Paperdoi:10.21203/rs.3.rs-11437581/v1

Pediatric ECG foundation model via adult-to-pediatric fine-tuning: downstream tasks and clinical evaluation

PreprintDated Nov 10, 2026Added Nov 16, 2026

  • YL

Authors: Yu Cao; Haiyan Xu; Xianhan Li; Jian Yang; Yuqing Feng; Wei Wang; Haomin Li

Research Square preprint (2026-11-10), not peer reviewed

TL;DR

Cao, Li et al. adapted an adult ECG foundation model to pediatric 12-lead ECGs (339,563 recordings from 238,189 children). Direct transfer was poor (mean AUC 0.8256), while pediatric fine-tuning reached 0.9695, above training from scratch (0.9661), and 0.9079 with only 1% of the training data. On 10,999 external ECGs, fine-tuning achieved mean AUC 0.9385 across 30 labels, plus age regression MAE 1.12 years and congenital heart disease AUC 0.9562. In 17 clinicians on 100 cases, diagnostic accuracy rose from 56.8% to 65.4%.

Abstract

Pediatric electrocardiogram (ECG) interpretation is complicated by age-dependent electrophysiological changes that limit direct transfer of adult AI models. We asked whether an adult ECG foundation model could be adapted to pediatric ECGs. Using 339,563 12-lead ECGs from 238,189 children, we compared direct application, pediatric fine-tuning, and training from scratch under an identical architecture. Direct transfer was poor (mean AUC 0.8256). Fine-tuning reached 0.9695, above de novo training (0.9661), and attained 0.9079 using only 1% of training data. On 10,999 external ECGs, fine-tuning achieved a mean AUC of 0.9385 across 30 labels. It generalized to age regression (MAE 1.12 years) and congenital heart disease classification (AUC 0.9562). Across 17 clinicians and 100 cases, diagnostic accuracy rose from 56.8% to 65.4% (p < 0.001) and did not decline when AI erred. Adult ECG foundation models require pediatric adaptation; fine-tuning offers a data-efficient route to scalable pediatric ECG AI.

Verbatim from the publisher's DOI record (opens in a new tab) · Publisher licence: https://creativecommons.org/licenses/by/4.0/

Results on ChatPPG leaderboards

All leaderboards

No results from this paper on ChatPPG leaderboards yet. Browse the leaderboards.

Tasks & signals

Signals

Keywords

ECG Foundation ModelPediatricFine TuningTransferCongenital Heart DiseaseAge RegressionReader Study

How it works

How Method works: Setup → Method → TargetSetup: Signals (ecg). Method: Method.SetupSignals: ecgSignalsecgMethodMethod

Cite this paper

Cite this paper with the fields recorded here. Check it against the publisher before submitting.

Cao, Y., Xu, H., Li, X., Yang, J., Feng, Y., Wang, W., & Li, H. (2026). Pediatric ECG foundation model via adult-to-pediatric fine-tuning: downstream tasks and clinical evaluation. Research Square preprint , not peer reviewed. https://doi.org/10.21203/rs.3.rs-11437581/v1
Cao, Yu, Haiyan Xu, Xianhan Li, Jian Yang, Yuqing Feng, Wei Wang, and Haomin Li. 2026. "Pediatric ECG foundation model via adult-to-pediatric fine-tuning: downstream tasks and clinical evaluation." Research Square preprint , not peer reviewed. https://doi.org/10.21203/rs.3.rs-11437581/v1.
Cao, Y., Xu, H., Li, X., Yang, J., Feng, Y., Wang, W. and Li, H. (2026) 'Pediatric ECG foundation model via adult-to-pediatric fine-tuning: downstream tasks and clinical evaluation', Research Square preprint , not peer reviewed. doi: 10.21203/rs.3.rs-11437581/v1.
Cao, Yu, et al. "Pediatric ECG foundation model via adult-to-pediatric fine-tuning: downstream tasks and clinical evaluation." Research Square preprint , not peer reviewed, 2026. https://doi.org/10.21203/rs.3.rs-11437581/v1.
Y. Cao et al., "Pediatric ECG foundation model via adult-to-pediatric fine-tuning: downstream tasks and clinical evaluation," Research Square preprint , not peer reviewed, 2026, doi: 10.21203/rs.3.rs-11437581/v1.
@misc{cao2026pediatric,
  title         = {{Pediatric ECG foundation model via adult-to-pediatric fine-tuning: downstream tasks and clinical evaluation}},
  author        = {Cao, Yu and Xu, Haiyan and Li, Xianhan and Yang, Jian and Feng, Yuqing and Wang, Wei and Li, Haomin},
  year          = {2026},
  howpublished  = {Research Square preprint , not peer reviewed},
  doi           = {10.21203/rs.3.rs-11437581/v1}
}

Added to the ChatPPG research hub