Deep neural networks using information from the 12-lead electrocardiogram (EKG) can predict the risk of developing atrial fibrillation (AF) in patients with no history of AF, a study in Circulation suggests.
In this study, a small team of researchers subsequently identified 1.6 million resting digital 12-lead ECG traces from 430,000 patients, which were recorded in a database from 1984 to 2019. None of the patients had pre-existing or simultaneous atrial fibrillation. The researchers trained deep neural networks to predict emerging atrial fibrillation in these patients within a year. The performance of the deep learning algorithms was assessed using areas below the receiver’s operating curve and the precision recall curve.
The researchers performed an incidence-free survival analysis over a 30 year period after the EKG, which was stratified based on the predictions made by the model. Before 2010, a separate model was trained using all of the EKGs to simulate the real deployment. This performance of this model was assessed using a test set of EKGs linked to a stroke registry, generated from 2010 to 2014.
For the prediction of a newly occurring AF within a 1-year period of an EKG, the respective area under the operating characteristic of the receiver and the area under the precision recall curve were 0.85 and 0.22. The deep neural network models showed a hazard ratio of 7.2 (95% CI, 6.9-7.6) for the predicted high and low risk groups over 30 years.
The model predicted emerging AF after 1 year with a sensitivity and specificity of 69% and 81%, respectively, in a simulated deployment scenario. The researchers found that the model suggested that 9 patients were required to be screened to find 1 new case of AF. In addition, the researchers found that 62% of patients with AF-related stroke were correctly predicted by the model as having a high risk of AF within 3 years of an EKG.
One possible limitation of this study was the use of single-center EKGs, particularly from a predominantly white patient population, suggesting that the results may be limited in their generalizability.
The researchers concluded that the preliminary data from their real-world scenario show that using this tool will identify a high-risk population for emerging AF that can be targeted for enhanced screening and may prove useful to treat AF-related ones Prevent strokes. ”
Raghunath S., Pfeifer JM, Ulloa-Cerna AE, et al. Deep neural networks can use the 12-lead electrocardiogram to predict emerging atrial fibrillation and help identify those at risk of AF-related stroke. Published online February 16, 2021. Edition. doi: 10.1161 / CIRCULATIONAHA.120.047829
This article originally appeared on The Cardiology Advisor