Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Synergy-Guided Regional Supervision of Pseudo Labels for Semi-Supervised Medical Image Segmentation

Tao Wang, Xinlin Zhang, Yuanbin Chen, Yuanbo Zhou, Longxuan Zhao, Tao Tan, Tong Tong

TL;DR

A novel Synergy-Guided Regional Supervision of Pseudo Labels (SGRS-Net) framework built upon the mean teacher network is introduced, which employs a Mix Augmentation module to enhance the unlabeled data.

Abstract

Semi-supervised learning has received considerable attention for its potential to leverage abundant unlabeled data to enhance model robustness. Pseudo labeling is a widely used strategy in semi supervised learning. However, existing methods often suffer from noise contamination, which can undermine model performance. To tackle this challenge, we introduce a novel Synergy-Guided Regional Supervision of Pseudo Labels (SGRS-Net) framework. Built upon the mean teacher network, we employ a Mix Augmentation module to enhance the unlabeled data. By evaluating the synergy before and after augmentation, we strategically partition the pseudo labels into distinct regions. Additionally, we introduce a Region Loss Evaluation module to assess the loss across each delineated area. Extensive experiments conducted on the LA dataset have demonstrated superior performance over state-of-the-art techniques, underscoring the efficiency and practicality of our framework.

Synergy-Guided Regional Supervision of Pseudo Labels for Semi-Supervised Medical Image Segmentation

TL;DR

A novel Synergy-Guided Regional Supervision of Pseudo Labels (SGRS-Net) framework built upon the mean teacher network is introduced, which employs a Mix Augmentation module to enhance the unlabeled data.

Abstract

Semi-supervised learning has received considerable attention for its potential to leverage abundant unlabeled data to enhance model robustness. Pseudo labeling is a widely used strategy in semi supervised learning. However, existing methods often suffer from noise contamination, which can undermine model performance. To tackle this challenge, we introduce a novel Synergy-Guided Regional Supervision of Pseudo Labels (SGRS-Net) framework. Built upon the mean teacher network, we employ a Mix Augmentation module to enhance the unlabeled data. By evaluating the synergy before and after augmentation, we strategically partition the pseudo labels into distinct regions. Additionally, we introduce a Region Loss Evaluation module to assess the loss across each delineated area. Extensive experiments conducted on the LA dataset have demonstrated superior performance over state-of-the-art techniques, underscoring the efficiency and practicality of our framework.

Paper Structure

This paper contains 29 sections, 20 equations, 6 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Relate Work
  3. Medical Image Segmentation
  4. Semi-supervised Medical Image Segmentation
  5. Method
  6. Overall architecture design
  7. Pseudo Label Generation Module
  8. Mix Augmentation Module
  9. Synergy Evaluation Module
  10. Regional Loss Evaluation Module
  11. Learning from $\mathcal{GT}$
  12. Learning from $\mathrm{\Omega}$
  13. Learning from $\mathrm{\Theta}$
  14. Total Loss Funciton
  15. Experience
  16. ...and 14 more sections

Figures (6)

  • Figure 1: Overview of our proposed SGRS-Net. The teacher network updates its parameters from the student network using exponential moving averages (EMA). The "MA" denotes the Mix Augmentation Module, and the "RLE" represents the Regional Loss Evaluation module.
  • Figure 2: Visualization of the segmentations results from different methods on the LA dataset. The red lines denote the boundary of GT and the green lines denote the boundary of predictions.
  • Figure 3: Visualization of the segmentations results from different methods on the Pancreas-CT dataset. The red lines denote the boundary of GT and the green lines denote the boundary of predictions.
  • Figure 4: Effect of the loss function for different regions.
  • Figure 5: Effect of the loss function for different regions.
  • ...and 1 more figures