The Accuracy of Artificial Intelligence-Based Models Applied to 12-Lead Electrocardiograms for the Diagnosis of Acute Coronary Syndrome: A Systematic Review

Q: What is the clinical evidence from this study?

Study design: Systematic Review. Population: Suspected Acute Coronary Syndrome (ACS) (n=118927). Intervention: Artificial intelligence (ML and DL) algorithms for 12-lead ECG interpretation vs. Clinician interpretation (retrospective or real-time), commercial rule-based ECG software, or other AI models. Primary outcome: Diagnostic accuracy (sensitivity and specificity) for acute coronary syndrome (ACS) (Pooled sensitivity 88.9% (range 68%-98%); Pooled specificity 88.7% (range 41%-98%), 95% CI 84.1%-92.4%).

Q: What does this research mean for the field?

Artificial intelligence models applied to 12-lead electrocardiograms demonstrate high diagnostic accuracy for acute coronary syndrome, with a pooled sensitivity of 88.9% and pooled specificity of 88.7%. Novelty: ClaimNovelty.SYNTHESIS. Consensus alignment: ConsensusAlignment.NEUTRAL.

Aly Fawzy; Aleena Malik; Juan Pablo Díaz-Martínez; Ani Orchanian‐Cheff; Sameer Masood

doi:10.1016/j.acepjo.2025.100240

August 22, 2025Journal of the American College of Emergency Physicians OpenOpen Access

The Accuracy of Artificial Intelligence-Based Models Applied to 12-Lead Electrocardiograms for the Diagnosis of Acute Coronary Syndrome: A Systematic Review

View Full Paper

Ask AI

Bookmark

Share

Key result

AI-based 12-lead ECG models achieve ~89% pooled sensitivity and specificity for acute coronary syndrome.

Pooled sensitivity 88.9% (range 68%-98%); Pooled specificity 88.7% (range 41%-98%)
95% CI 84.1%-92.4%
n=118,927

Figures

Why the study?

The diagnostic accuracy of AI algorithms in detecting ACS using 12-lead ECGs needed systematic evaluation.

Do artificial intelligence-based models applied to 12-lead ECGs improve the diagnostic accuracy (sensitivity and specificity) of acute coronary syndrome detection in adult patients with suspected ACS compared with clinician interpretation or rule-based commercial software?

Key Points

To evaluate the diagnostic accuracy and effectiveness of artificial intelligence algorithms interpreting 12-lead electrocardiograms to detect acute coronary syndrome.
Conducted a systematic review following PRISMA guidelines across Ovid MEDLINE, Ovid Embase, Cochrane Central, and Cochrane Database of Systematic Reviews through June 15, 2023.
Included 24 studies of adult patients (≥18 years) with suspected acute coronary syndrome evaluated with AI-based 12-lead ECG interpretation in acute care settings.
Assessed risk of bias across participants, predictors, outcomes, and statistical analyses using the Prediction model Risk Of Bias Assessment Tool (PROBAST).
AI-based sensitivity for acute coronary syndrome ranged from 68% to 98% (pooled sensitivity, 88.9% [95% CI, 84.1%-92.4%]) and specificity ranged from 41% to 98% (pooled specificity, 88.7% [95% CI, 82.1%-93.1%]).
For ST-elevated myocardial infarction/occlusion myocardial infarction subgroups, sensitivity ranged from 68% to 97% and specificity from 68% to 99%.
AI exceeded clinicians interpreting ECGs retrospectively in sensitivity (90% of studies) and PPV (100% of studies), but clinicians had better NPV (70% of studies); 38% of studies had high risk of bias, 50% had unclear risk, 3 reported code availability, and 1 compared AI to real-time emergency department physician interpretations.

Study Design

Type

Systematic Review (n=118,927)

Multicenter

Yes

Structured PICO

Do artificial intelligence-based models applied to 12-lead ECGs improve the diagnostic accuracy (sensitivity and specificity) of acute coronary syndrome detection in adult patients with suspected ACS compared with clinician interpretation or rule-based commercial software?

P

Population

Adults (≥18 years) across 24 studies (total n=118,927) presenting with suspected acute coronary syndrome in acute care settings who underwent 12-lead ECG interpretation using AI algorithms.

I

Intervention

Artificial intelligence (AI) and machine learning (ML) algorithms applied to 12-lead ECG interpretation, including deep learning methods (66.7%, 16/24 studies: artificial neural networks and convolutional neural networks), traditional ML methods (25.0%, 6/24 studies: logistic regression, gradient boosting machine, random forest, long short-term memory, and LASSO), and hybrid methods (3%, 2/24 studies: adaptive logic computing network and fusions).

C

Comparator

Clinician interpretation (retrospective expert clinician review blinded to clinical outcomes in 8/24 studies, or real-time emergency department/coronary care unit physician interpretation in 2 studies), commercial non-ML ECG software using predefined rule-based criteria (6/24 studies), or other AI models (3/24 studies).

