Deep learning electrocardiogram model for risk stratification of coronary revascularization need in the emergency department

Q: What is the clinical evidence from this study?

Study design: Observational. Population: Acute coronary syndrome (n=199359). Intervention: Deep learning electrocardiogram model vs. Clinician ECG interpretation and cardiac troponin T. Primary outcome: Area under the receiver operating characteristic curve (AUROC) for revascularization (AUROC 0.91, 95% CI 0.91-0.91).

Antonius Büscher; Lucas Plagwitz; Kemal Yildirim; Tobias Brix; Philipp Neuhaus; Lucas Bickmann; Amélie Friederike Menke; Vincent F van Almsick; Hermann Pavenstädt; Philipp Kümpers; Dominik Heider; Julian Varghese; Lars Eckardt

doi:10.1093/eurheartj/ehaf254

March 29, 2025European Heart JournalOpen Access

Deep learning electrocardiogram model for risk stratification of coronary revascularization need in the emergency department

View Full Paper

Ask AI

Bookmark

Share

Key result

AI ECG model outperforms clinicians for emergency revascularization prediction with an AUROC of 0.91 vs 0.65.

AUROC 0.91
95% CI 0.91-0.91
n=199,359

Why the study?

Identification of patients with acute coronary syndrome requiring coronary revascularization is challenging due to inconclusive ECG findings or biomarker results.

Does a deep learning ECG model improve the prediction of coronary revascularization and Type 1 MI compared to clinician interpretation and troponin testing in ED patients?

Key Points

To develop and externally validate a deep learning model to predict the need for coronary revascularization and identify Type 1 myocardial infarction from emergency department electrocardiograms.
Trained a convolutional neural network on 144,691 emergency department visits from a US cohort (mean age 60 ± 19 years; 53% female; 0.6% revascularization rate) and evaluated it in a separate internal test cohort (n = 35,995).
Externally validated model performance on 18,673 emergency department visits from Europe (mean age 55 ± 21 years; 49% female; 1.5% revascularization; 1% Type 1 myocardial infarction).
Benchmarked primary discrimination (area under the receiver operating characteristic curve, AUROC) against clinician electrocardiogram interpretation and conventional/high-sensitivity cardiac troponin T assays.
In the test cohort, the deep learning model achieved an AUROC of 0.91 (95% CI 0.91–0.91) for revascularization, outperforming clinician interpretation (AUROC 0.65, 95% CI 0.54–0.76) and conventional troponin T (AUROC 0.71).
In external validation, the model achieved an AUROC of 0.81 (95% CI 0.81–0.82) for revascularization and 0.85 (95% CI 0.84–0.85) for Type 1 myocardial infarction, exceeding clinician interpretation (0.70, 95% CI 0.57–0.83 and 0.74, 95% CI 0.56–0.92, respectively).
Compared with high-sensitivity troponin T (AUROC 0.85 for revascularization; 0.87 for Type 1 myocardial infarction), the model demonstrated higher specificity but lower sensitivity.

Study Design

Type

Observational (n=199,359)

Multicenter

Yes

Structured PICO

Does a deep learning ECG model improve the prediction of coronary revascularization and Type 1 MI compared to clinician interpretation and troponin testing in ED patients?

P

Population

199,359 emergency department visits across US and European cohorts evaluated for acute coronary syndrome requiring coronary revascularization.

E

Exposure

Deep learning convolutional neural network model for ECG interpretation

C

Comparator

Clinician ECG interpretation and cardiac troponin T (conventional and high-sensitivity)

O

Outcome

Area under the receiver operating characteristic curve (AUROC) for predicting coronary revascularization and Type 1 myocardial infarctionsurrogate

Main Result

Effect estimate: AUROC 0.91 (95% CI 0.91-0.91)

Absolute Event Rate: 0.91% vs 0.65%

A deep learning ECG model demonstrated competitive performance in predicting the need for coronary revascularization and Type 1 MI in the emergency department, outperforming clinician ECG interpretation.

Implication

May complement ED risk stratification; leaves open prospective validation and outcome impact.

Discussion

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

Cite This Study

Büscher et al. (2025) conducted an observational in Acute coronary syndrome (n=199,359). Deep learning electrocardiogram model vs. Clinician ECG interpretation and cardiac troponin T was evaluated on Area under the receiver operating characteristic curve (AUROC) for revascularization (AUROC 0.91, 95% CI 0.91-0.91). A deep learning ECG model outperformed clinician interpretation for detecting the need for coronary revascularization in the emergency department (AUROC 0.91 vs 0.65).

synapsesocial.com/papers/69d774e35f9a1dad534903ef https://doi.org/10.1093/eurheartj/ehaf254

View Full Paper

Ask AI

Bookmark

Share

Deep learning electrocardiogram model for risk stratification of coronary revascularization need in the emergency department

Key Points

Study Design

Structured PICO

Main Result

Implication

Discussion

Cite This Study

Deep learning electrocardiogram model for risk stratification of coronary revascularization need in the emergency department

Key Points

Study Design

Structured PICO

Main Result

Implication

Discussion

Cite This Study

Keep reading

Keep reading