LAEF: A Lead-Agnostic ECG Foundation Model Towards Point-of-Care Diagnostics

2026-08-04

AI Summary

This work addresses the limitation of existing ECG foundation models, which rely on fixed 12-lead inputs and suffer significant performance degradation when applied to portable devices that capture only one or two leads. To overcome this, we propose the first ECG foundation model natively supporting arbitrary lead subsets by formulating the ECG as a variable-scale spatiotemporal graph. Leveraging graph attention networks, physiologically inspired intra- and inter-lead connectivity strategies, and masked node modeling, our approach naturally generalizes to any lead configuration without zero-padding or architectural modifications. The model is pretrained on large-scale 12-lead ECGs using random lead subsampling to learn configuration-robust representations. Evaluated across 18 downstream tasks, it achieves an average AUROC improvement of 3.2 points over current zero-padding baselines under single- and dual-lead settings, while matching the performance of specialized models with over 12 times more parameters in the full 12-lead scenario.

Abstract

Point-of-care cardiac devices such as smartwatches and handheld ECG recorders typically capture 1--2 leads, yet existing ECG foundation models are architecturally constrained to fixed 12-lead inputs, degrading or failing under these reduced configurations. We introduce LAEF (Lead-Agnostic ECG Foundation), a 7M-parameter ECG foundation model that can natively process any lead subset without zero-padding or architectural modification. LAEF represents ECGs as variable-size spatiotemporal graphs with physiologically motivated intra- and inter-lead connectivity, processed by a Graph Attention Network that scales naturally with active lead count.Pre-trained on 9.2M 12-lead ECGs via masked node modelling with stochastic lead sampling, LAEF learns representations robust to lead configuration. Across 18 downstream datasets, LAEF is on par with specialized 12-lead baselines over 12$\times$ larger at full lead availability. Under direct point-of-care-oriented diagnostics (1--2 leads), it outperforms all zero-padded alternatives on 17 out of 18 datasets with with a single randomly sampled lead and on 14 out of 18 with 2 leads, with an average AUROC gain of +3.2 points. Representation analysis links this advantage to architectural lead-agnosticism, and a lead-importance study across 164 cardiovascular conditions shows population-level performance is stable across single standard input leads while still recovering established clinically lead-condition associations.

Problem

• ECG foundation model
• lead-agnostic
• point-of-care diagnostics
• reduced-lead ECG
• cardiac monitoring

Innovation

• lead-agnostic
• ECG foundation model
• spatiotemporal graph
• graph attention network
• point-of-care diagnostics