Related Experiment Video
Updated: May 3, 2026
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
42.4K
CUAMT: A MRI semi-supervised medical image segmentation framework based on contextual information and mixed
Hanguang Xiao1, Yangjian Wang1, Shidong Xiong1
1School of Artificial Intelligence, Chongqing University of Technology, Chongqing 401135, China.
Computer Methods and Programs in Biomedicine
|April 30, 2025
Summary
This study introduces a new semi-supervised medical image segmentation framework that improves boundary accuracy by utilizing contextual information and hybrid uncertainty. The model effectively enhances segmentation performance, outperforming existing methods on benchmark datasets.
Area of Science:
- Medical Image Analysis
- Machine Learning
- Computer Vision
Background:
- Semi-supervised learning reduces data labeling in medical image segmentation.
- Existing methods struggle with volumetric contextual information, leading to ambiguous boundaries.
Purpose of the Study:
- To propose a novel semi-supervised medical image segmentation framework.
- To enhance boundary definition by effectively utilizing contextual and uncertainty information.
Main Methods:
- Developed a hybrid uncertainty network (CUAMT) incorporating a contextual information extraction (CIE) module.
- Proposed a hybrid uncertainty module (HUM) to focus on segmentation boundary details.
- CIE extracts multi-scale semantic features to learn image context.
Main Results:
- Achieved Dice: 89.84%, Jaccard: 79.89%, 95HD: 8.73 on left atrial and brain tumor datasets.
- Demonstrated significant outperformance compared to state-of-the-art semi-supervised methods.
- Validated the effectiveness of the proposed CIE and HUM strategies.
Conclusions:
- The proposed framework offers an effective approach to semi-supervised medical image segmentation.
- The integration of contextual and uncertainty modules significantly improves segmentation accuracy, especially at boundaries.
