Fully Automated Diagnosis of Acute Myocardial Infarction Using Electrocardiograms and Multimodal Deep Learning
Authors' conclusion
The deep learning model demonstrated strong AMI detection across diverse patient groups. A randomized trial is needed to compare its performance with emergency physicians.
Study context
- Population
- Patients with chest pain or dyspnea who received a prehospital or in-hospital electrocardiogram (ECG), Västra Götaland County, Sweden (January 2015-June 2023)individuals (208,366 ECGs)
- Participants
- 104,507 patients
- Index test or model
- Deep learning model vs acute myocardial infarction (AMI) defined by International Classification of Diseases codes at discharge
- Target condition
- acute myocardial infarction (AMI)
Methodological detailsdevelopment and validation of a deep learning model, 2 centers for training and validation, third center for external testing
- Development and validation of a deep learning model
- 2 centers for training and validation
- Third center for external testing
Performance
As reported in the abstract
| Test or model | AUROC |
|---|---|
| Internal validation | 0.8221 ± 0.0101 |
| External validation | 0.8314 ± 0.0085 |
AUROC
Internal validation0.8221 ± 0.0101
External validation0.8314 ± 0.0085
Further findings
AMI detection in ambulance-arriving patients
- Area under the receiver operating characteristic
- 0.8081 ± 0.0095
- Slightly lower for ambulance-arriving patients
Performance by sex
- Performance was consistent across sex
Proportion diagnosed with AMI
- Training set
- 8.17%
- External set
- 8.59%
Saliency maps
- Saliency maps highlighted focus on ST segments and T waves
Original publication
Fully Automated Diagnosis of Acute Myocardial Infarction Using Electrocardiograms and Multimodal Deep Learning. JACC Adv. Published 17 July 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →