Related Experiment Video
Updated: Aug 22, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.9K
An attention-guided network for surgical instrument segmentation from endoscopic images
Lei Yang1, Yuge Gu1, Guibin Bian2
1School of Electrical and Information Engineering, Zhengzhou University, Henan Province, 450001, China; Robot Perception and Control Engineering Laboratory of Henan Province, 450001, China.
Computers in Biology and Medicine
|November 10, 2022
Summary
This study introduces an improved surgical instrument segmentation network to enhance robot-assisted surgery. The new model effectively processes local features and global contexts, achieving high accuracy in segmenting surgical instruments.
Area of Science:
- Medical Image Analysis
- Computer Vision
- Robotics
Background:
- Accurate surgical instrument segmentation is crucial for robot-assisted surgery.
- Existing methods like U-Net variants struggle with local feature processing and class imbalance.
- Enhanced segmentation aids surgeons in precise operation judgment.
Purpose of the Study:
- To propose an effective surgical instrument segmentation network for end-to-end detection.
- To address limitations in local feature processing and class imbalance in current models.
- To improve the accuracy and reliability of surgical instrument segmentation in robotic procedures.
Main Methods:
- An encoder-decoder network incorporating a residual path for enhanced low-level feature propagation.
- Integration of a non-local attention block at the bottleneck for global context acquisition.
- Implementation of a dual-attention module (DAM) to fuse high-level and low-level features for precise instrument highlighting.
Main Results:
- Achieved a 95.77% Dice score and 92.13% mIOU on the Kvasir-instrument dataset.
- Attained a 95.60% Dice score and 92.74% mIOU on the Endovis2017 dataset.
- Demonstrated superior performance compared to other advanced segmentation models.
Conclusions:
- The proposed network effectively addresses local feature processing and class imbalance issues.
- The model provides superior surgical instrument segmentation accuracy, aiding robotic surgery.
- This work offers a valuable reference for developing intelligent surgical robots.
