Project Summary:
This project aims to develop AI-based tools for early detection and risk assessment of ocular and systemic diseases, with future potential for commercial translation. We will integrate UK Biobank ocular imaging data with proprietary datasets of retinal fundus photographs and OCT scans to train scalable, generalizable AI models. This combined data will enhance model robustness, enable discovery of novel imaging biomarkers, and support clinical application.
Research Questions:
How can deep learning and explainable AI improve sensitivity, specificity, and interpretability across diverse subgroups?
Which retinal features predict systemic diseases such as cardiovascular conditions?
How can longitudinal data support tracking disease progression?
Objectives:
Develop and validate deep learning models using multi-source retinal images.
Identify retinal biomarkers linked to systemic health.
Evaluate model generalizability, fairness, and clinical readiness.
Enable scalable deployment in public health settings, including entry-level facilities in low-resource regions.
Scientific Rationale:
The retina offers a non-invasive view into systemic health, reflecting early microvascular and neurodegenerative changes. Combining UK Biobank data with curated internal datasets will support the development of clinically relevant AI models. The focus is on accessible, affordable tools for early detection, especially suited for public health programs in developing countries. Findings will inform solutions aligned with regulatory requirements and real-world clinical validation.