Related Experiment Video
Updated: Jun 25, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.7K
Quality-driven deep cross-supervised learning network for semi-supervised medical image segmentation
Zhenxi Zhang1, Heng Zhou2, Xiaoran Shi1
1The Ministry of Education, Key Laboratory of Electronic Information Counter-measure and Simulation, Xidian University, Xi'an 710071, China; School of Electronic Engineering, Xidian University, Xi'an 710071, China.
Computers in Biology and Medicine
|May 21, 2024
Summary
This study introduces Quality-driven Deep Cross-supervised Learning Network (QDC-Net) for efficient semi-supervised medical image segmentation. QDC-Net improves accuracy by managing sub-network disagreement and enhancing training reliability.
Area of Science:
- Medical Image Analysis
- Machine Learning
- Computer Vision
Background:
- Semi-supervised learning reduces annotation burden in medical image segmentation.
- Existing cross-supervised methods struggle with sub-network disagreement and training efficiency.
Purpose of the Study:
- Introduce a novel Quality-driven Deep Cross-supervised Learning Network (QDC-Net).
- Address challenges in sub-network disagreement and training reliability for semi-supervised medical image segmentation.
Main Methods:
- QDC-Net employs an evidential and a vanilla sub-network to manage disagreement.
- Real-time quality estimation and directional cross-training with directional weights enhance reliability.
- Truncated sample-wise loss weighting mitigates inaccurate predictions.
Main Results:
- QDC-Net demonstrated superior performance in semi-supervised medical image segmentation.
- Experiments on LA and Pancreas-CT datasets confirmed QDC-Net's effectiveness.
- The proposed methods significantly improved segmentation accuracy and training efficiency.
Conclusions:
- QDC-Net offers a robust and efficient solution for semi-supervised medical image segmentation.
- The framework effectively handles sub-network disagreement and improves training reliability.
- QDC-Net represents a significant advancement in automated medical image analysis.
