🛰️ Daily AI Frontier
‹ back to 2026-08-05

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

Research Medical/Healthcare AI

Ranking

Overall 76
Content 100
Popularity 19

Observed public metrics from 1 member.

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

Merged summary

TL;DR - LAEF is a 7M-parameter, lead-agnostic ECG foundation model that processes any subset of ECG leads as a spatiotemporal graph. It improves reduced-lead diagnostics for smartwatches and handheld devices while remaining competitive on full 12-lead ECGs.

  • Pre-trained on 9.2M ECGs using masked node modeling and stochastic lead sampling.
  • Uses physiologically motivated graph connectivity and attention without zero-padding or architectural changes.
  • With one randomly sampled lead, it beat zero-padded alternatives on 17 of 18 datasets; with two leads, on 14 of 18.
  • Across 18 datasets, it matched specialized 12-lead baselines more than 12× its size and achieved a mean 3.2-point AUROC gain in reduced-lead settings.

Sources (1)

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

arXiv cs.LG Edoardo Coppola, Stefano Fiorini, Pietro Liò, Mattia Savardi, Alberto Signoroni 2026-08-04 arXiv:2608.03690
Public signals Semantic Scholar citations 0 · Semantic Scholar influential citations 0
Providers: Hugging Face · N/A OpenAlex · N/A Publisher · N/A Semantic Scholar · Citations 0 · Influential citations 0 X · N/A Fetched 2026-08-15 14:22:39.916997 UTC

TL;DR - LAEF is a 7M-parameter, lead-agnostic ECG foundation model that processes any subset of ECG leads as a spatiotemporal graph. It improves reduced-lead diagnostics for smartwatches and handheld devices while remaining competitive on full 12-lead ECGs.

  • Pre-trained on 9.2M ECGs using masked node modeling and stochastic lead sampling.
  • Uses physiologically motivated graph connectivity and attention without zero-padding or architectural changes.
  • With one randomly sampled lead, it beat zero-padded alternatives on 17 of 18 datasets; with two leads, on 14 of 18.
  • Across 18 datasets, it matched specialized 12-lead baselines more than 12× its size and achieved a mean 3.2-point AUROC gain in reduced-lead settings.
item →