Related Experiment Video
Updated: Nov 1, 2025
06:18
Author Spotlight: Segmentation and VR for Advanced Neurovascular Interventions
Published on: April 5, 2024
1.3K
Accurate instance segmentation of surgical instruments in robotic surgery: model refinement and cross-dataset
Xiaowen Kong1,2, Yueming Jin3, Qi Dou3,4
1Department of Precision Machinery and Precision Instrumentation, University of Science and Technology of China, Hefei, China.
Summary
This study introduces an improved Mask R-CNN model for precise surgical instrument segmentation in robot-assisted surgery. The model achieves state-of-the-art results and demonstrates effective cross-dataset generalization capabilities.
Area of Science:
- Computer Vision
- Robotic Surgery
- Medical Imaging Analysis
Background:
- Context awareness is crucial in robot-assisted minimally invasive surgery.
- Accurate segmentation of surgical instruments enhances surgical precision and safety.
Purpose of the Study:
- To develop an instance segmentation model for accurate segmentation and type identification of surgical instruments.
- To improve context awareness in robot-assisted minimally invasive surgery.
Main Methods:
- Re-formulated instrument segmentation as an instance segmentation problem.
- Optimized Mask R-CNN with anchor optimization and improved Region Proposal Network.
- Conducted cross-dataset evaluation using various sampling strategies.
Main Results:
- Achieved new state-of-the-art performance on two segmentation tasks in the MICCAI 2017 Endoscopic Vision Challenge dataset.
- Demonstrated improved performance through cross-dataset training compared to testing on a public dataset.
Conclusions:
- The proposed instance segmentation network effectively segments surgical instruments.
- The model exhibits cross-dataset generalization capabilities, further enhanced by cross-dataset training.
- Empirical findings offer guidance for efficient annotation cost allocation in new dataset labeling.
