Truace
Subscribe

AI reading heart tracings spots serious heart attacks more reliably than clinicians in pooled studies

Study title: Diagnostic Accuracy of Artificial Intelligence for Detecting ST-Elevation Myocardial Infarction on 12-Lead Electrocardiograms: A Systematic Review and Meta-Analysis

Researchers pooled 26 studies covering 539,876 electrocardiograms to see how well AI identifies ST-elevation myocardial infarction, a serious type of heart attack, from standard 12-lead tracings. They searched six databases covering January 2010 to March 10, 2025, and analysed three separate questions rather than one combined result.

Read the study abstract

Artificial intelligence (AI) may improve electrocardiographic detection of ST-elevation myocardial infarction (STEMI), but available studies address distinct diagnostic questions. We systematically reviewed studies evaluating AI-based STEMI detection from 12-lead electrocardiograms and separately synthesized AI performance against a clinical reference standard, AI and clinician performance against a clinical reference, and AI agreement with physician electrocardiogram (ECG) interpretation. PubMed, Embase, Web of…

By the Truvace desk1 min read1 sourceEvidence: Moderate · 70/100 — Systematic review or meta-analysis · n = 539,876 · 1 independent sourcePeer-reviewedTRV-2026-2312· cite & verify
Diagnostic Accuracy of Artificial Intelligence for Detecting ST-Elevation Myocardial Infarction on 12-Lead Electrocardiograms: A Systematic Review and Meta-Analysis
A key to the electrocardiogram, by Bishop, Louis Faugeres, 1864-1941. Public domain

In brief

Faster and more sensitive ECG reading could matter for a condition where timely recognition affects treatment. However, the evidence is split across different comparison types, and the abstract reports only retrospective accuracy figures, so it does not show whether using AI changes patient outcomes in practice.

Main points

  1. Twenty studies compared AI with a clinical reference standard: pooled sensitivity 0.95, specificity 0.95, AUC 0.98.
  2. In 8 studies with both AI and clinicians, AI sensitivity was 0.91 versus 0.78 for clinicians, with specificity 0.91 for both.
  3. In 7 studies using physician ECG interpretation as the comparator, AI sensitivity was 0.85 and specificity 0.99.
  4. Sensitivity analyses excluding case-control designs and influential studies gave similar findings.

The gain

Across 8 studies comparing both against a clinical reference, AI had higher sensitivity for STEMI than clinicians (0.91 versus 0.78) at equal specificity of 0.91.

The rundown

The review searched PubMed, Embase, Web of Science, Scopus, Cochrane Library and ClinicalTrials.gov, and used bivariate random-effects models to pool sensitivity, specificity and summary AUC. It deliberately synthesised three diagnostic questions separately because the studies were not answering the same one.

Against a clinical reference, the positive likelihood ratio was 19.2 and the negative likelihood ratio 0.05. In the clinician comparison, AUCs were 0.95 for AI and 0.90 for clinicians. Results held when restricting to deep learning models or raw waveform ECGs.

Sources

  1. Peer-reviewedCardiology in Review2026-10-29

The debate