Beyond Conventional Retinal Biomarkers: Mining the Human Retina

Delivering Precision Medicine in Diabetes using Big Data: A global health Initiative

The NIHR Unit on Global Diabetes Outcomes Research at the University of Dundee seeks ten PhD Students with statistical and computing background. Students will be registered with the University of Dundee and will be jointly supervised by academics from the Madras Diabetes Research Foundation in Chennai.

This project aims to develop VAMPIRE research on biomarkers for risk assessment of high-incidence conditions. Traditional retinal biomarker featuring in the literature involve morphometric measurements of the retinal vasculature, and less frequently lesions. However there is scope for investigating further and non-semantic (i.e. without a clear clinical or anatomical meaning) features as possible biomarkers, and compare their predictive value with that of conventional measures.

This strongly interdisciplinary project involves the following;

• Familiarization with VAMPIRE software and tools, with assistance from the VAMPIRE research team in Dundee and in Edinburgh.
• Investigation of novel representations for retinal measurements, e.g. spatial distributions, and where possible spatiotemporal distrivutions, as opposed to local values in specific, limited locations or regions.
• Investigation of computer vision features used in contemporary classification and image mining research, e.g. in the context of deep learning.
• Evaluation of the above as predictors of risk or indicators of disease progression, using the retinal image sets made available to the project.
• Assessment of the relative predictive power of semantic and non-semantic features, and their complementarity.

The project builds on the solid platform of the VAMPIRE research program on multi-modal retinal image analysis for biomarkers discovery, including recent work on eye disease stratification via assessment of density and tortuosity of corneal nerve fibres using
machine learning.

The student will join the Computer Vision and Image Processing team (SSE, Computing) who has been working for 8 years in close collaboration with researchers and clinicians at Ninewells NHS and University Hospital (consistently rated one the best medical school for research and teaching in the UK).

The student will access images of the retinal vascular network observed by conventional as well as portable fundus cameras, and possibly scanning laser ophthalmoscopes and the novel OCT-angiography. The student will become part of the core VAMPIRE research team, a joint initiative of the universities of Dundee and Edinburgh.