Can artificial intelligence highlight those at greater risk of stroke?

Strokes represent one of the leading causes of death and disability in the UK, impacting around 100,000 patients each year.

However, with one in seven strokes seen as preventable, a new project aims to enhance the ability to predict whether a person is at an increased risk of stroke.

Funded by a £300,000 grant from the Medical Research Council, the ABSTRACT project is being led by experts at the University of Plymouth, University Hospitals Plymouth NHS Trust and the University of Exeter. They will also be working with two commercial providers of medical investigations, Express Diagnostics and Ultracardiac.

It will assess past brain scans and other medical test results of stroke survivors, and aim to establish if there are patterns which could have identified them as being at higher risk of stroke.

The researchers will then look to develop a series of artificial intelligence models that can predict whether someone is at greater risk of experiencing a stroke at any point over the next decade.

With the first five years of care post stroke costing the NHS around £3.6billion, and 13.7% of strokes regarded as being preventable, the project team hopes its work will not only improve lives but also prove cost effective at a time when the health system is under greater financial pressure than ever.

Dr Stephen Mullin , Associate Professor in Neurology in the University of Plymouth’s Peninsula Medical School and Consultant Neurologist at University Hospitals Plymouth NHS Trust, is the project’s principal investigator.

The project, funded by the Medical Research Council, brings together researchers with expertise in health data science, data governance, statistics, radiology/imaging, and neurology.

It will initially build a database of results from patients seen at University Hospitals Plymouth NHS Trust, and benefits from the support and collaboration of the national Sentinel Stroke National Audit Programme (SSNAP) and NIHR Applied Research Collaboration South West Peninsula (PenARC).

This database will include the results of magnetic resonance imaging (MRI) and computer tomography (CT) brain scans, electrocardiograms (ECG) and echocardiograms (ECGs), standard tests currently carried out when a person is suspected of having had a stroke.

It will be used to train an artificial intelligence computer model, which the researchers hope can predict who will later develop strokes based on patterns within the data collected.

The work will leverage state of the art techniques with artificial intelligence, collectively known as explainability. These tools allow visualisation and identification of the factors driving predictions. This allows researchers to be certain that the predictions being made are accurate and hopefully will identify new factors which contribute to the risk of developing a stroke.

An opportunity will be given for patients at UHPNT to opt out of use of their data in this project, but the research team is currently compiling the database which will be used in the analysis.