Volumetric medical image segmentation via scribble annotations and shape priors

Weakly-supervised image segmentation, particularly using scribbles, has become increasingly prominent in both computer vision and medical image analysis. This popularity stems from the relative ease of obtaining scribble annotations compared to the extensive effort required for precise pixel- or voxel-level labeling. Despite their advantages, scribble-based methods often struggle with accurate boundary localization due to insufficient structural supervision of regions of interest (ROIs). Additionally, many existing approaches are primarily designed for 2D segmentation and fail to effectively utilize volumetric data when applied directly to individual image slices. In this paper, we propose a scribble-based volumetric image segmentation, Scribble2D5, which tackles 3D anisotropic image segmentation and aims to improve boundary prediction. To achieve this, we augment a 2.5D attention UNet with a proposed label propagation module to extend semantic information from scribbles and use a combination of static and active boundary prediction to learn ROI’s boundary and regularize its shape. Also, we propose an optional add-on component, which incorporates the shape prior information from unpaired segmentation masks to improve model accuracy further. Extensive experiments on three public datasets and one private dataset demonstrate our Scribble2D5 achieves state-of-the-art performance on volumetric image segmentation using scribbles and shape prior if available. Our source code is available online:https://github.com/Qybc/Scribble2D5.