Related Experiment Video
Updated: Jul 7, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
Enhanced U-Net with GridMask (EUGNet): A Novel Approach for Robotic Surgical Tool Segmentation
Mostafa Daneshgar Rahbar1, Seyed Ziae Mousavi Mojab2
1Department of Electrical and Computer Engineering, Lawrence Technological University, Southfield, MI 48075, USA.
Journal of Imaging
|December 22, 2023
Summary
Enhanced U-Net with GridMask (EUGNet) improves medical image segmentation accuracy and robustness. This novel approach significantly boosts inference speed, crucial for real-time robotic surgery applications.
Area of Science:
- Medical Image Analysis
- Computer Vision
- Machine Learning
Background:
- U-Net architecture limitations in medical image segmentation.
- Need for robust segmentation in robotic surgery.
Purpose of the Study:
- Introduce Enhanced U-Net with GridMask (EUGNet) for improved medical image segmentation.
- Address U-Net's limitations using GridMask augmentation.
Main Methods:
- Incorporation of GridMask augmentation for pixel manipulation.
- Deep contextual encoder, residual connections, class-balancing loss, adaptive feature fusion.
- Evaluation on a dataset of robotic surgical scenarios and instruments.
Main Results:
- 1.6% increase in balanced accuracy for foreground segmentation.
- 1.7% improvement in Intersection over Union (IoU).
- 1.7% improvement in mean Dice Similarity Coefficient (DSC).
- Reduced inference speed from 0.163 ms to 0.097 ms.
Conclusions:
- EUGNet enhances segmentation accuracy and robustness in medical imaging.
- GridMask augmentation improves adaptability to occlusions and local features.
- Significant improvements in accuracy metrics and inference speed for real-time applications.
