cs.CVAug 12, 2026

A Vision Transformer for ECG-Based Detection of Left Ventricular Systolic Dysfunction Across Multiple Clinical Sites

Authors: Burcu Ozek, Aruna Mohan, David Vorchheimer, Daniel Weiss, Eyal Kedar, Tamar Sobol, Or Zilbershot, Fatemeh Afghah

Organizations: Walkky LLC, Boston, MA, USA · Northwell Cardiovascular Institute, Lenox Hill Hospital, New York, NY, USA · Weiss Consulting Groups, Boca Raton, FL, USA · St. Lawrence Health, Canton, NY, USA · Clemson University, Clemson, SC, USA

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Reduced left ventricular ejection fraction (LVEF) is frequently asymptomatic and often detected only after advanced heart failure develops. Electrocardiograms are recorded routinely yet underused for this condition, because reduced LVEF has no single diagnostic waveform. We trained an ensemble of vision transformers from scratch to detect reduced LVEF (≤\leq40%) from 12-lead ECGs, analyzing each heartbeat individually, using 10,142 patients across seven sites in three US health systems. In a held-out external cohort of 4,092 patients from three geographically independent US clinical sites at a real-world reduced-LVEF prevalence of 8.72%, the model achieved an AUROC of 0.88 (95% CI 0.86-0.89), sensitivity 81.2%, specificity 81.0%, and negative predictive value 97.8%. Sensitivity remained high across sex, race, ethnicity, and comorbidity subgroups, while specificity was lower in older patients and those with atrial fibrillation or cardiomyopathy. Beat-level attention maps provided interpretability into the model's predictions, showing consistent focus on the QRS complex rather than the P wave. These findings support the potential of routine ECGs as a scalable first-pass triage step to identify patients who should undergo echocardiography for reduced ejection fraction across diverse patient populations.

