External validation of an artificial intelligence–based ECG algorithm for the early diagnosis of acute myocardial infarction: a prospective diagnostic cohort study

Background A novel artificial intelligence–based electrocardiogram (AI-ECG) model (AiTiA MI v2.00.00) for the early diagnosis of acute myocardial infarction (AMI) was developed using data from Korean patients. Its performance in a European population remains unknown. Methods Patients presenting to the emergency department (ED) with suspected AMI were enrolled in two prospective European cohorts (APACE & APACE-Light; ClinicalTrials.gov number, NCT00470587). The primary outcome was AMI, centrally adjudicated according to the fourth Universal Definition of MI. The AI-ECG model was applied to 12-lead ECGs recorded at ED presentation. Its diagnostic performance for AMI was directly compared to that of the BASEL score, high-sensitivity cardiac troponin T (hs-cTnT), the recalibrated HEART score (all at ED presentation), and integrated physician judgment at 90 min, with the last two only available in APACE. Predefined AI-ECG triage thresholds for early rule-out and rule-in of AMI were also evaluated. Findings Among 12,630 patients with an available digital 12-lead ECG (5047 in APACE and 7583 in APACE-Light), adjudicated AMI was diagnosed in 19% (968/5047) and 18% (1386/7583) of patients, respectively. In patients with hs-cTnT at presentation (n = 5031 and n=7565), the AI-ECG model showed good diagnostic performance for AMI with an area under the receiver operating characteristic curve (AUC) of 0.78 (95% CI: 0.77–0.80) and 0.84 (0.83–0.85) and an area under the precision–recall curve (AUPRC) of 0.56 (0.53–0.59) and 0.65 (0.62–0.67). Its discrimination was inferior to hs-cTnT (AUC 0.94 [0.93–0.95] and 0.93 [0.92–0.94]). The AI-ECG model provided some incremental diagnostic value when combined with hs-cTnT (AUC 0.95 [0.94–0.96] and 0.95 [0.94–0.95], p < 0.001). The predefined triage thresholds ruled out 14% (706/5031) and 19% (1437/7565) of patients, with a sensitivity for AMI of 97.6% (96.5–98.5) and 97.5% (96.6–98.3), respectively. Interpretation AI-ECG showed high, but not very high, diagnostic accuracy for AMI. As a diagnosis of AMI is unlikely to be ruled out on ECG alone, combining AI-ECG with hs-cTnT may improve decision making in the ED, warranting prospective studies. Funding The Swiss National Science Foundation, the Swiss Heart Foundation, Innosuisse, the University Hospital Basel, the University of Basel, Abbott, Beckman Coulter, Idorsia, LSI Medience Corporation, Roche, Ortho Clinical Diagnostics, Quidel, Siemens, SpinChip, Upstream, and Singulex, the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grant number: RS-2026-25543703).

Authors

Institutions

Publication Details

Journal
The Lancet Regional Health - Europe
Published
2026-11-07
DOI
https://doi.org/10.1016/j.lanepe.2026.101912
Primary Topic
ECG Monitoring and Analysis
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

External validation of an artificial intelligence–based ECG algorithm for the early diagnosis of acute myocardial infarction: a prospective diagnostic cohort study

Stephan Steuer, Esther Rodríguez Adrada, Hak Seung Lee, Yong‐Yeon Jo et al.
3 links
The Lancet Regional Health - Europe
ECG Monitoring and Analysis
article

External validation of an artificial intelligence–based ECG algorithm for the early diagnosis of acute myocardial infarction: a prospective diagnostic cohort study

