Beyond Patient Invariance: Learning Cardiac Dynamics via Action-Conditioned JEPAs

2026-04-24

AI Summary

This work addresses the limitations of conventional self-supervised learning, which relies on patient invariance and struggles to capture transient pathological dynamics in cardiac electrophysiology. We propose an action-conditioned world model that, for the first time, explicitly represents pathology as dynamic transition vectors acting on latent states, thereby disentangling stable anatomical features from variable pathological factors. Building upon the LeJEPA framework, we design an event-conditioned joint-embedding predictive architecture to learn temporal cardiac dynamics from the MIMIC-IV-ECG dataset. By abandoning the assumption of static invariance, our approach surpasses fully supervised baselines on critical triage tasks, achieving an AUROC improvement of over 0.05 under low-resource settings. These results demonstrate that dynamic modeling yields denser and more robust supervisory signals for clinical representation learning.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics for personalization
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Self-supervised learning in healthcare has largely relied on invariance-based objectives, which maximize similarity between different views of the same patient. While effective for static anatomy, this paradigm is fundamentally misaligned with clinical diagnosis, as it mathematically compels the model to suppress the transient pathological changes it is intended to detect. We propose a shift towards Action-Conditioned World Models that learn to simulate the dynamics of disease progression, or Event-Conditioned. Adapting the LeJEPA framework to physiological time-series, we define pathology not as a static label, but as a transition vector acting on a patient's latent state. By predicting the future electrophysiological state of the heart given a disease onset, our model explicitly disentangles stable anatomical features from dynamic pathological forces. Evaluated on the MIMIC-IV-ECG dataset, our approach outperforms fully supervised baselines on the critical triage task. Crucially, we demonstrate superior sample efficiency: in low-resource regimes, our world model outperforms supervised learning by over 0.05 AUROC. These results suggest that modeling biological dynamics provides a dense supervision signal that is far more robust than static classification. Source code is available at https://github.com/cljosegfer/lesaude-dynamics

Problem

• self-supervised learning
• patient invariance
• cardiac dynamics
• pathological changes
• clinical diagnosis

Innovation

• Action-Conditioned World Models
• Self-supervised Learning
• Cardiac Dynamics
• Pathology as Transition
• LeJEPA