance improved when models were trained on both normal and diseased eyes compared to models trained on normal eyes alone.

• Study 1 assessed how models trained on normal eyes generalize to diseased eyes.
• Study 2 tested whether incorporating disease labels improves performance and systemic associations.
• Incorporating disease labels in training improved model performance for biological age prediction.

Predicted biological age showed significantly stronger correlations with the Charlson Comorbidity Index (CCI) than chronological age across both studies.

• The association was observed across both Study 1 and Study 2.
• Linear regressors were trained on chronological and biological features to infer CCI.
• The CCI is described as a 'validated measure of mortality.'
• The stronger correlation held across both normal and diseased eye training paradigms.

Models were developed to predict biological age from both fundus photography and optical coherence tomography (OCT) using a multimodal approach.

• Models were fine-tuned to the imaging dataset to predict biological age.
• Both fundus photography and OCT were used as input modalities.
• Gradient-weighted regression activation mapping generated heatmaps to identify the model's region of focus.

Gradient-weighted regression activation mapping was used to generate heatmaps identifying which retinal regions the model focused on when predicting biological age.

• Gradient-weighted regression activation mapping (Grad-RAM) was applied to the trained models.
• Heatmaps were generated to provide interpretability of model predictions.
• This approach identified the spatial regions of the retinal images most influential in age prediction.

The retinal aging clock algorithm may provide insight into systemic health burdens beyond that of traditional risk assessments.

• The algorithm's association with CCI supports its potential clinical relevance.
• CCI is described as a validated measure of mortality.
• The authors conclude the algorithm captures systemic health information not fully reflected by chronological age alone.