Using deep learning to explore the predictive power of 180,000 eye images for dementia outcomes

This project will examine the ability of non-invasive eye imaging to predict dementia-related outcomes, cognitive tests and structural changes in MRI images. To do this we will use the 180,000 eye images in UK Biobank, one of the world’s largest eye-imaging studies.

In contrast with the existing literature, we will use deep learning methodologies to improve the ability to extract information from structural eye imaging. We will focus on prediction of a range of cognition outcomes, reflecting real world usage. The UK Biobank also contains brain MRI imaging. We can use this to explore the ability of structural changes in the eye to predict structural changes in the brain.