EchoJEPA: A Latent Predictive Foundation Model for Echocardiography

2026-02-02

AI Summary

This study addresses the challenge of disentangling anatomical structures from speckle noise and acquisition artifacts in echocardiography by introducing, for the first time in this domain, a latent prediction paradigm. Leveraging 18 million unlabeled images, a large-scale foundation model is developed to learn robust representations of cardiac anatomy through a latent prediction objective. The approach employs a frozen-backbone multi-view probing framework combined with a physics-informed acoustic perturbation evaluation strategy. Remarkably, using only 1% of labeled data, the model achieves 79% accuracy in view classification and improves estimation performance by approximately 20% for left ventricular ejection fraction and 17% for right ventricular systolic pressure. It demonstrates exceptional zero-shot generalization—particularly outperforming fine-tuned baselines on pediatric patients—and exhibits superior robustness to perturbations compared to existing methods.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Other Foundations of Computer Vision
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Foundation models for echocardiography often struggle to disentangle anatomical signal from the stochastic speckle and acquisition artifacts inherent to ultrasound. We present EchoJEPA, a foundation model trained on 18 million echocardiograms across 300K patients, representing the largest pretraining corpus for this modality to date. By leveraging a latent predictive objective, EchoJEPA learns robust anatomical representations that ignore speckle noise. We validate this using a novel multi-view probing framework with frozen backbones, where EchoJEPA outperforms leading baselines by approximately 20% in left ventricular ejection fraction (LVEF) estimation and 17% in right ventricular systolic pressure (RVSP) estimation. The model also exhibits remarkable sample efficiency, reaching 79% view classification accuracy with only 1% of labeled data versus 42% for the best baseline trained on 100%. Crucially, EchoJEPA demonstrates superior generalization, degrading by only 2% under physics-informed acoustic perturbations compared to 17% for competitors. Most remarkably, its zero-shot performance on pediatric patients surpasses fully fine-tuned baselines, establishing latent prediction as a superior paradigm for robust, generalizable medical AI.

Problem

• echocardiography
• speckle noise
• acquisition artifacts
• anatomical signal disentanglement
• ultrasound imaging

Innovation

• latent predictive modeling
• echocardiography foundation model
• speckle-robust representation
• zero-shot generalization
• sample efficiency