Figures & tables

CharacteristicInternal ValidationExternal Validation
 (n = 3,295)(n = 4,092)
LVEF ≤40%\leq 40\% , % (n)8.71% (287)8.72% (357)
Age, mean (SD), years66.5 (15.8)65.2 (16.6)
Female, %51.1%50.1%
Race/ethnicity, %  
White, non-Hispanic51.1%62.3%
Table 1: Patient characteristics of the internal and external validation sets.
MetricInternal ValidationExternal Validation
 (n = 3,295)(n = 4,092)
AUROC0.880 (0.859–0.897)0.878 (0.858–0.894)
Sensitivity82.9% (78.2–87.0)81.2% (76.9–85.0)
Specificity80.4% (78.9–81.8)81.0% (79.8–82.3)
PPV28.7% (25.7–31.8)29.1% (26.3–32.0)
NPV98.0% (97.4–98.5)97.8% (97.3–98.3)
Table 2: ViT ensemble performance on internal and external validation sets. All metrics computed at threshold = 0.45. 95% CIs from BCa bootstrap (10,000 resamples).
Figure 1: Calibration of the ViT ensemble on the external validation set (n = 4,092). The raw ensemble (orange) overestimated risk, falling below the diagonal. Post-hoc Platt scaling (blue), fit on the internal validation set and applied to the external set, restored calibration to the diagonal without affecting discrimination (AUROC) or the binary operating point.
CategorySubgroupnPrev (%)Sens (%)Spec (%)AUC
SexMale2,04211.381.479.10.873
 Female2,0506.181.082.80.879
Age<< 60 years1,3498.280.089.30.918
 60–75 years1,4658.777.381.00.863
 >> 75 years1,2789.386.672.30.850
Race/EthnicityWhite, non-Hispanic2,5488.277.579.60.849
Table 3: Subgroup performance on the external validation set. All metrics computed at threshold = 0.45. Prev = LVEF ≤40%\leq 40\% prevalence within subgroup. Sens = sensitivity; Spec = specificity; AUC = AUROC. ∗* Fewer than 10 positive cases; estimates should be interpreted with caution.
Figure 2: Attention maps of the 12-lead ViT by prediction category. These maps show which parts of each heartbeat the model learned to focus on, rather than relying on predefined rules. Attention weights were extracted from the final transformer encoder layer, averaged across 4 attention heads, 3 ensemble models, and all patients per category. The Lead II signal (mean ±\pm 1 SD) is overlaid on the attention heatmap (red intensity). Attention weights are derived from the full 12-lead input; the Lead II signal is shown for display purposes only. TP = true positive; TN = true negative; FN = false negative; FP = false positive.
Figure 3: Decision curve analysis for the ViT-ECG ensemble (external validation, n = 4,092). Panel (A) shows the standard DCA: net benefit across threshold probabilities from 5% to 16% for the ViT-ECG model (blue), treat-all (red dashed), and treat-none (green dotted). Panel (B) shows the net benefit gain: the improvement of the model over the best simple strategy at each threshold (treat-all where its net benefit is positive, treat-none otherwise).
VariantCWLSL2AugOSFPSSensSpecAUCΔ\Delta AUC
Full ensemble (reference)✓0.1✓0.1525081.2%81.0%0.8775—
Label smoothing✓0.0✓0.1525076.8%79.9%0.8640++ 1.35
L2 regularization✓0.1×\times0.1525079.3%78.9%0.8679++ 0.96
Augmentation✓0.1✓0.025079.8%79.7%0.8698++ 0.77
Class weight×\times0.1✓0.1525073.7%85.7%0.8744++ 0.31
Oversampling✓0.1✓0.1515075.6%84.6%0.8750++ 0.25
Table 4: Ablation study: contribution of each training component. Δ\Delta AUC = AUC(full ensemble) −- AUC(variant), expressed in AUROC percentage points, representing the benefit of including each component (positive = component improves performance). Each variant is a 3-model ensemble with one component removed. Threshold = 0.45. External validation set. CW = class weighting; LS = label smoothing; L2 = L2 regularization; Aug = augmentation ratio; OSF = oversampling factor; PS = patch size. A check mark indicates the component was included; a cross indicates it was removed.
ModelOSFAugLSExternalInternal
SensSpecAUCSensSpecAUC
CNN vanilla (5,000 signals)10077.351.70.72676.350.70.688
CNN (1,500, best single model)10078.477.30.86078.076.90.849
CNN (1,500, ensemble, ViT params)20.150.195.840.40.86796.540.70.868
CNN (1,500, best ensemble)10079.677.60.85377.477.70.854
ViT (3-model ensemble)20.150.181.281.00.87882.980.40.880
Table 5: CNN baseline configurations vs. ViT ensemble. All metrics at threshold = 0.45. OSF = oversampling factor; Aug = augmentation ratio; LS = label smoothing.
Figure 4: Vision Transformer architecture for ECG-based LVEF classification. ECG beat windows are divided into patches and projected to embeddings. A learnable CLS token is prepended, and the sequence is processed by Transformer Encoder layers with pre-norm multi-head self-attention and residual connections. The CLS token representation is passed to an MLP head to predict LVEF ≤40%\leq 40\% or >40%>40\% .
Figure 5: Ensemble training strategy. Three ViT models were trained on independent random subsamples of the training data and combined into an ensemble by averaging their predicted probabilities.
Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Figure S1: Mean attention maps stratified by atrial fibrillation status. Each row corresponds to a prediction category (true positive, false negative, false positive) and each column to AFib vs. non-AFib patients in the external validation set. The blue line and shaded band show the mean ±\pm SD of the normalized RR-segmented ECG signal (Lead II); the red heatmap overlaid on each panel shows the normalized mean attention weight from the last transformer layer (4 heads, 3 models averaged). The number of patients ( nn ) contributing to each panel is shown in the subplot title.
Figure S2: Mean attention maps stratified by age group ( <60{<}60 , 6060 – 7575 , >75{>}75 years). Each row corresponds to a prediction category; columns correspond to the three age groups. Layout otherwise identical to Figure S1 .
Figure S3: Mean attention maps stratified by sex (female vs. male). Layout identical to Figure S1 .

Explore similar work

CardsList

  1. Detecting low left ventricular ejection fraction from ECG using an interpretable and scalable predictor-driven framework

    Mar 30, 2026Ya Zhou, Tianxiang Hao, Ziyi Cai +5Left Ventricular Ejection Fraction EstimationElectrocardiography

    Department of Information Center, Fuwai Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China · The Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, China · Institute of Statistics and Big Data, Renmin University of China, Beijing, China +2

  2. A Multimodal and Explainable Machine Learning Approach to Diagnosing Multi-Class Ejection Fraction from Electrocardiograms

    Apr 17, 2026Catherine Ning, Yu Ma, Cindy Beini Wang +4Left Ventricular Ejection Fraction EstimationMultimodal Classification

    Operations Research Center, Massachusetts Institute of Technology, Cambridge, MA, USA · Wisconsin School of Business, University of Wisconsin–Madison, Madison, WI, USA · Heart and Vascular Institute, Hartford Hospital, Hartford, CT, USA

  3. Echo-E3^3Net: Efficient Endocardial Spatio-Temporal Network for Ejection Fraction Estimation

    Mar 21, 2025Moein Heidari, Afshin Bozorgpour, AmirHossein Zarif-Fakharnia +5Efficient Neural Network InferenceLeft Ventricular Ejection Fraction Estimation

    University of British Columbia, Vancouver, BC, Canada · University of Regensburg, Regensburg, Germany · Iran University of Science and Technology, Tehran, Iran +1