Single lead electrocardiographic detection of left ventricular systolic dysfunction in pediatric and congenital heart disease

Q: What are the key findings of this study?

A noise-adapted AI-enabled single-lead ECG model accurately predicted moderate left ventricular systolic dysfunction (LVEF ≤40%) in pediatric and congenital heart disease patients, maintaining an AUROC of 0.91 under simulated real-world noise.

Q: What is the clinical evidence from this study?

Study design: Observational. Population: Left ventricular systolic dysfunction in pediatric and congenital heart disease (n=113494). Intervention: Noise-adapted AI-enabled single-lead ECG model vs. Baseline AI-ECG model (without noise adaptation). Primary outcome: Prediction of left ventricular ejection fraction ≤40% under simulated real-world noise (SNR 0.5) (AUROC 0.91, p=<0.0001).

Q: What does this research mean for the field?

A noise-adapted AI-enabled single-lead ECG model accurately detects left ventricular systolic dysfunction (LVEF ≤ 40%) in pediatric and congenital heart disease patients across diverse clinical settings. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Q: What question did this study set out to answer?

To develop and evaluate a noise-adapted, single-lead artificial intelligence electrocardiogram model for detecting left ventricular systolic dysfunction in pediatric and congenital heart disease patients.

Joshua Mayourian; Ivor Asztalos; William La Cava; Ryan Kobayashi; Sunil J. Ghelani; Victoria L. Vetter; Rachel M. Wald; Anne Marie Valente; Tal Geva; John K. Triedman

doi:10.1038/s41746-026-02646-x

July 11, 2026npj Digital MedicineOpen Access

Single lead electrocardiographic detection of left ventricular systolic dysfunction in pediatric and congenital heart disease

View Full Paper

Ask AI

Bookmark

Share

Key result

Noise-adapted AI single-lead ECG predicts LVEF ≤40% with 0.91 AUROC in pediatric and CHD patients.

AUROC 0.91
P<0.0001
n=113,494

Why the study?

Lifelong care challenges and disparities in congenital heart disease underscore the need for portable, scalable AI-enabled smartwatch solutions tailored to this population.

Does a noise-adapted single-lead AI-ECG model accurately predict left ventricular systolic dysfunction in pediatric and congenital heart disease patients?

Key Points

To develop and evaluate a noise-adapted, single-lead artificial intelligence electrocardiogram model for detecting left ventricular systolic dysfunction in pediatric and congenital heart disease patients.
Conducted external testing on 76,400 ECG-echocardiogram pairs (42,984 patients) from Children's Hospital of Philadelphia and 629 ECG-cardiac MRI pairs (284 repaired tetralogy of Fallot patients) from Toronto General Hospital.
At a simulated real-world signal-to-noise ratio of 0.5, the noise-trained model maintained high discrimination for LVEF ≤ 40% (AUROC 0.91 vs. AUROC 0.44 for baseline model; p<0.0001).
Across institutions, the low-risk threshold for LVEF ≤ 40% yielded 91–92% sensitivity and NPV ≥99.8%, while the high-risk threshold achieved 97–98% specificity with positive likelihood ratios between 15.7 and 27.3.
Patients classified as high risk by the model demonstrated a 7.9-fold increased hazard of all-cause mortality compared to those classified as low risk.

Study Design

Type

Observational (n=113,494)

Multicenter

Yes

Structured PICO

Does a noise-adapted single-lead AI-ECG model accurately predict left ventricular systolic dysfunction in pediatric and congenital heart disease patients?

P

Population

113,494 pediatric and congenital heart disease patients from three tertiary centers with paired ECG and echocardiogram or CMR data used to train and validate an AI model for detecting left ventricular systolic dysfunction.

E

Exposure

Noise-adapted single-lead AI-ECG model (convolutional neural network) trained on noisy single-lead ECGs to simulate real-life wearable device data.

C

Comparator

Baseline single-lead AI-ECG model trained without added noise.

O

Outcome

Left ventricular systolic dysfunction defined as LVEF ≤40%.surrogate

Main Result

Effect estimate: AUROC 0.91

Absolute Event Rate: 0.91% vs 0.44%

p-value: p=<0.0001

A noise-adapted single-lead AI-ECG model can accurately detect left ventricular systolic dysfunction in pediatric and congenital heart disease patients, offering a scalable tool for remote monitoring via wearable devices.

Limitations

Clinical ECG lead I signals may differ from those captured by portable devices
Selection bias may exist due to reliance on documented LVEF
Model is limited to assessing left, not right, ventricular health
Lower performance for select cases such as ventricular pacing
External validation was limited to tertiary centers
Reliance on institutional coding systems introduces risk of lesion and mortality misclassification

Implication

May support AI-ECG screening in pediatric CHD; hypothesis-generating for wearable use pending prospective trials.

Abstract

Portable, scalable, and accessible artificial intelligence (AI)-enabled smartwatch technology shows promise as a cardiovascular risk stratification strategy in the general adult population. The challenges of lifelong care in congenital heart disease (CHD)—inclusive of regional and socioeconomic disparities worldwide—underscore the need for similar solutions tailored to this population. Herein, we present the first noise-adapted single-lead ECG model for predicting left ventricular systolic dysfunction (left ventricular ejection fraction [LVEF] ≤ 40%) in pediatric and CHD patients. The internal cohort was comprised of 70,226 patients. External test groups included Children’s Hospital of Philadelphia (CHOP; 42,984 patients) and Toronto General Hospital (TGH; 284 repaired tetralogy of Fallot patients). Our model had strong performance across a broad range of CHD lesions, races, ages, and healthcare systems. Our findings support the potential of AI-enabled wearables to expand global access to CHD care. Prospective studies utilizing wearable ECG devices in pediatric and CHD patients are warranted.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Mayourian et al. (2026) conducted an observational in Left ventricular systolic dysfunction in pediatric and congenital heart disease (n=113,494). Noise-adapted AI-enabled single-lead ECG model vs. Baseline AI-ECG model (without noise adaptation) was evaluated on Prediction of left ventricular ejection fraction ≤40% under simulated real-world noise (SNR 0.5) (AUROC 0.91, p=<0.0001). A noise-adapted AI-enabled single-lead ECG model accurately predicted moderate left ventricular systolic dysfunction (LVEF ≤40%) in pediatric and congenital heart disease patients, maintaining an AUROC of 0.91 under simulated real-world noise.

synapsesocial.com/papers/6a534ff4985393734a594cff https://doi.org/10.1038/s41746-026-02646-x

View Full Paper

Ask AI

Bookmark

Share

Single lead electrocardiographic detection of left ventricular systolic dysfunction in pediatric and congenital heart disease

Key Points

Study Design

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Single lead electrocardiographic detection of left ventricular systolic dysfunction in pediatric and congenital heart disease

Key Points

Study Design

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading