Systematic Review of Artificial Intelligence–Based ECG Algorithms for Early Detection of Left Ventricular Dysfunction

Resumen

Background: Left ventricular dysfunction (LVD), particularly left ventricular systolic dysfunction (LVSD), is a major precursor to heart failure and is often underdiagnosed due to reliance on imaging modalities such as echocardiography. Artificial intelligence (AI) applied to electrocardiography (ECG) has emerged as a promising non-invasive, cost-effective screening tool for early detection.

Objective: To systematically review the diagnostic performance, clinical utility, and limitations of AI-based ECG algorithms for early detection of LVD.

Methods: A systematic literature review was conducted across major databases including PubMed, Scopus, and Web of Science. Studies evaluating AI-enabled ECG models for detecting LVD were included. Data extracted included study design, population, model type, and performance metrics such as area under the receiver operating characteristic curve (AUROC), sensitivity, and specificity.

Results: AI-ECG algorithms demonstrated strong diagnostic performance across multiple studies, with reported sensitivities up to 95.6% and high negative predictive values . Deep learning models were capable of detecting subclinical LVSD up to two years before clinical diagnosis . Both single-lead and 12-lead ECG-based models showed promising results, although most studies relied on retrospective datasets. Challenges identified included limited external validation, variability in datasets, and lack of interpretability.

Conclusion: AI-enabled ECG represents a transformative approach for early LVD detection, with potential to improve screening and risk stratification. However, further prospective validation and integration into clinical workflows are required.