Computer modelling of the retinal microvasculature: a move towards personalised management of kidney disease and brain health through improved understanding of vascular dysfunction and injury

Aims

To extract superior information about the state of the microvasculature at the back of the eye than has previously been achievable and use this to investigate: (1) detecting and tracking of disease in the kidney and brain, (2) monitoring response to treatment, and (3) predicting patient outcomes.

The project will utilise recent technological advancements from our group in processing retinal images as the basis for building novel 3D patient-specific computer models of the microvasculature. It will leverage existing longitudinal multi-modal retinal imaging linked to clinical data that we have previously collected. This includes healthy volunteers (n=250), those with CKD (n=500) and those with CKD who have received a kidney transplant and therefore their kidney function has returned to a healthy level (n=250). Clinical biomarkers of CKD (estimated glomerular filtration rate [eGFR] and proteinuria) are also available. Additionally, a mid-life cohort (n=180) currently returning for a third wave of tests and who are currently in good health but with low/medium/high risk of dementia in later life based on ApoE status, family history, and cardiovascular risk factors. Plus, MRI derived measurements of cSVD (i.e., lacunes, microbleeds, white matter hyperintensities, and enlarged perivascular spaces) and cognitive tests.

The student will develop the computational processes for re-creating the microvascular anatomy contained within image data for a person and then use these detailed models to examine geometric differences (that might be missed by conventional retinal image analysis) between patient groups as well as longitudinal changes for individuals in response to disease progression (i.e., worsening eGFR/proteinuria; increase in MRI markers of cSVD) and treatment (e.g., therapies to reduce systemic and renal inflammation or vasoactive medications). They will also experiment with deriving boundary conditions and simulating blood flow velocities and vessel wall dynamics to generate a physiological dimension to the computer modelling and investigate what additional information this reveals about changes to microvascular health in relation to CKD and cSVD. The student will explore trajectories of change in the retinas and use machine learning classification techniques (e.g., logistic regression, decision trees, deep neural networks) to attempt the stratification of patients/groups. Finally, the project will look at predicting outcomes such as associated cardiovascular disease events (i.e., myocardial infarction, stroke, heart failure and cardiovascular mortality) and deteriorating brain health (e.g., increasing lesion volume on MRI, decreasing cognitive scores) to contribute to future personalised risk-based intervention.