The Project

A stroke is a sudden loss of oxygen to the brain, caused by a clot blocking the blood supply or a bleed within the brain itself.

Stroke is the leading cause of disability in the UK and worldwide, but research suggests:
Around 3 in 10 strokes are unexplained.
Around 1 in 10 of strokes could be prevented
if known risk factors for stroke were treated.

We know that in many cases, the information necessary to predict a future stroke is present in medical tests undertaken for other reasons, however, often healthcare professionals do not have the time to look for and act on this information. Moreover, around 30% of the strokes cannot be explained by known risk factors for stroke.

We believe that Artificial Intelligence technologies may allow us to automatically detect those at higher risk of stroke, by identifying patterns within these investigations which predict a stroke. These patterns may allow us to identify new and known risk factors. We hope to create an artificial intelligence ‘model’ which will analyse health records and and medical tests, to create a tailored, personal assessment of stroke risk. This will allow treatments which reduce this risk to be started and, we hope, new treatments to be developed.

2024 February

PHASE I - model training

Funded by a grant from the UKRI Medical Research Council, we are currently building a database of anonymised information drawn from health records and medical tests libraries of University Hospitals Plymouth NHS Trust. This will comprise 10,000 people who have had a stroke and up to 110,000 people who have not. We think this will allow us to generate sophisticated estimates of the future risk of stroke at 1,3, 5 and 10 years. Using techniques known as ‘explainable AI’, we also hope to be able to identify new markers of future stroke, which will may enable new treatments to be developed

2027 (planned)

PHASE II - External Validation

In order to ensure the predictions our model is making are accurate, we plan to test whether it can accurately predict stroke in populations representing the diversity of the United Kingdom. This will involve building new datasets extracted from other UK Healthcare providers and applying the model to them to see if it works.

2029 (planned)


Having engaged with stroke survivors, carers, members of the public and healthcare professionals to understand the best way to communicate the predictions made in the model, we hope to use them to target personalized treatments to reduce the risk of future stroke.