Automated screening for paroxysmal atrial fibrillation from sinus rhythm ECGs.
Principal Investigator: Dr James Fraser
Approved Research ID: 35452
Approval date: April 24th 2018
We aim to develop a low-cost automated screening test for paroxysmal atrial fibrillation (PAF), a common, treatable disease associated with stroke. This will allow detection of PAF sufferers from their sinus rhythm ECGs. Our proposed research follows from preliminary work in a small clinical study that demonstrated a change in the complexity of sinus rhythm ECGs in PAF sufferers. The UK Biobank has a unique record of resting and exercise ECGs and cardiovascular diagnoses that will now allow us to develop and test an algorithm that is suitable for screening for PAF in the normal population. Our research aligns with the UK Biobank?s purpose by aiming to improve the diagnosis of a common, serious but underdiagnosed condition. Paroxysmal atrial fibrillation (PAF) increases the risk of ischaemic stroke roughly five-fold but around half of cases are undiagnosed. Cases are frequently missed because PAF often occurs in short episodes interposed with long periods of normal sinus rhythm, and can be symptomless. Our work aims to prevent potentially life-threatening consequences of PAF by allowing many more sufferers to receive appropriate anti-clotting therapy. Resting ECGs from the UK Biobank will be analysed using a new algorithm. In a small study, we have shown that this algorithm can detect an abnormal heart rhythm called paroxysmal atrial fibrillation, even if the ECG appears normal. We now need to test this in many more subjects and assess whether other cardiovascular conditions also influence ECG complexity. We will begin by analysing half of the resting ECGs within the UK Biobank (~2500 records) to learn how complexity measures correlate with the cardiovascular diagnoses of the same subjects. We will then blind-test this tool against the remaining ECGs. The study will initially look at all subjects with a resting 12-lead ECG (5082 records). It will then look at a 10,000 patient subset of the exercise ECG records to explore whether exercise ECGs might be more predictive of paroxysmal atrial fibrillation than resting ECGs.