Dataset Information

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.

ABSTRACT:

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 1 year) in patients without a history of AF. Performance was evaluated using areas under the receiver operating characteristic curve and precision-recall curve. 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 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% 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 1 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.

SUBMITTER: Raghunath S

PROVIDER: S-EPMC7996054 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:CONTEXT:Although the heritability of atrial fibrillation (AF) is established, the contribution of familial AF to predicting new-onset AF remains unknown. OBJECTIVE:To determine whether familial occurrence of AF is associated with new-onset AF beyond established risk factors. DESIGN, SETTING, AND PARTICIPANTS:The Framingham Heart Study, a prospective community-based cohort study started in 1948. Original and Offspring Cohort participants were aged at least 30 years, were free of AF at the baseline examination, and had at least 1 parent or sibling enrolled in the study. The 4421 participants in this analysis (mean age, 54 [SD, 13] years; 54% women) were followed up through December 31, 2007. MAIN OUTCOME MEASURES:Incremental predictive value of incorporating different features of familial AF (any familial AF, premature familial AF [onset ?65 years old], number of affected relatives, and youngest age of onset in a relative) into a risk model for new-onset AF. RESULTS:Across 11,971 examinations during the period 1968-2007, 440 participants developed AF. Familial AF occurred among 1185 participants (26.8%) and premature familial AF occurred among 351 participants (7.9%). Atrial fibrillation occurred more frequently among participants with familial AF than without familial AF (unadjusted absolute event rates of 5.8% and 3.1%, respectively). The association was not attenuated by adjustment for AF risk factors (multivariable-adjusted hazard ratio, 1.40; 95% confidence interval [CI], 1.13-1.74) or reported AF-related genetic variants. Among the different features of familial AF examined, premature familial AF was associated with improved discrimination beyond traditional risk factors to the greatest extent (traditional risk factors, C statistic, 0.842 [95% CI, 0.826-0.858]; premature familial AF, C statistic, 0.846 [95% CI, 0.831-0.862]; P = .004). Modest changes in integrated discrimination improvement were observed with premature familial AF (2.1%). Net reclassification improvement (assessed using 8-year risk thresholds of <5%, 5%-10%, and >10%) did not change significantly with premature familial AF (index statistic, 0.011; 95% CI, -0.021 to 0.042; P = .51), although categoryless net reclassification was improved (index statistic, 0.127; 95% CI, 0.064-0.189; P = .009). CONCLUSIONS:In this cohort, familial AF was associated with an increased risk of AF that was not attenuated by adjustment for AF risk factors including genetic variants. Assessment of premature familial AF was associated with a very slight increase in predictive accuracy compared with traditional risk factors.