Identification of high-risk imaging features in hypertrophic cardiomyopathy using electrocardiography: A deep-learning approach

Abstract

Background: Patients with hypertrophic cardiomyopathy (HCM) are at risk of sudden death, and individuals with ≥1 major risk markers are considered for primary prevention implantable cardioverter-defibrillators. Guidelines recommend cardiac magnetic resonance (CMR) imaging to identify high-risk imaging features. However, CMR imaging is resource intensive and is not widely accessible worldwide.

Objective: The purpose of this study was to develop electrocardiogram (ECG) deep-learning (DL) models for the identification of patients with HCM and high-risk imaging features.

Methods: Patients with HCM evaluated at Tufts Medical Center (N = 1930; Boston, MA) were used to develop ECG-DL models for the prediction of high-risk imaging features: systolic dysfunction, massive hypertrophy (≥30 mm), apical aneurysm, and extensive late gadolinium enhancement. ECG-DL models were externally validated in a cohort of patients with HCM from the Amrita Hospital HCM Center (N = 233; Kochi, India).

Results: ECG-DL models reliably identified high-risk features (systolic dysfunction, massive hypertrophy, apical aneurysm, and extensive late gadolinium enhancement) during holdout testing (c-statistic 0.72, 0.83, 0.93, and 0.76) and external validation (c-statistic 0.71, 0.76, 0.91, and 0.68). A hypothetical screening strategy using echocardiography combined with ECG-DL-guided selective CMR use demonstrated a sensitivity of 97% for identifying patients with high-risk features while reducing the number of recommended CMRs by 61%. The negative predictive value with this screening strategy for the absence of high-risk features in patients without ECG-DL recommendation for CMR was 99.5%.

Conclusion: In HCM, novel ECG-DL models reliably identified patients with high-risk imaging features while offering the potential to reduce CMR testing requirements in underresourced areas.

Keywords: Cardiac magnetic resonance imaging; Deep-learning; Electrocardiography; Hypertrophic cardiomyopathy; Risk prediction; Sudden cardiac death.

Figure 1:

VAE based ECG-DL model for prediction of high-risk imaging features in HCM patients.

Figure 2:. Saliency analysis.

Representative ECGs are obtained by averaging latent variables for each high-risk (HR) feature and converting to 12-lead, median-beat ECG voltage wave-forms using the pre-trained VAE decoder.