Log in to MyACC
Machine Learning Prediction of New Atrial Fibrillation Based on ECG
Quick Takes
- A deep learning model can identify patients at high risk for new-onset atrial fibrillation (AF).
- In patients with no history of AF who have an AF-related stroke, nearly two thirds would have been predicted to be high-risk for AF before the stroke by the deep learning model.
- A deep learning model capable of predicting future AF could be used in conjunction with a systematic monitoring strategy to find AF early and potentially prevent AF-related stroke.
Study Questions:
Can deep neural network predict new-onset atrial fibrillation (AF) from the resting 12-lead electrocardiogram (ECG), and does this prediction help identify those at risk of AF-related stroke?
Methods:
A total of 1.6 M resting 12-lead digital ECG traces from 430,000 patients were analyzed. Deep neural networks were trained to predict new-onset AF (within 1 year) in patients without a history of AF. The authors then simulated a deployment scenario of this model retrospectively to demonstrate its potential to identify patients who later have an AF-related stroke.
Results:
The area under the receiver operating characteristic curve and area under the precision-recall curve were 0.85 and 0.22, respectively, for predicting new-onset AF within 1 year of an ECG. The hazard ratio for the predicted high- versus low-risk groups over a 30-year span was 7.2 (95% confidence interval, 6.9–7.6). In a simulated deployment scenario, the model predicted new-onset AF at 1 year with a sensitivity of 69% and specificity of 81%. The number needed to screen to find one new case of AF was 9. This model predicted patients at high risk for new-onset AF in 62% of all patients who experienced an AF-related stroke within 3 years of the index ECG.
Conclusions:
Deep learning can predict new-onset AF from the 12-lead ECG in patients with no previous history of AF. This prediction may help identify patients at risk for AF-related strokes.
Perspective:
AF is underdiagnosed because AF is often minimally symptomatic or asymptomatic, which leads to underutilization of anticoagulation. In some patient populations, opportunistic pulse and ECG screening have been advocated. Given the relatively low incidence of the disease, and often times its paroxysmal nature, a cost-effective approach to identifying patients with AF and at high risk of stroke has been challenging to establish. The authors report a remarkable potential of deep neural networks to predict AF within a year of the sinus rhythm ECG. If reproducible, this approach could inform targeted screening of a select group with ambulatory ECG monitors to establish the diagnosis of AF with greater cost-effectiveness and facilitate an appropriate anticoagulation strategy.
Clinical Topics: Anticoagulation Management, Arrhythmias and Clinical EP, Prevention, Anticoagulation Management and Atrial Fibrillation, Implantable Devices, SCD/Ventricular Arrhythmias, Atrial Fibrillation/Supraventricular Arrhythmias
Keywords: Anticoagulants, Arrhythmias, Cardiac, Artificial Intelligence, Atrial Fibrillation, Cost-Benefit Analysis, Electrocardiography, Electrocardiography, Ambulatory, Learning, Neural Networks, Computer, Risk Assessment, Secondary Prevention, Stroke, Vascular Diseases
< Back to Listings
