PhD: Retinal Biomarkers

PhD project on retinal biomarkers, infections and vaccines to predict dementia risk using causal modelling and large scale health and imaging datasets.

Title: Retinal Biomarkers, Respiratory Infections, and Vaccinations: Causal Links to Dementia Risk and Prevention Strategies

Synopsis: This project seeks to establish causal relationships between respiratory infections, vaccinations, and dementia risk by analysing retinal biomarkers, aiming to identify modifiable factors for dementia prevention and informing public health strategies to reduce neurodegenerative disease incidence.

Background: Dementia represents a major global health challenge, with projections exceeding 150 million affected individuals by 2050. Recent evidence underscores the potential of respiratory infections such as COVID-19, respiratory syncytial virus (RSV), and influenza in elevating dementia risk. Concurrently, there is mounting evidence that vaccinations against these infections, including AS01-adjuvanted RSV vaccines, may reduce dementia incidence. Retinal imaging offers a non-invasive method to observe neurovascular changes that are indicative of neurodegeneration. The supervisory team has access to SCONe, a unique dataset of 1.5M retinal images collected in optometry practices throughout Scotland over the past 20 years linked to national healthcare records including hospital, prescription, and death records. No other comparable dataset exists globally. The supervisory team includes world experts in retinal biomarkers of dementia, epidemiology of respiratory infections, and causal inference.

Objective: This project aims to use causal inference techniques to elucidate the relationship between respiratory infections, vaccinations, and dementia risk by employing retinal biomarkers. By integrating retinal imaging with comprehensive healthcare data on infections and vaccinations for the same cohort, we will explore causal pathways and potentially modifiable factors in dementia progression.

Methodology 

1. Data Integration: We will utilise retinal images and detailed healthcare records—including respiratory infection episodes and vaccination status—collected from the same individuals. The cohort will comprise patients with both pre- and post-infection/vaccination imaging for whom we have a comprehensive record of retinal images over two decades.
2. Retinal Imaging Analysis: We will employ colour fundus photographs (CFPs) to quantify structural and vascular changes in the retina that have been shown to describe dementia risk. Primary biomarkers will include validated approaches such as fractal dimension analysis of retinal vessels, retinal age gap, and our newly developed method for estimating peripapillary retinal nerve fibre layer (RNFL) and macular ganglion cell–inner plexiform layer (GC-IPL) thinning in CFPs.
3. Infection and Vaccination Assessment: We will analyse comprehensive healthcare data to establish incidence and severity of infections. Vaccination data will be utilised to explore temporal relationships with retinal changes and potential protective effects against neurodegeneration.
4. Causal Modelling: We wish to attribute the causal impact of exposure to infections/vaccines on the outcome of developing another dementia. At the same time, we have semi-regular analysis of risk of developing the disease through time-to-event predictors. This allows us to model disease progression in a longitudinal fashion. The proposed meta-risk model is an extension of a latent-state dynamical system with regular observations of (1) covariates which are known to effect the disease, (2) retina-based risk predictors (as a proxy to disease state) and (3) knowledge of when the individual is exposed to the infection/vaccine of interest. This framework leverages statistical strength from observational data to allow us to determine causal effects of the exposure on the disease progression.

Expected Outcomes 

• Causal Links: Determination of causal pathways linking respiratory infections and vaccinations to changes in dementia risk, mediated by retinal biomarkers.
• Predictive Models: Validated models predicting dementia risk based on retinal imaging, infection history, and vaccination status.
• Public Health Insights: Evidence supporting the protective effect of vaccinations against dementia, potentially informing vaccination policies for neuroprotection.

Potential impact: The project will advance understanding of infection-related dementia risk and validate retinal imaging as a diagnostic tool, potentially informing public health policies on vaccination strategies to reduce neurodegenerative disease incidence, thereby improving population

Training: The student will receive comprehensive training across multiple scientific domains, focusing on image analysis, causal modelling, and health data science, all within the context of neuroscience and brain disease. This interdisciplinary approach supports the development of a “T-shaped” researcher, combining specialised expertise with the versatility to collaborate across fields.

Core training elements include state-of-the-art techniques in image analysis for interpreting retinal imaging data, alongside advanced health data modelling skills to understand causal relationships between infections, vaccinations, and dementia risk. Data science training will provide proficiency in managing and analysing large datasets, essential for deriving key insights.

Throughout the project, the student will gain career development opportunities by engaging in collaborative research and presenting at conferences, enhancing communication skills and professional networks. Additional workshops on research ethics, grant writing, and intellectual property will prepare the student

for diverse career paths. This training framework aims to equip the student with the skills necessary to address complex challenges in brain disease research.

Recruitment: A good undergraduate degree in a STEM discipline with postgraduate training in health data science or Artificial Intelligence. Evidence of having undertaken original research in health data science or related topic.

Apply: All applications must be submitted through the Future Medicine PhD fellowships website.