Related Experiment Video
Updated: Jul 30, 2025
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
457
A novel medical image segmentation approach by using multi-branch segmentation network based on local and global
Shangzhu Jin1, Sheng Yu2, Jun Peng3
1Information Office, Chongqing University of Science and Technology, Chongqing, 401331, China.
Scientific Reports
|May 15, 2023
Summary
This study introduces MBSNet, a novel medical image segmentation network that is lighter, faster, and more accurate than existing methods. MBSNet effectively captures both local and global image information for improved segmentation performance.
Area of Science:
- Medical Image Analysis
- Computer Vision
- Deep Learning
Background:
- Existing medical image segmentation methods often neglect network parameters and real-time performance.
- Deep encoders and focus on local information limit accuracy, especially in boundary regions.
- Global information is crucial for accurate medical image segmentation.
Purpose of the Study:
- To propose a novel multi-branch medical image segmentation network (MBSNet).
- To address limitations in network efficiency and boundary segmentation accuracy.
- To improve the capture of both local and global image features.
Main Methods:
- Designed a multi-branch network (MBSNet) with parallel residual mixer (PRM) and dilate convolution blocks.
- Incorporated SE-Block and spatial attention modules to enhance features.
- Utilized a cross-fusion method to combine features from different branches and layers.
Main Results:
- MBSNet demonstrated superior performance across five datasets (ISIC2018, Kvasir, BUSI, COVID-19, LGG).
- Achieved significantly lower FLOPs (10.68G) and higher F1-Score ([Formula: see text]) compared to UNet++ (216.55G FLOPs).
- TOPSIS analysis confirmed MBSNet's better overall performance based on F1-Score, IOU, and G-mean.
Conclusions:
- MBSNet offers a lighter, faster, and more accurate solution for medical image segmentation.
- The proposed architecture effectively balances local and global feature extraction.
- MBSNet represents a significant advancement in efficient and accurate medical image segmentation.
