Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DiffRect: Latent Diffusion Label Rectification for Semi-supervised Medical Image Segmentation

Xinyu Liu, Wuyang Li, Yixuan Yuan

TL;DR

The paper tackles semi-supervised medical image segmentation under label scarcity by addressing pseudo label reliability and latent distribution misalignment. It introduces DiffRect, a diffusion-based framework with two modules: LCC calibrates class relationships in pseudo labels, while LFR uses a latent DDPM to progressively rectify label distributions from strong to weak to ground-truth supervision. The method demonstrates substantial improvements on ACDC, MS-CMRSEG 2019, and Decathlon Prostate, including achieving 82.40% Dice with only 1% labeled data on ACDC and surpassing state-of-the-art methods by notable margins. The approach offers a robust, generalizable way to utilize unlabeled data in medical image segmentation with a publicly released implementation.

Abstract

Semi-supervised medical image segmentation aims to leverage limited annotated data and rich unlabeled data to perform accurate segmentation. However, existing semi-supervised methods are highly dependent on the quality of self-generated pseudo labels, which are prone to incorrect supervision and confirmation bias. Meanwhile, they are insufficient in capturing the label distributions in latent space and suffer from limited generalization to unlabeled data. To address these issues, we propose a Latent Diffusion Label Rectification Model (DiffRect) for semi-supervised medical image segmentation. DiffRect first utilizes a Label Context Calibration Module (LCC) to calibrate the biased relationship between classes by learning the category-wise correlation in pseudo labels, then apply Latent Feature Rectification Module (LFR) on the latent space to formulate and align the pseudo label distributions of different levels via latent diffusion. It utilizes a denoising network to learn the coarse to fine and fine to precise consecutive distribution transportations. We evaluate DiffRect on three public datasets: ACDC, MS-CMRSEG 2019, and Decathlon Prostate. Experimental results demonstrate the effectiveness of DiffRect, e.g. it achieves 82.40\% Dice score on ACDC with only 1\% labeled scan available, outperforms the previous state-of-the-art by 4.60\% in Dice, and even rivals fully supervised performance. Code is released at \url{https://github.com/CUHK-AIM-Group/DiffRect}.

DiffRect: Latent Diffusion Label Rectification for Semi-supervised Medical Image Segmentation

TL;DR

The paper tackles semi-supervised medical image segmentation under label scarcity by addressing pseudo label reliability and latent distribution misalignment. It introduces DiffRect, a diffusion-based framework with two modules: LCC calibrates class relationships in pseudo labels, while LFR uses a latent DDPM to progressively rectify label distributions from strong to weak to ground-truth supervision. The method demonstrates substantial improvements on ACDC, MS-CMRSEG 2019, and Decathlon Prostate, including achieving 82.40% Dice with only 1% labeled data on ACDC and surpassing state-of-the-art methods by notable margins. The approach offers a robust, generalizable way to utilize unlabeled data in medical image segmentation with a publicly released implementation.

Abstract

Semi-supervised medical image segmentation aims to leverage limited annotated data and rich unlabeled data to perform accurate segmentation. However, existing semi-supervised methods are highly dependent on the quality of self-generated pseudo labels, which are prone to incorrect supervision and confirmation bias. Meanwhile, they are insufficient in capturing the label distributions in latent space and suffer from limited generalization to unlabeled data. To address these issues, we propose a Latent Diffusion Label Rectification Model (DiffRect) for semi-supervised medical image segmentation. DiffRect first utilizes a Label Context Calibration Module (LCC) to calibrate the biased relationship between classes by learning the category-wise correlation in pseudo labels, then apply Latent Feature Rectification Module (LFR) on the latent space to formulate and align the pseudo label distributions of different levels via latent diffusion. It utilizes a denoising network to learn the coarse to fine and fine to precise consecutive distribution transportations. We evaluate DiffRect on three public datasets: ACDC, MS-CMRSEG 2019, and Decathlon Prostate. Experimental results demonstrate the effectiveness of DiffRect, e.g. it achieves 82.40\% Dice score on ACDC with only 1\% labeled scan available, outperforms the previous state-of-the-art by 4.60\% in Dice, and even rivals fully supervised performance. Code is released at \url{https://github.com/CUHK-AIM-Group/DiffRect}.
Paper Structure (12 sections, 14 equations, 1 figure, 5 tables)

This paper contains 12 sections, 14 equations, 1 figure, 5 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Preliminary: Conditional DDPM
  4. Label Context Calibration Module (LCC)
  5. Latent Feature Rectification Module (LFR)
  6. Loss Function
  7. Experiments
  8. Experimental Setup
  9. Comparison with State-of-the-art Methods
  10. Further Analysis
  11. Conclusion
  12. Acknowledgments.

Figures (1)

  • Figure 1: Overall framework of DiffRect. (a) Label Context Calibration Module (LCC). (b) Latent Feature Rectification Module (LFR). (c) Segmentation Network.