Related Experiment Video
Updated: Jul 30, 2025
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
457
CVCL: Context-aware Voxel-wise Contrastive Learning for label-efficient multi-organ segmentation.
1Institute of Medical Robotics, School of Biomedical Engineering, Shanghai Jiao Tong University, No. 800, Dongchuan Road, Shanghai, 200240, China.
Computers in Biology and Medicine
|May 15, 2023
Summary
This study introduces Context-aware Voxel-wise Contrastive Learning (CVCL) for multi-organ segmentation. CVCL effectively utilizes both labeled and unlabeled data, improving segmentation performance in label-scarce scenarios.
Area of Science:
- Medical Imaging
- Computer Vision
- Machine Learning
Background:
- Supervised deep learning significantly enhances multi-organ segmentation but requires extensive labeled data, limiting practical applications in disease diagnosis and treatment planning.
- Acquiring expert-annotated, densely labeled multi-organ datasets is challenging, driving interest in label-efficient segmentation techniques like partially supervised and semi-supervised learning.
- Existing label-efficient methods often overlook or inadequately address challenging unlabeled regions during training, hindering optimal performance.
Purpose of the Study:
- To propose a novel Context-aware Voxel-wise Contrastive Learning (CVCL) method for multi-organ segmentation.
- To leverage both labeled and unlabeled data effectively in label-scarce datasets.
- To improve the performance of multi-organ segmentation, particularly in scenarios with limited annotations.
Main Methods:
- Developed a novel Context-aware Voxel-wise Contrastive Learning (CVCL) approach.
- Designed CVCL to fully exploit information from both labeled and unlabeled data within medical image datasets.
- Focused on enhancing segmentation accuracy in the context of multi-organ identification.
Main Results:
- Experimental results indicate that the proposed CVCL method achieves superior performance compared to existing state-of-the-art methods.
- Demonstrated the effectiveness of CVCL in improving multi-organ segmentation accuracy, especially with limited labeled data.
- Validated the approach on challenging medical image segmentation tasks.
Conclusions:
- The novel CVCL method offers a promising solution for label-efficient multi-organ segmentation.
- CVCL effectively addresses the limitations of previous methods by utilizing unlabeled data more comprehensively.
- The findings suggest significant potential for CVCL in practical medical image analysis for diagnosis and treatment planning.
