Related Experiment Video
Updated: Sep 7, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.9K
Semi-supervised medical image segmentation via uncertainty rectified pyramid consistency
Xiangde Luo1, Guotai Wang1, Wenjun Liao2
1School of Mechanical and Electrical Engineering, University of Electronic Science and Technology of China, Chengdu, China; Shanghai AI Lab, Shanghai, China.
Medical Image Analysis
|June 22, 2022
Summary
Uncertainty Rectified Pyramid Consistency (URPC) improves semi-supervised medical image segmentation by using unlabeled data. This method achieves better results than existing approaches with a simpler pipeline.
Area of Science:
- Medical Image Analysis
- Computer Vision
- Machine Learning
Background:
- Convolutional Neural Networks (CNNs) excel at medical image segmentation but require extensive labeled data.
- Acquiring labeled medical images is costly and time-consuming, hindering model training.
- Semi-supervised learning offers a solution by leveraging abundant unlabeled data alongside limited labeled samples.
Purpose of the Study:
- To introduce a novel, efficient semi-supervised approach for medical image segmentation.
- To address the challenge of limited labeled data in training deep learning models for medical imaging.
- To enhance segmentation performance by effectively utilizing unlabeled medical image datasets.
Main Methods:
- Developed Uncertainty Rectified Pyramid Consistency (URPC), a consistency regularization method.
- Employed a pyramid-prediction network generating multi-scale segmentation predictions.
- Implemented multi-scale uncertainty rectification to refine consistency loss by mitigating outlier pixel influence.
Main Results:
- URPC demonstrated significant performance improvements in medical image segmentation tasks by incorporating unlabeled data.
- The method achieved superior or comparable results against five established semi-supervised techniques.
- URPC offers a simpler and more efficient pipeline compared to existing approaches.
Conclusions:
- URPC effectively enhances semi-supervised medical image segmentation performance.
- The proposed method provides a valuable tool for leveraging unlabeled data in medical imaging.
- A publicly available codebase (https://github.com/HiLab-git/SSL4MIS) is provided to foster further research.