O

Outcome

Diagnostic accuracy (sensitivity and specificity) for acute coronary syndrome (ACS) and its subtypes (STEMI, NSTEMI, and occlusion MI [OMI]).surrogate

Main Result

Effect estimate: Pooled sensitivity 88.9% (range 68%-98%); Pooled specificity 88.7% (range 41%-98%) (95% CI 84.1%-92.4%)

AI models applied to 12-lead ECGs demonstrate high pooled sensitivity (88.9%) and specificity (88.7%) for diagnosing acute coronary syndrome and generally outperform retrospective clinician and rule-based software sensitivity, though clinical adoption is currently limited by high risk of bias, lack of code transparency, and minimal real-time validation.

Limitations

Excluded studies that used ECGs from large public databases or had missing sensitivity/specificity metrics
Literature search ended in June 2023 in a rapidly evolving field
High heterogeneity in study designs, AI methodologies, and reference standards for ACS diagnosis
High risk of bias in 38% of studies and unclear risk of bias in 50% of studies per PROBAST
Limited code availability (only 3 of 24 studies published code publicly) and minimal comparison to real-time clinician interpretation (only 1 study in ED setting)
Exclusion of studies that used ECGs from public databases or had missing sensitivity and specificity values
Literature search was conducted up to June 2023, potentially missing more recent developments in AI-based diagnostic tools
Majority of comparative studies used retrospective clinician ECG annotations rather than real-time emergency department physician interpretations (only 1 study compared directly with real-time ED physician interpretation)
Considerable heterogeneity in outcomes, AI design methodologies, patient populations, and reference standards for diagnosing STEMI/OMI
Limited transparency and code availability (only 3 of 24 studies shared code publicly), and high (38%) or unclear (50%) risk of bias across studies

Ask about this paper

Answers are grounded in this paper and cite their sources.

Implication

May enhance ACS triage in acute settings; extends diagnostic evidence but leaves open prospective validation needs.

Abstract

Objectives: This systematic review aims to evaluate the diagnostic accuracy of artificial intelligence (AI) algorithms in acute coronary syndrome (ACS) detection using 12-lead electrocardiograms (ECGs). Methods: Adhering to Preferred Reporting Items for Systematic Reviews guidelines, Ovid MEDLINE, Ovid Embase, Cochrane Central, and Cochrane Database of Systematic Reviews were searched up to June 15, 2023. Eligible studies involved adults with suspected ACS and employed AI for 12-lead ECG interpretation. The primary outcomes were sensitivity and specificity, with secondary outcomes including positive predictive value (PPV), negative predictive value (NPV), and accuracy. Risk of bias was evaluated using Prediction model Risk Of Bias Assessment Tool (PROBAST). Results: From 2051 records, 24 studies were included. The sensitivity of AI-based diagnosis for ACS among the 24 studies varied from 68% to 98%, and the specificity varied from 41% to 98%. For subgroup analysis of ST-elevated myocardial infarction/occlusion myocardial infarction, sensitivity ranged from 68% to 97% and specificity from 68% to 99%. AI models outperformed clinicians interpreting ECGs retrospectively without knowledge of outcomes in sensitivity (90% of studies) and PPV (100% of studies), whereas clinicians had better NPV (70% of studies). One study compared AI with real-time emergency department physician interpretations. Three studies reported code availability. Thirty-eight percentage of studies showed a high risk of bias, with 50% showing unclear risk, although applicability concerns were minimal. Conclusion: AI models show high diagnostic accuracy for ACS using 12-lead ECGs, with potential to enhance early diagnosis. However, variability in performance, transparency challenges with limited code availability, a high risk of bias in some studies, and minimal real-time comparisons underscore the necessity for standardized reporting and open-access practices.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Fawzy et al. (2025) conducted a systematic review in Suspected Acute Coronary Syndrome (ACS) (n=118,927). Artificial intelligence (ML and DL) algorithms for 12-lead ECG interpretation vs. Clinician interpretation (retrospective or real-time), commercial rule-based ECG software, or other AI models was evaluated on Diagnostic accuracy (sensitivity and specificity) for acute coronary syndrome (ACS) (Pooled sensitivity 88.9% (range 68%-98%); Pooled specificity 88.7% (range 41%-98%), 95% CI 84.1%-92.4%). AI-based 12-lead ECG models demonstrated a pooled sensitivity of 88.9% and pooled specificity of 88.7% for diagnosing acute coronary syndrome across 24 studies.

synapsesocial.com/papers/6a9150c6b108026aa1fb6199 https://doi.org/10.1016/j.acepjo.2025.100240

View Full Paper

Ask AI

Bookmark

Share

The Accuracy of Artificial Intelligence-Based Models Applied to 12-Lead Electrocardiograms for the Diagnosis of Acute Coronary Syndrome: A Systematic Review

Figures

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

The Accuracy of Artificial Intelligence-Based Models Applied to 12-Lead Electrocardiograms for the Diagnosis of Acute Coronary Syndrome: A Systematic Review

Figures

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading