AI-ECG for Predicting CMR-Defined Myocardial Injury in Acute Myocardial Infarction

Artificial Intelligence-Enhanced ECG for Predicting Cardiac Magnetic Resonance-Defined Myocardial Injury in Acute Myocardial Infarction: An External Validation Study

Status

Completed

Phases

Unknown

Study type

Observational

Source

ClinicalTrials.gov

Registry ID

NCT07751809

Enrollment

461

Registered

2026-08-07

Start date

2020-04-01

Completion date

2025-06-30

Last updated

2026-08-07

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

Acute Myocardial Infarction (AMI)

Keywords

Artificial Intelligence, Electrocardiography, cardiac magnetic resonance, Acute Myocardial Infarction

Brief summary

This retrospective, single-center external validation study evaluates whether two commercially approved artificial intelligence-enhanced electrocardiography (AI-ECG) algorithms (AiTiA LVSD and AiTiA MI; Medical AI Co., Ltd.), applied to a single pre-percutaneous coronary intervention (PCI) 12-lead ECG, predict cardiac magnetic resonance (CMR)-defined myocardial injury in patients with acute myocardial infarction (AMI). The primary endpoint is a large infarct (late gadolinium enhancement \>17.9% of left ventricular mass); secondary endpoints are CMR left ventricular ejection fraction (LVEF) ≤40% and microvascular obstruction (MVO).

Detailed description

Consecutive patients with acute myocardial infarction undergoing PCI and subsequent CMR at Yongin Severance Hospital (April 2020-June 2025) were identified. Two approved AI-ECG algorithms were applied to the first 12-lead ECG obtained at emergency-department presentation (pre-PCI). CMR was performed a median of 3 days (IQR 2-5) after PCI. The diagnostic performance of each AI-ECG score for the CMR-defined endpoints was assessed by the area under the ROC curve (AUC), with sensitivity, specificity, and predictive values at prespecified and cohort-optimal (Youden) cutoffs, and by logistic regression. The final analysis cohort comprised 461 patients.

Interventions

None listed

Study design

Observational model

COHORT

Time perspective

RETROSPECTIVE

Eligibility

Sex/Gender

ALL

Age

19 Years to No maximum

Healthy volunteers

Inclusion criteria

* Acute myocardial infarction (STEMI or NSTEMI) undergoing PCI * Subsequent cardiac magnetic resonance imaging performed * Valid pre-PCI 12-lead ECG processable by both AI-ECG algorithms

Exclusion criteria

* Duplicate AMI enrollment * Invalid ECG-CMR date linkage * No valid pre-PCI 12-lead ECG processable by both algorithms

Design outcomes

Primary

Measure	Time frame	Description
Large myocardial infarct on CMR	CMR performed a median of 3 days (IQR 2-5) after PCI, during the index hospitalization.	: Late gadolinium enhancement (LGE) \>17.9% of left ventricular mass, assessed on CMR. Discrimination of the pre-PCI AI-ECG score reported as AUC.

Secondary

Measure	Time frame	Description
CMR-defined LV systolic dysfunction	CMR performed a median of 3 days (IQR 2-5) after PCI.	CMR-derived left ventricular ejection fraction (LVEF) ≤40%.
Microvascular obstruction (MVO)	CMR performed a median of 3 days (IQR 2-5) after PCI.	Presence of microvascular obstruction on CMR.

Countries

South Korea

Outcome results