Related Experiment Video
Updated: Aug 23, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.9K
UCR-Net: U-shaped context residual network for medical image segmentation
Qi Sun1, Mengyun Dai1, Ziyang Lan1
1Digital Fujian Research Institute of Big Data for Agriculture and Forestry, College of Computer and Information Sciences, Fujian Agriculture and Forestry University, Fuzhou 350002, China.
Computers in Biology and Medicine
|October 28, 2022
Summary
UCR-Net enhances medical image segmentation by preserving spatial information lost in U-Net. This novel network captures richer context and high-level features, improving segmentation accuracy in clinical applications.
Area of Science:
- Biomedical image analysis
- Deep learning for medical imaging
- Computer-aided diagnosis
Background:
- Medical image segmentation is crucial for clinical applications.
- U-Net is a popular deep learning model for segmentation.
- U-Net can lose spatial information due to pooling and downsampling.
Purpose of the Study:
- To propose a novel U-shaped context residual network (UCR-Net) for improved medical image segmentation.
- To address the loss of spatial information in traditional U-Net architectures.
- To enhance the capture of context and high-level features in medical images.
Main Methods:
- Developed UCR-Net, an encoder-decoder framework.
- Incorporated Context Attention Exploration (CAE) modules to capture multi-scale context.
- Integrated a Global and Spatial Attention (GSA) module for enhanced global and semantic features.
Main Results:
- UCR-Net effectively recovers high-level semantic features.
- The network fuses context attention information from CAE and GSA modules.
- Experiments showed UCR-Net outperforms U-Net and other advanced methods on retinal vessel, femoropopliteal artery stent, and polyp datasets.
Conclusions:
- UCR-Net offers superior performance in medical image segmentation compared to existing methods.
- The proposed architecture effectively preserves and utilizes spatial and contextual information.
- UCR-Net demonstrates significant potential for clinical applications requiring accurate image segmentation.
