Shape Transformation Driven by Active Contour for Class-Imbalanced Semi-Supervised Medical Image Segmentation

TL;DR

A novel Shape Transformation driven by Active Contour (STAC), that enlarges smaller organs to alleviate imbalanced class distribution across different organs and significantly outperforms some state-of-the-art methods.

Abstract

Annotating 3D medical images demands expert knowledge and is time-consuming. As a result, semi-supervised learning (SSL) approaches have gained significant interest in 3D medical image segmentation. The significant size differences among various organs in the human body lead to imbalanced class distribution, which is a major challenge in the real-world application of these SSL approaches. To address this issue, we develop a novel Shape Transformation driven by Active Contour (STAC), that enlarges smaller organs to alleviate imbalanced class distribution across different organs. Inspired by curve evolution theory in active contour methods, STAC employs a signed distance function (SDF) as the level set function, to implicitly represent the shape of organs, and deforms voxels in the direction of the steepest descent of SDF (i.e., the normal vector). To ensure that the voxels far from expansion organs remain unchanged, we design an SDF-based weight function to control the degree of deformation for each voxel. We then use STAC as a data-augmentation process during the training stage. Experimental results on two benchmark datasets demonstrate that the proposed method significantly outperforms some state-of-the-art methods. Source code is publicly available at https://github.com/GuGuLL123/STAC.

Figures (5)

• Figure 1: Motivation of STAC: enlarging smaller organs helps to alleviate imbalanced class distribution across different organs. The second and third rows are magnified images of the corresponding color boxes in the first row. The green curve in the Image represents the corresponding edges of the ground truth for categories with smaller pixel proportions. The white curve in the Transformed Image represents the edges of the corresponding ground-truth category, after processing with STAC.
• Figure 2: The pipeline of the proposed STAC framework, that is used for data-augmentation. For unlabeled samples, the Fastest Descent Direction (FDD) is defined as the gradient of the Signed Distance Function (SDF), which represents the direction of Adaptive Deformation Map (ADM). The degree of the ADM is SDF-based weight function. The shape transformation for enlarging smaller organs is obtained by interpolating the unlabeled images and pseudo labels according to the ADM. For labeled samples, the ADM is directly obtained using the label.
• Figure 3: Some qualitative segmentation results of STAC and some other methods on AMOS dataset. The first and third rows are 3D views, and the second and fourth rows are 2D slices.
• Figure 4: Plot of the SDF-based weight function with different parameters $\alpha$ and $\beta$.
• Figure 5: Dice score of using different values for parameters $\alpha$ and $\beta$ on AMOS dataset using 5% labeled images under the DHC wang2023dhc baseline. The numerical values in the heatmap represent Dice scores.