Related Experiment Video
Updated: Sep 22, 2025
06:48
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
9.0K
FGAM: A pluggable light-weight attention module for medical image segmentation.
Zhongxi Qiu1, Yan Hu1, Jiayi Zhang1
1Department of Computer Science and Engineering, Southern University of Science and Technology, Shenzhen, 51805, Guangdong, China.
Computers in Biology and Medicine
|May 24, 2022
Summary
We introduce the Feature Guided Attention Module (FGAM), a simple, pluggable, and effective tool to enhance medical image segmentation. FGAM improves accuracy by leveraging encoder-decoder features, offering a computationally efficient solution.
Area of Science:
- Medical Imaging
- Computer Vision
- Artificial Intelligence
Background:
- Medical image segmentation is crucial for computer-aided diagnosis and surgery.
- Existing attention modules often suffer from high computational costs and limited framework applicability.
Purpose of the Study:
- To propose a novel, efficient, and versatile attention module for medical image segmentation.
- To address the limitations of current attention mechanisms in terms of computation and applicability.
Main Methods:
- Introduced the Feature Guided Attention Module (FGAM), a pluggable module designed to enhance encoder-decoder networks.
- FGAM utilizes shallow decoder features as a queryable dictionary to extract rich feature representations.
- The module features a parameter-free activator and is removable post-training.
Main Results:
- Demonstrated the effectiveness of FGAM across various encoder-decoder architectures.
- Validated FGAM's performance on five diverse datasets, including publicly available and in-house medical image data.
Conclusions:
- FGAM offers a simple yet effective solution for improving medical image segmentation.
- The module's pluggable nature and parameter-free design enhance its applicability and efficiency in deep learning frameworks.
