Base44
Mini toolsEchoJEPA
By bowang-lab
EchoJEPA is a latent predictive foundation model for echocardiography. It learns by predicting future echo views and frames in latent space across 18M exams from 300K patients, then is probed for tasks such as LVEF/RVSP estimation, chamber segmentation, view classification and quality control, improving performance and robustness over supervised baselines while avoiding label bias.
New
Video
Gen 4
Overview
Self-supervised echocardiography foundation model that predicts future ultrasound views in latent space, enabling zero-shot cardiac measurements, disease classification, motion tracking and image enhancement from standard clinical echo exams.
Tools using EchoJEPA
No tools found for this model yet.
Last updated: February 10, 2026
