Diagnostic accuracy of artificial-intelligence-based electrocardiogram algorithm to estimate heart failure with reduced ejection fraction: A systematic review and meta-analysis

Abstract

Introduction

AI-based ECG has shown good accuracy in diagnosing heart failure. However, due to the heterogeneity of studies regarding cutoff points, its precision for specifically detecting heart failure with left ventricle reduced ejection fraction (LVEF <40 %) is not yet well established. What is the sensitivity and specificity of artificial-based electrocardiogram to diagnose heart failure with low ejection fraction (cut-off of 40 %. Aims: We conducted a meta-analysis and systematic review to evaluate the accuracy of artificial intelligence electrocardiograms in estimating an ejection fraction below 40 %.

Methods

We searched PubMed, Embase, and Cochrane Library for studies evaluating the performance of AI ECGs in diagnosing heart failure with reduced ejection fraction. We computed true positives, true negatives, false positives, and false negatives events to estimate pooled sensitivity, specificity, and area under the curve, using R software version 4.3.1, under a random-effects model.

Results

We identified 9 studies, including patients with a paired artificial intelligence-enabled electrocardiogram with an echocardiography. patients had an ejection fraction below 40 % according to the echocardiogram. The AI-ECG data yielded areas under the receiver operator of, the sensitivity of), specificity of, and area under the curve of. The mean/median age ranged from 60±9 to 68.05± 11.9 years.

Conclusions

In this systematic review and meta-analysis, the use of electrocardiogram-based artificial intelligence models demonstrated high sensitivity and specificity to estimate a left ventricular ejection fraction below 40 %.