Stephan Steuer, Esther Rodríguez Adrada, Hak Seung Lee, Yong‐Yeon Jo, Christian Puelacher, Christian Hans Nickel, Ivo Strebel, Piotr Muzyk, Arnold von Eckardstein, Pedro López‐Ayala, Damian Kawecki, Michael Christ, Min Sung Lee, Ksenija Slankamenac, Felix Mahfoud, KATHARINA M. RENTSCH, Kathrin Meissner, Philippe Claude Cattin, Luca Crisanti, Thomas Nestelberger, Kyuseok Kim, Desiree Nadine Wussler, Jude Formambuh, Luca Koechlin, Angelika Hammerer‐Lercher, Roland M Bingisser, Joon‐myoung Kwon, Tobias Zimmermann, Nicolas Geigy, Eliška Potluková, Jasper Boeddinghaus, Paolo Bima, Karin S. Wildi, Gemma Martínez‐Nadal, Óscar Miró, Danielle Menosi Gualandro, Carlos Spagnuolo, Niclas Geldermann, Jonas Glaeser, Sunyoung Yoon, Dagmar Keller, Julia Reinhardt, Eliza Kaplan, Christian Eugen Mueller, Andreas Buser, Maria Rubini Gimenez, Beatriz López, Lourdes Herraiz Recuenco, Gabrielle Huré, Jeanne du Fay de Lavallaz, Arnaud Champetier, Tzu-Ying Tsao, Marco Nani, Markku Mund, Kyung Su Kim
article en

Abstract

Background A novel artificial intelligence–based electrocardiogram (AI-ECG) model (AiTiA MI v2.00.00) for the early diagnosis of acute myocardial infarction (AMI) was developed using data from Korean patients. Its performance in a European population remains unknown. Methods Patients presenting to the emergency department (ED) with suspected AMI were enrolled in two prospective European cohorts (APACE & APACE-Light; ClinicalTrials.gov number, NCT00470587). The primary outcome was AMI, centrally adjudicated according to the fourth Universal Definition of MI. The AI-ECG model was applied to 12-lead ECGs recorded at ED presentation. Its diagnostic performance for AMI was directly compared to that of the BASEL score, high-sensitivity cardiac troponin T (hs-cTnT), the recalibrated HEART score (all at ED presentation), and integrated physician judgment at 90 min, with the last two only available in APACE. Predefined AI-ECG triage thresholds for early rule-out and rule-in of AMI were also evaluated. Findings Among 12,630 patients with an available digital 12-lead ECG (5047 in APACE and 7583 in APACE-Light), adjudicated AMI was diagnosed in 19% (968/5047) and 18% (1386/7583) of patients, respectively. In patients with hs-cTnT at presentation (n = 5031 and n=7565), the AI-ECG model showed good diagnostic performance for AMI with an area under the receiver operating characteristic curve (AUC) of 0.78 (95% CI: 0.77–0.80) and 0.84 (0.83–0.85) and an area under the precision–recall curve (AUPRC) of 0.56 (0.53–0.59) and 0.65 (0.62–0.67). Its discrimination was inferior to hs-cTnT (AUC 0.94 [0.93–0.95] and 0.93 [0.92–0.94]). The AI-ECG model provided some incremental diagnostic value when combined with hs-cTnT (AUC 0.95 [0.94–0.96] and 0.95 [0.94–0.95], p < 0.001). The predefined triage thresholds ruled out 14% (706/5031) and 19% (1437/7565) of patients, with a sensitivity for AMI of 97.6% (96.5–98.5) and 97.5% (96.6–98.3), respectively. Interpretation AI-ECG showed high, but not very high, diagnostic accuracy for AMI. As a diagnosis of AMI is unlikely to be ruled out on ECG alone, combining AI-ECG with hs-cTnT may improve decision making in the ED, warranting prospective studies. Funding The Swiss National Science Foundation, the Swiss Heart Foundation, Innosuisse, the University Hospital Basel, the University of Basel, Abbott, Beckman Coulter, Idorsia, LSI Medience Corporation, Roche, Ortho Clinical Diagnostics, Quidel, Siemens, SpinChip, Upstream, and Singulex, the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grant number: RS-2026-25543703).

The Lancet Regional Health - EuropeVol. 71
University of Basel (CH), University of Zurich (CH), University Hospital of Basel (CH), University Hospital of Zurich (CH), Hospital Clínic de Barcelona (ES), Luzerner Kantonsspital (CH), CHA University Bundang Medical Center (KR), CHA University (KR)
Universitätsspital Basel
Good health and well-being
Openalex Percentile: Top 13%
ECG Monitoring and Analysis
DOIOpen Access PDFOpenAlex
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon