Daysy – Wearable Devices
May 1, 2023New medicine, thanks to AI
May 1, 2023Article of the Month – May 2023
Deep Neural Networks Can Predict New-Onset Atrial Fibrillation From the 12-Lead ECG and Help Identify Those at Risk of Atrial Fibrillation–Related Stroke...
Deep Neural Networks Can Predict New-Onset Atrial Fibrillation From the 12-Lead ECG and Help Identify Those at Risk of Atrial Fibrillation–Related Stroke
Sushravya Raghunath, John M. Pfeifer, Alvaro E. Ulloa-Cerna, Arun Nemani, Tanner Carbonati, Linyuan Jing, David P. vanMaanen, Dustin N. Hartzel, Jeffrey A. Ruhl, Braxton F. Lagerman, Daniel B. Rocha, Nathan J. Stoudt, Gargi Schneider, Kipp W. Johnson, Noah Zimmerman, Joseph B. Leader, H. Lester Kirchner, Christoph J. Griessenauer, Ashraf Hafez, Christopher W. Good, Brandon K. Fornwalt, and Christopher M. Haggerty
Background
Atrial fibrillation (AF) is associated with substantial morbidity, especially when it goes undetected. If new-onset AF could be predicted, targeted screening could be used to find it early. We hypothesized that a deep neural network could predict new-onset AF from the resting 12-lead ECG and that this prediction may help identify those at risk of AF-related stroke.
Methods
We used 1.6 M resting 12-lead digital ECG traces from 430 000 patients collected from 1984 to 2019. Deep neural networks were trained to predict new-onset AF (within one year) in patients without a history of AF. Performance was evaluated using areas under the receiver operating characteristic and precision-recall curves. We performed an incidence-free survival analysis for a period of 30 years following the ECG stratified by model predictions. To simulate real-world deployment, we trained a separate model using all ECGs before 2010 and evaluated model performance on a test set of ECGs from 2010 through 2014 that were linked to our stroke registry. We identified the patients at risk for AF-related stroke among those predicted to be high risk for AF by the model at different prediction thresholds.
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 one year of an ECG. The hazard ratio for the predicted high- versus low-risk groups over a 30-year span was 7.2 (95% CI, 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 three 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.
Relevance to Healthcare Field
Atrial fibrillation (AF) is the most common sustained rhythm disturbance and its prevalence increases as the population ages. It is associated with two non-favorable outcomes: stroke and heart failure. Although it is associated with heart disease, many patients with no history of heart disease develop AF. This study is focused on predicting new-onset AF in asymptomatic patients to prevent AF-related complications (strokes). They performed two separate experiments, the proof-of-concept model and simulated deployment model. The proof-of-concept model used a deep neural network (DNN) model designed to analyze the electrocardiogram (ECG) signals to generate a predicted risk score for new-onset AF within one year of the ECG. Two versions of this model design were compared, with and without age and sex (DNN-ECG and DNN-ECG-AS). Both models were evaluated by the area under the receiving operating characteristic curve (AUROC), where higher AUROC signifies higher performance. They implemented an extreme gradient boosting model (XGBoost) using only sex and age for comparison against the DNN models. Secondly, the simulated deployment model used the DNN-ECG and DNN-ECG-AS to determine who would have an AF-related stroke. It was shown that this DNN (AUROC, 0.85) outperformed already existing models within this dataset, CHARGE – AF (AUROC, 0.77) and XGBoost, but also other models in previous studies: atherosclerosis risk in communities (ARIC) (AUROC, 0.78) and Framingham heart study (AUROC, 0.78). With this DNN, we can predict AF and treat the patients to prevent AF-related complications beforehand and prolong the survival rate.