Size Aware Cross-shape Scribble Supervision for Medical Image Segmentation

TL;DR

This paper introduces size-aware cross-shape scribble supervision techniques for medical image segmentation, significantly improving accuracy and addressing challenges like annotation consistency and target scale variation.

Contribution

The paper presents three novel methods—cross-shape scribble annotation, pseudo mask creation, and size-aware multi-branch architecture—that enhance weakly supervised medical image segmentation.

Findings

Achieved significant mDice score improvements across multiple datasets.

Outperformed existing state-of-the-art scribble supervision methods.

Demonstrated robustness to varying target scales in medical images.

Abstract

Scribble supervision, a common form of weakly supervised learning, involves annotating pixels using hand-drawn curve lines, which helps reduce the cost of manual labelling. This technique has been widely used in medical image segmentation tasks to fasten network training. However, scribble supervision has limitations in terms of annotation consistency across samples and the availability of comprehensive groundtruth information. Additionally, it often grapples with the challenge of accommodating varying scale targets, particularly in the context of medical images. In this paper, we propose three novel methods to overcome these challenges, namely, 1) the cross-shape scribble annotation method; 2) the pseudo mask method based on cross shapes; and 3) the size-aware multi-branch method. The parameter and structure design are investigated in depth. Experimental results show that the proposed methods have achieved significant improvement in mDice scores across multiple polyp datasets. Notably, the combination of these methods outperforms the performance of state-of-the-art scribble supervision methods designed for medical image segmentation.