Scribble2D5: Weakly-Supervised Volumetric Image Segmentation via Scribble Annotations

Abstract

Image segmentation using weak annotations like scribbles has gained great attention, since such annotations are easier to obtain compared to time-consuming and labor-intensive labeling at the pixel/voxel level. However, scribbles lack structure information of the region of interest (ROI), thus existing scribble-based methods suffer from poor boundary localization. Moreover, current methods are mostly designed for 2D image segmentation, which do not fully leverage volumetric information. In this paper, we propose a scribble-based volumetric image segmentation, Scribble2D5, which tackles 3D anisotropic image segmentation and improves boundary predictions. To achieve this, we augment a 2.5D attention UNet with a proposed label propagation module to extend semantic information from scribbles and a combination of static and active boundary prediction to learn ROI’s boundaries and regularize its shape. Extensive experiments on three public datasets demonstrate Scribble2D5 significantly outperforms existing scribble-based methods and approaches the performance of fully-supervised ones. Our code is available at https://github.com/Qybc/Scribble2D5.