Key result
Multimodal deep learning model detects acute MI from 12-lead ECGs with an AUROC of ~0.85.
Why the study?
Most prior deep learning studies for automated ECG analysis in acute myocardial infarction relied on small, curated datasets with limited external validation, limiting clinical applicability.
Does a multimodal deep learning model (Conv-BiLSTM-Attn) integrating 12-lead ECG waveforms and patient demographics improve the automated detection of acute myocardial infarction compared to benchmark deep learning architectures in emergency department patients presenting with chest pain or dyspnoea?
Cohort (n=96,813)
Yes
Does a multimodal deep learning model (Conv-BiLSTM-Attn) integrating 12-lead ECG waveforms and patient demographics improve the automated detection of acute myocardial infarction compared to benchmark deep learning architectures in emergency department patients presenting with chest pain or dyspnoea?
A multimodal deep learning model combining 12-lead ECG waveforms and basic demographics achieved high negative predictive value (97.6%) and strong cross-hospital generalizability (AUROC 0.801-0.849) for detecting acute myocardial infarction in emergency department patients, supporting its potential as an early triage and rule-out tool.
Answers are grounded in this paper and cite their sources.
May support ED AMI rule-out triage; leaves open prospective validation before adoption.
Aims: Acute myocardial infarction (AMI) remains a leading global cause of mortality, where timely diagnosis is critical to enable early intervention. The 12-lead electrocardiogram (ECG) is a critical tool for AMI detection. While deep learning (DL) models show promise for automated ECG analysis, most prior studies rely on small, curated datasets with limited external validation, limiting their clinical applicability. Methods and results: We developed a multimodal DL model (Conv-BiLSTM-Attn) integrating convolutional and recurrent neural networks with an attention mechanism. Using a large, multi-centre dataset of 145 656 ECGs from 96 813 patients across three Swedish hospitals. We trained the model to detect AMI using raw 12-lead ECG signals and demographic inputs (age, sex). Model performance was evaluated under two external validation protocols: generalization across hospitals (GAH) and leave-one-hospital-out (LOHO). The model achieved an area under the receiver operating characteristic (AUROC) of 0.848 (95% CI: 0.84-0.86) and an area under the precision-recall (AUPRC) of 0.456, reflecting class imbalance (∼6-10% AMI prevalence) under the GAH protocol. Subgroup AUROCs ranged from 0.79 to 0.92 across age and sex groups. At the Youden-optimized threshold (0.439), the model showed sensitivity of 0.736, specificity of 0.793, negative predictive value of 0.976, and weighted F1-score of 0.837. Under the LOHO protocol, AUROCs ranged from 0.801-0.849. At Youden thresholds (0.34-0.50), sensitivity ranged from 0.671 to 0.776 and specificity from 0.651 to 0.801, confirming generalizability across sites. Conv-BiLSTM-Attn outperformed benchmark models, with Score-CAM highlighting relevant ST-T segments. Conclusion: This DL model can support accurate and generalizable AMI detection from routine ECGs, with the Conv-BiLSTM-Attn architecture outperforming current benchmark approaches.
No takes yet. Share an insight, caveat, or question.
A 2025 study conducted a cohort in Suspected acute myocardial infarction (n=96,813). Multimodal deep learning model (Conv-BiLSTM-Attn) integrating raw 12-lead ECG signals and demographics (age, sex) vs. Benchmark deep learning architectures (ResNet-based Model A, CODE pretrained Model B, Model B-TF, and ECG-SMART-NET) was evaluated on Detection of acute myocardial infarction (ICD-10 code I21) assessed by area under the receiver operating characteristic curve (AUROC) under generalization across hospitals (GAH) protocol (95% CI 0.84-0.86, p=<0.001). The multimodal Conv-BiLSTM-Attn model detected acute myocardial infarction from 12-lead ECGs with an AUROC of 0.848, outperforming benchmark architectures.
Synapse has enriched 10 related papers. Consider them for comparative context: