EchoFM: Foundation Model for Generalizable Echocardiogram Analysis

TL;DR

EchoFM is a novel foundation model for echocardiogram videos that employs self-supervised learning to effectively capture spatio-temporal features, enabling superior performance across various cardiac imaging tasks.

Contribution

We introduce EchoFM, a self-supervised foundation model specifically designed for echocardiography videos, capable of generalizing across multiple tasks without labeled data.

Findings

Outperforms state-of-the-art methods on four downstream tasks.

Effectively captures spatio-temporal dynamics of echocardiography videos.

Pre-trained on over 290,000 videos with 20 million frames.

Abstract

Foundation models have recently gained significant attention because of their generalizability and adaptability across multiple tasks and data distributions. Although medical foundation models have emerged, solutions for cardiac imaging, especially echocardiography videos, are still unexplored. In this paper, we introduce EchoFM, a foundation model specifically designed to represent and analyze echocardiography videos. In EchoFM, we propose a self-supervised learning framework that captures both spatial and temporal variability patterns through a spatio-temporal consistent masking strategy and periodic-driven contrastive learning. This framework can effectively capture the spatio-temporal dynamics of echocardiography and learn the representative video features without any labels. We pre-train our model on an extensive dataset comprising over 290,000 echocardiography videos covering 26 scan views across different imaging modes, with up to 20 million frames of images. The pre-trained EchoFM can then be easily adapted and fine-tuned for a variety of downstream tasks, serving as a robust backbone model. Our evaluation was systemically designed for four downstream tasks after the echocardiography examination routine. Experiment results show that EchoFM surpasses state-of-the-art methods, including specialized echocardiography methods, self-supervised pre-training models, and general-purposed pre-trained foundation models, across all downstream tasks.