Reference-guided Pseudo-Label Generation for Medical Semantic Segmentation

TL;DR

This paper introduces a reference-guided pseudo-labeling method for medical image segmentation that leverages visually similar regions between labeled and unlabeled images, reducing the need for extensive annotations while maintaining high accuracy.

Contribution

It presents a novel pseudo-label generation approach that uses reference images to improve semi-supervised segmentation without architectural changes.

Findings

Achieves comparable performance to fully supervised models with 95% fewer labeled images.

Outperforms existing methods in retinal fluid segmentation by up to 15% mean IoU.

Avoids confirmation bias common in prediction-based pseudo-labeling.

Abstract

Producing densely annotated data is a difficult and tedious task for medical imaging applications. To address this problem, we propose a novel approach to generate supervision for semi-supervised semantic segmentation. We argue that visually similar regions between labeled and unlabeled images likely contain the same semantics and therefore should share their label. Following this thought, we use a small number of labeled images as reference material and match pixels in an unlabeled image to the semantics of the best fitting pixel in a reference set. This way, we avoid pitfalls such as confirmation bias, common in purely prediction-based pseudo-labeling. Since our method does not require any architectural changes or accompanying networks, one can easily insert it into existing frameworks. We achieve the same performance as a standard fully supervised model on X-ray anatomy segmentation, albeit 95% fewer labeled images. Aside from an in-depth analysis of different aspects of our proposed method, we further demonstrate the effectiveness of our reference-guided learning paradigm by comparing our approach against existing methods for retinal fluid segmentation with competitive performance as we improve upon recent work by up to 15% mean IoU.