Key result
Single-lead ECG AI detects CAD with AUC of ~0.74, nearly matching standard 12-lead models.
Why the study?
Wearable and handheld devices record a Lead I-like ECG, and it was unclear whether a single lead retains the information that deep learning extracts from a 12-lead ECG to detect coronary artery disease.
Does a deep learning model restricted to a single Lead I ECG retain the diagnostic accuracy of a 12-lead ECG deep learning model for detecting coronary artery disease/coronary heart disease in a hospital-based adult population?
Observational (n=161,306)
Does a deep learning model restricted to a single Lead I ECG retain the diagnostic accuracy of a 12-lead ECG deep learning model for detecting coronary artery disease/coronary heart disease in a hospital-based adult population?
A deep learning model using only a single Lead I ECG retains most of the diagnostic information of a full 12-lead ECG for detecting coronary artery disease, supporting potential screening utility in wearable devices.
Answers are grounded in this paper and cite their sources.
Lead I ECG may suffice for DL-based CAD detection; leaves open need for external validation before any clinical adoption.
Wearable and handheld devices record a Lead I-like electrocardiogram (ECG), but it is unclear whether a single lead retains the information that deep learning extracts from the 12-lead ECG for coronary disease. We developed two one-dimensional residual network models, oneusingthe12-leadECGandonerestrictedtoLeadI,withidenticalpreprocessing, patient-level partitions and five-fold cross-validation, to detect coded coronary artery disease/coronary heart disease (CAD/CHD). The cohort comprised the most recent ECG of 161,306 adults in MIMIC-IV-ECG linked to MIMIC-IV, a hospital-based population spanning emergency, inpatient (including intensive care) and outpatient settings, in which controls included patients with other cardiovascular disease. In a shared holdout cohort of 31,497 patients (943 cases), the area under the receiver operating characteristic curve (AUC) was 0.764 (95% CI 0.751-0.777) for the 12-lead model and 0.744 (0.730-0.759) for Lead I (p < 0.001). Predictions were strongly correlated (Pearson r = 0.78), binary classifications agreed in 81.8% of patients, and sensitivity did not differ between models (75.0% vs. 74.0%, p = 0.55). Both models also discriminated heart failure with reduced ejection fraction, including after excluding CAD/CHD. Grad-CAM showed that both models shifted attribution towards the ST segment in ECGs classified as positive, although their attribution profiles were only partly shared. Adding the Lead I prediction to age, sex and race/ethnicity increased the AUC from 0.734 to 0.773. A single Lead I ECG retained most, but not all, of the CAD/CHD-related information available to the 12-lead model.
No takes yet. Share an insight, caveat, or question.
Borg et al. (2026) conducted an observational in Coronary artery disease/coronary heart disease (CAD/CHD) (n=161,306). Lead I ECG deep learning model (one-dimensional residual network) vs. 12-lead ECG deep learning model (one-dimensional residual network) was evaluated on Detection of coded coronary artery disease/coronary heart disease (CAD/CHD) via AUC (AUC 0.744 vs 0.764, 95% CI 0.730-0.759, p=< 0.001). A single-lead (Lead I) ECG deep learning model detected CAD/CHD with an AUC of 0.744 (95% CI 0.730-0.759) compared to 0.764 (95% CI 0.751-0.777) for a 12-lead model (p < 0.001).
Synapse has enriched 10 related papers. Consider them for comparative context: