Federated learning for eye health

Our project is focused on improving early detection and diagnosis of eye and neurological conditions by applying artificial intelligence to real-world clinical data. By partnering with optometry clinics in both rural and urban areas, we aim to collect and standardize a wide variety of vision-related data, including clinical observations, health records, and imaging of the eye, into a common framework that enables secure, privacy-respecting research and model development.

This work addresses a growing challenge in healthcare: how to make use of fragmented, often unstructured data collected in everyday clinical settings. These data hold valuable insights but are difficult to analyze at scale using traditional methods. By applying machine learning and data harmonization techniques, our platform will help uncover patterns that may be linked to early signs of eye diseases like glaucoma or diabetic retinopathy, or to broader neurological issues affecting eye movement and visual processing.

All data used in the project is de-identified and securely managed through a cloud-based system that complies with national privacy standards. We use federated learning, an approach that allows models to be trained across multiple clinics without sharing raw patient data, to ensure both data privacy and equitable representation.

By the end of the project, we aim to develop AI-driven tools that can support clinicians in real time, reduce diagnostic delays, and improve access to care across diverse populations. This initiative contributes to the broader goals of precision medicine by demonstrating how responsible AI and data sharing can lead to smarter, safer, and more inclusive healthcare.