Generating and Weighting Semantically Consistent Sample Pairs for Ultrasound Contrastive Learning

TL;DR

This paper introduces Meta-USCL, a meta-learning-based contrastive learning method for ultrasound images that improves representation learning by generating semantically consistent sample pairs, reducing domain gap issues.

Contribution

It proposes a novel meta-learning approach for generating and weighting sample pairs in ultrasound contrastive learning, enhancing medical image representation without labeled data.

Findings

Achieves state-of-the-art results on multiple CAD tasks

Effectively reduces domain gap between natural and medical images

Improves ultrasound image representations using self-supervised learning

Abstract

Well-annotated medical datasets enable deep neural networks (DNNs) to gain strong power in extracting lesion-related features. Building such large and well-designed medical datasets is costly due to the need for high-level expertise. Model pre-training based on ImageNet is a common practice to gain better generalization when the data amount is limited. However, it suffers from the domain gap between natural and medical images. In this work, we pre-train DNNs on ultrasound (US) domains instead of ImageNet to reduce the domain gap in medical US applications. To learn US image representations based on unlabeled US videos, we propose a novel meta-learning-based contrastive learning method, namely Meta Ultrasound Contrastive Learning (Meta-USCL). To tackle the key challenge of obtaining semantically consistent sample pairs for contrastive learning, we present a positive pair generation module along with an automatic sample weighting module based on meta-learning. Experimental results on multiple computer-aided diagnosis (CAD) problems, including pneumonia detection, breast cancer classification, and breast tumor segmentation, show that the proposed self-supervised method reaches state-of-the-art (SOTA). The codes are available at https://github.com/Schuture/Meta-USCL.