Introduction: Identification of individuals at risk for heart failure (HF) can lead to early medical intervention. However, large scale screening for incident HF in the general population is insufficient, partly due to the limited capacities in health care systems and number of health care specialists. The application of deep learning algorithms on 1-lead ECGs can be a means to overcome this, especially as wearables capable of deriving this biomarker are becoming more popular in the population. This holds the potential to identify individuals at risk who could benefit from a referral to a HF specialist.

Purpose: We examined the ability to predict three-year HF incidence from single-lead ECGs using a deep learning model that was trained on data from the UK Biobank.

Methods: The UK Biobank was established in 2006 and contains data from 500,000 participants from the British general population. 100,551 1-lead ECGs from 91,898 unique participants were used for this study. This includes 57,897 3-lead resting ECGs written during ergometry exams between 2006 and 2010 (Exam 1; median age 56.7 years, 46.2% men) and 9,826 3-lead resting ECGs at the first follow-up from 2010-2013 (Exam 2; median age 61.3 years, 49% men) and 32,828 12-lead resting ECGs which were acquired between 2014 and 2020 (Exam 3; median age 63.5 years, 48.4% men). Lead I was extracted from the ECGs for all analysis. A 1D-Convolutional Neural Network was used for prediction of the 3-year risk of incident HF. A comparison with the CHARGE-AF score was performed. A nested 6-fold cross-validation scheme was employed.

Results: In all three exams, the Neural Network yielded good detection of 3-year incident HF (AUC 0.738, 0.822, 0.738). CHARGE-AF provided slightly better results at Exam 1 and Exam 3 (AUC 0.766, 0.748), but worse results in Exam 2 (AUC 0.736). Combining both models led to increased discriminatory properties in all exams (AUC 0.812, 0.832, 0.788).

Conclusion: A deep learning model to predict incident HF from 1-lead ECGs shows high discriminatory ability, which can be further elevated when combining with the CHARGE-AF score. This study represents a further step towards the applications of deep learning-based analysis of the ECG to advance health care.