iCAIRD Work package 4: Stroke and AI
Research type
Research Study
Full title
Industrial Centre for Artificial Intelligence Research in Digital Diagnosis (iCAIRD): Stroke Thrombolysis AI Treatment Decision Support Evaluation and Federated Learning
IRAS ID
297671
Contact name
Mary Joan Macleod
Contact email
Sponsor organisation
University of Aberdeen
Clinicaltrials.gov Identifier
DASH 388, Grampian Safe Haven ID (DASH)
Duration of Study in the UK
1 years, 11 months, 30 days
Research summary
Time is critical when making treatment decisions for patients with a suspected stroke. Clot-busting drugs that dissolve blood clots in the brain can potentially restore blood flow to the affected brain region, leading to a good recovery for some patients if given within 4.5 hours of stroke symptom onset. Without careful patient selection, these drugs may cause fatal bleeds or more damage to brain tissue. Clinicians therefore must quickly balance the benefit versus risk of treatment based on the patients’ brain scans and medical history. Decision making in stroke has been made even more complex during the Coronavirus (COVID-19) pandemic not only because of changes in admission patterns, but because of the known and unknown effects of COVID-19 on the brain.
This project aims to train and evaluate two artificial intelligence (AI) applications developed by Canon Medical Research Europe Ltd that will potentially minimise the time required to make treatment decisions. The first AI application will automatically analyse routine brain computerized tomography (CT) images to detect signs that may be associated with a higher chance of bleeding (such as previous bleeds and tumours). The second AI application will automatically search medical records within different databases and electronic documents for information relevant to treatment decisions (such as medical history and blood test results).
The two AI applications are currently being developed and tested on a suspected stroke cohort of 10,000 cases in Greater Glasgow & Clyde (GG&C) using data within the NHS Greater Glasgow & Clyde Safe Haven (NHSGGC). Extending the project to NHS Grampian will allow us to evaluate the performance of the AI application in a new stroke cohort in NHS Grampian, as well as assess federated learning, where an AI application is simultaneously trained with patient data from both Grampian and GG&C – without any patient data leaving its originating centre.REC name
South Central - Berkshire Research Ethics Committee
REC reference
21/SC/0339
Date of REC Opinion
27 Sep 2021
REC opinion
Favourable Opinion