Related Experiment Video
Updated: Jul 8, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
MSDE-Net: A Multi-Scale Dual-Encoding Network for Surgical Instrument Segmentation
IEEE Journal of Biomedical and Health Informatics
|December 20, 2023
Summary
This study introduces MSDE-Net, a novel deep learning model for precise surgical instrument segmentation. The network effectively combines local and global features, significantly improving accuracy in minimally invasive surgery.
Area of Science:
- Medical Imaging
- Computer Vision
- Surgical Robotics
Background:
- Minimally invasive surgery requires precise image segmentation for safety and efficiency.
- Accurate segmentation of surgical instruments is challenging due to complex surgical environments.
Purpose of the Study:
- To introduce a novel multiscale dual-encoding segmentation network (MSDE-Net) for automatic and precise segmentation of surgical instruments.
Main Methods:
- MSDE-Net utilizes a dual-branch encoder (CNN and transformer) for local and global feature extraction.
- Incorporates an attention fusion block (AFB) for feature complementarity.
- Employs multilayer context fusion (MCF) and multi-receptive field fusion (MRF) blocks to enhance feature extraction capabilities.
Main Results:
- MSDE-Net demonstrated superior performance in segmenting surgical instruments compared to existing methods.
- Experiments were conducted on two publicly available datasets for validation.
Conclusions:
- The proposed MSDE-Net effectively addresses the challenges in surgical instrument segmentation.
- This network offers a promising solution for enhancing precision in minimally invasive surgical procedures.
