Key result
FCN deep learning model detects myocardial ischemia on 12-lead ECGs with ~98% accuracy.
Why the study?
Early detection of myocardial ischemia remains difficult, and manual interpretation of the widely used ECG is time-consuming and subject to inter-observer variability.
Does a Fully Convolutional Network combined with LIME accurately classify myocardial ischemia from 12-lead ECG median beats and provide clinically consistent explanations compared with alternative machine learning models?
Observational (n=10,135)
Yes
Does a Fully Convolutional Network combined with LIME accurately classify myocardial ischemia from 12-lead ECG median beats and provide clinically consistent explanations compared with alternative machine learning models?
A Fully Convolutional Network applied to 12-lead ECG median beats detects myocardial ischemia with high internal and external accuracy while providing LIME-based explanations that quantitatively align with the ST-T region and cardiologist-annotated ischemic territories.
Answers are grounded in this paper and cite their sources.
Supports FCN-based ECG ischemia detection with explainable outputs; leaves open prospective validation before clinical adoption.
Myocardial ischemia is a major cause of illness and death worldwide, and its early detection remains difficult. The electrocardiogram (ECG) is the most widely used test for ischemia, but manual interpretation is time-consuming and subject to inter-observer variability. This study develops a deep learning framework based on a Fully Convolutional Network (FCN) that classifies myocardial ischemia from 12-lead ECGs and explains its decisions with Local Interpretable Model-Agnostic Explanations (LIME). The public PTB-XL database was used, comprising 11,018 ECGs from 10,135 patients (2,441 ischemia and 8,577 normal). Signals were cleaned, segmented into beats and summarized as lead-wise median beats. Training, validation and independent test sets were separated at the patient level, performance was estimated over ten patient-level bootstrap iterations, and the model was externally validated on 2,033 ECGs from the Georgia 12-lead ECG Challenge database. On the independent test set, the FCN achieved an accuracy of 97.82% ± 0.19%, precision of 95.98%, recall (sensitivity) of 94.23%, specificity of 98.86%, F1-score of 95.10% and an area under the receiver operating characteristic curve (AUC) of 0.995 (95% CI 0.995–0.996). In external validation, the AUC was 0.959 and the accuracy was 90.31%. The FCN was compared with an attention-based encoder, a simple convolutional neural network, a Transformer encoder, random forest, support vector machine and XGBoost under the same protocol. LIME explanations placed 64.7% of the positive importance of ischemic ECGs in the ST–T region (chance level 48%) and 57.3% in the cardiologist-annotated ischemia territory (chance level 47.9%), and removing the most important segments reduced the predicted ischemia probability significantly more than removing random segments. The framework therefore combines competitive performance with explanations that are quantitatively consistent with clinical knowledge. Prospective, multicenter validation is required before clinical use.
No takes yet. Share an insight, caveat, or question.
Aslan et al. (2026) conducted an observational in Myocardial ischemia (n=10,135). Fully Convolutional Network (FCN) with Local Interpretable Model-Agnostic Explanations (LIME) applied to lead-wise median ECG beats vs. Attention-based encoder, simple CNN, Transformer encoder, random forest, support vector machine (SVM), and XGBoost was evaluated on Classification of myocardial ischemia on independent internal test set (area under the receiver operating characteristic curve [AUC] and accuracy) (AUC 0.995, 95% CI 0.995-0.996). The FCN model classified myocardial ischemia from 12-lead ECGs with 97.82% accuracy and an AUC of 0.995 (95% CI 0.995-0.996) on internal testing and 0.959 AUC in external validation.
Synapse has enriched 10 related papers. Consider them for comparative context: