Related Experiment Video
Updated: Jul 25, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
WRANet: wavelet integrated residual attention U-Net network for medical image segmentation
Yawu Zhao1, Shudong Wang1, Yulin Zhang2
1School of Computer Science and Technology, China University of Petroleum, Qingdao, Shandong China.
Complex & Intelligent Systems
|June 26, 2023
Summary
This study introduces the Wavelet Residual Attention Network (WRANet) for robust medical image segmentation. WRANet effectively reduces noise and improves segmentation accuracy for conditions like aneurysms and polyps.
Area of Science:
- Medical image analysis
- Deep learning for medical imaging
Background:
- Deep convolutional neural networks (CNNs) excel in medical image segmentation but are vulnerable to noise and training instability.
- Noise interference can significantly degrade CNN performance, and deeper networks face gradient issues.
Purpose of the Study:
- To enhance the robustness and segmentation performance of deep learning models for medical images.
- To address noise susceptibility and gradient problems in CNNs for medical image segmentation.
Main Methods:
- Proposed Wavelet Residual Attention Network (WRANet) for medical image segmentation.
- Replaced standard downsampling with discrete wavelet transform to decompose features and remove high-frequency noise.
- Integrated an attention mechanism to mitigate feature loss.
Main Results:
- Achieved 78.99% Dice score for aneurysm segmentation and 88.89% Dice score for polyp segmentation.
- Demonstrated competitive performance compared to state-of-the-art techniques.
- WRANet effectively reduces noise and improves segmentation accuracy.
Conclusions:
- WRANet offers a robust and effective solution for medical image segmentation.
- The proposed method shows significant improvements in accuracy and noise resistance.
- WRANet is a competitive alternative for medical image segmentation tasks.
