Validation of an Atrial Fibrillation detection algorithm
Research type
Research Study
Full title
Stroke Risk Monitoring Service
IRAS ID
277886
Contact name
Oliver Faust
Contact email
Sponsor organisation
Sheffield Hallam University
Duration of Study in the UK
0 years, 3 months, 21 days
Research summary
We have developed a deep learning algorithm that can detect symptoms of atrial fibrillation(AF) in RR interval signals. With benchmark data from publicly accessible databases, the algorithm achieves a detection accuracy of 98%. We have designed a clinical study which aims to validate the detection accuracy and to establish the diagnostic relevance. AF can also contribute to an increased risk of stroke, heart attack and dementia.In addition, AF prevalence increases the risk of stroke five-folds. \n\n\nTherefore, we proposes stroke risk monitoring service that detects AF periods in real time. As such, that system can benefit stroke survivors and patients at risk of stroke during diagnosis and treatment monitoring. As part of this study, we plan to recruit 20 patients who have had stroke or transient ischaemic attack: 10 who are known to have AF, and 10 who are not known to have the disease. We plan to\nmeasure the electrical activity of the patients with two sensors. The Holter monitor will measure electrocardiogram and the Lifetouch sensor will record RR intervals. Once all recordings are collected, an experienced cardiologist or stroke physician will analyse the electrocardiogram signal to establish whether or not AF is present in a specific signal segment. We plan to compare this ground truth with the deep learning results. The study will be considered successful if the accuracy of the deep learning prediction is above 80%.
REC name
South Central - Oxford B Research Ethics Committee
REC reference
20/SC/0320
Date of REC Opinion
4 Jan 2021
REC opinion
Further Information Favourable Opinion