Key result
AI ECG model identifies actionable echocardiographic pathology with ~0.80 AUC on external validation.
Why the study?
Ongoing cardiac sonographer shortages at over 70% of US institutions limit necessary TTEs, and whether deep learning models applied to ECG data can triage TTEs at low risk of identifying actionable pathology was uncertain.
Does a deep learning algorithm applied to ECG waveform inputs (ASTRAL-ECG) accurately predict a composite of actionable pathology on paired same-day transthoracic echocardiograms to help triage TTE requests?
Observational (n=123,238)
Yes
Does a deep learning algorithm applied to ECG waveform inputs (ASTRAL-ECG) accurately predict a composite of actionable pathology on paired same-day transthoracic echocardiograms to help triage TTE requests?
A novel AI-ECG deep learning model demonstrated high discrimination and negative predictive value for ruling out actionable pathology on paired TTEs, offering a potential tool to triage echocardiogram urgency during sonographer shortages.
May aid ECG-based triage for echocardiography referral; leaves open whether implementation improves outcomes or efficiency.
Background Over 70% of US institutions report ongoing cardiac sonographer shortages, limiting necessary echocardiograms (TTEs). Whether deep learning (DL) models applied to ECG data could be used to triage TTEs with low-risk of identifying actionable pathology is uncertain. Purpose We aimed to derive and externally validate a DL algorithm, ASTRAL-ECG, using ECG waveform inputs to predict the composite of LVEF < 50% or wall motion abnormalities, ≥ moderate pericardial effusion, intracardiac mass, acute aortic syndrome, VSD or ASD, 2+ or greater valvular disease, ≥ moderate pulmonary hypertension, aortic aneurysm, or ≥ moderate RV dysfunction. Methods Data from the Medical Information Mart for Intensive Care, 4th version (MIMIC-IV) were used to identify those with same day ECG and TTE at our institution. A 2D convolutional neural network, ASTRAL-ECG, was used to classify ECGs as positive with respect to the composite outcome with internal validity assessed using 10-fold cross-validation. Gradient-weighted Class Activation Mapping (Grad-CAM) maps were generated for interpretation. Results were externally validated in a multi-center ECG-TTE dataset from the EchoNext study and cutoffs chosen to maximize sensitivity. Results In MIMIC-IV (N = 45,321) and EchoNext (N = 77,917) samples, 15,410 (34.0%) and 32,676 (41.9%) had the composite outcome respectively. Model discrimination and positive test performance in the MIMIC-IV leave-out sample was excellent (accuracy 78.4%, AUC = 0.847, area under the precision recall curve [AUPRC] = 0.745, sensitivity 68.3%, specificity 83.6%). Grad-CAM identified the p-wave and QRS complex as the primary model focus for classification [Figure 1]. After similar waveform reprocessing, performance in the EchoNext validation sample was excellent (accuracy = 76.8%, AUC = 0.804, AUPRC = 0.705, sensitivity = 39.9%, specificity = 95.1%) [Figure 2]. An operating threshold of 0.098 had a sensitivity of 95.6%, specificity 40.8%, PPV 45.4%, and NPV 94.7% for the composite in the derivation leave-out test sample. Conclusions A novel AI-ECG model had high performance for ruling out actionable pathology on paired TTE. While requiring further clinical validation, this model could be used to triage the urgency of TTE requests by deprioritizing low-risk studies.
No takes yet. Share an insight, caveat, or question.
Strom et al. (2026) conducted an observational in Actionable echocardiographic pathology (n=123,238). ASTRAL-ECG deep learning algorithm vs. Transthoracic echocardiogram (TTE) findings was evaluated on Composite of LVEF < 50% or wall motion abnormalities, >= moderate pericardial effusion, intracardiac mass, acute aortic syndrome, VSD or ASD, 2+ or greater valvular disease, >= moderate pulmonary hypertension, aortic aneurysm, or >= moderate RV dysfunction (AUC 0.847 (derivation); AUC 0.804 (validation)). The ASTRAL-ECG deep learning model predicted actionable TTE pathology with an AUC of 0.847 in derivation and 0.804 in external validation, achieving 94.7% NPV at a high-sensitivity threshold.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: