Related Experiment Video
Updated: Aug 16, 2025
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
480
Transformer-based multilevel region and edge aggregation network for magnetic resonance image segmentation.
Shaolong Chen1, Lijie Zhong2, Changzhen Qiu1
1School of Electronics and Communication Engineering, Sun Yat-sen University, Shenzhen, 518107, China.
Computers in Biology and Medicine
|December 21, 2022
Summary
This study introduces a new transformer-based network for magnetic resonance (MR) image segmentation, improving edge and region feature aggregation for better accuracy. The novel approach enhances segmentation quality in medical imaging applications.
Area of Science:
- Medical imaging
- Artificial intelligence
- Computer vision
Background:
- Convolutional neural networks (CNNs) face limitations in capturing long-range dependencies for MR image segmentation due to their inherent locality.
- Existing methods for MR image edge segmentation struggle to effectively integrate region and edge information.
Purpose of the Study:
- To propose a novel transformer-based multilevel region and edge aggregation network for enhanced MR image segmentation.
- To address the limitations of CNNs in modeling long-range information for improved segmentation accuracy.
Main Methods:
- A dual-branch module extracts multilevel region and edge features.
- Multiple transformer-based inference modules aggregate these features to create complementary representations.
- An attention feature selection module integrates complementary and level-specific features for decoding.
Main Results:
- The proposed method achieved a Dice score of 93.2% on the ASC dataset and 91.9% on the IPFP dataset.
- Significant improvements were observed compared to other 2D segmentation methods, with a 0.6% increase for ASC and 3.0% for IPFP.
Conclusions:
- The transformer-based multilevel network effectively aggregates region and edge information for superior MR image segmentation.
- This approach represents a significant advancement in leveraging transformer architectures for medical image analysis.
