Estimating chronological age from flicker-induced retinal vascular reactivity: a machine learning approach

This is the authors' abstract. We don't add key points for this paper.

This study proposes a novel, non-invasive approach to chronological age (CA) estimation by leveraging the Extra Trees Regressor to analyze retinal vessel reactivity to flicker light stimulation as a unique biomarker. By capturing and interpreting the dynamic responses of retinal blood vessels, we aim to develop an accessible and reliable tool for assessing chronological age, contributing to the emerging field of personalized anti-aging medicine through individualized insights into aging. Our method focuses on vascular function analysis using non-invasive dynamic vessel response measurements, offering a unique contribution compared to existing techniques. In this paper, multiple machine learning and deep learning models are employed to conduct experiments that estimate patient age using retinal arterial and venous vessel diameter changes across three flicker cycles. The best-performing model achieved a mean absolute error (MAE) of 9.3, a root mean squared error (RMSE) of 12.4, and a Mean Absolute Percentage Error (MAPE) value of 17.7%, demonstrating acceptable reliability and accuracy. Additionally, the application of Local Interpretable Model-agnostic Explanations (LIME) to our time series age estimation model provided valuable insights into the key features influencing predictions. To our knowledge, this is the first study to demonstrate that retinal vascular reactivity can serve as a functional biomarker for estimating chronological age, providing a foundation for future research on non-invasive markers of aging.