Boundary-aware Information Maximization for Self-supervised Medical Image Segmentation

TL;DR

This paper introduces a novel unsupervised pre-training framework for medical image segmentation that leverages boundary-aware information maximization and over-segmentation, improving performance with limited labeled data.

Contribution

It proposes a boundary-aware information maximization approach that avoids contrastive learning drawbacks for unsupervised medical image segmentation.

Findings

Improves segmentation accuracy with limited labeled data.

Effective on benchmark medical datasets.

Outperforms existing unsupervised pre-training methods.

Abstract

Unsupervised pre-training has been proven as an effective approach to boost various downstream tasks given limited labeled data. Among various methods, contrastive learning learns a discriminative representation by constructing positive and negative pairs. However, it is not trivial to build reasonable pairs for a segmentation task in an unsupervised way. In this work, we propose a novel unsupervised pre-training framework that avoids the drawback of contrastive learning. Our framework consists of two principles: unsupervised over-segmentation as a pre-train task using mutual information maximization and boundary-aware preserving learning. Experimental results on two benchmark medical segmentation datasets reveal our method's effectiveness in improving segmentation performance when few annotated images are available.