Related Experiment Video
Updated: Aug 30, 2025
05:56
Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application
Published on: April 14, 2023
2.6K
Polyp segmentation with consistency training and continuous update of pseudo-label.
Hyun-Cheol Park1, Sahadev Poudel1, Raman Ghimire1
1Department of IT Convergence Engineering, Gachon University, Seongnam, 13120, South Korea.
Scientific Reports
|August 26, 2022
Summary
This study introduces a semi-supervised learning approach to improve polyp image segmentation by leveraging unlabeled data. The novel method enhances segmentation performance, addressing challenges in medical dataset labeling.
Area of Science:
- Medical image analysis
- Computer vision
- Machine learning
Background:
- Supervised learning has advanced polyp segmentation.
- Acquiring large labeled medical datasets is difficult.
- Semi-supervised methods can utilize unlabeled data.
Purpose of the Study:
- To improve polyp image segmentation using semi-supervised learning.
- To address the challenge of limited labeled medical data.
- To enhance the performance of polyp segmentation models.
Main Methods:
- An encoder-decoder network for polyp segmentation.
- A teacher-student model with exponential averaging.
- Consistency regularization on perturbed unlabeled data.
- An improved pseudo-labeling technique with continuous updates.
Main Results:
- The proposed method achieves better results in semi-supervised settings.
- Demonstrated efficacy across various polyp datasets.
- Successful propagation of information from unlabeled data.
Conclusions:
- Semi-supervised learning effectively enhances polyp segmentation.
- The proposed techniques improve model performance with limited labels.
- This approach offers a viable solution for medical image segmentation challenges.
