Scaling ECG Foundation Models and Identifying a Threshold for Effective Representation Learning

We conducted a scaling evaluation of unlabeled pretraining for electrocardiogram foundation model performance. One-dimensional vision transformer (1D-ViT) masked autoencoders were pretrained across increasing ECG volumes and fine-tuned for rhythm, morphology, diagnostic, and structural heart disease tasks. Models pretrained at ≤400,000 ECGs failed to consistently exceed controls without self-supervised pre-training, whereas 600,000-800,000 ECGs improved AUROC across tasks, suggesting a minimum threshold for effective ECG representation learning.