Related Experiment Video
Updated: Sep 10, 2025
Robotic-Guided Stereoelectroencephalography for Invasive Epilepsy Monitoring
Published on: June 13, 2025
Breakthrough Recognition in Robotic-Assisted UBE Surgery Based on Force Sensing and VAE-MLP
Xuquan Ji1, Yonghong Zhang2, Yuanyuan Zhu3
1Key Laboratory of Biomechanics and Mechanobiology (Beihang University), Ministry of Education, Key Laboratory of Innovation and Transformation of Advanced Medical Devices, Ministry of Industry and Information Technology, National Medical Innovation Platform for Industry-Education Integration in Advanced Medical Devices (Interdiscipline of Medicine and Engineering), School of Biological Science and Medical Engineering, Beihang University, Beijing, China.
This study introduces a novel force and Variational Autoencoder-Multilayer Perceptron (VAE-MLP) method for real-time penetration recognition in robotic-assisted unilateral biportal endoscopic (UBE) surgery. The technique enhances safety by accurately detecting drill penetration, minimizing nerve damage during procedures.
Area of Science:
- Robotics
- Surgical Technology
- Machine Learning
Background:
- Robotic-assisted unilateral biportal endoscopic (UBE) surgery offers improved accuracy and safety over traditional open procedures.
- Accurate penetration recognition during ultrasonic drilling is a critical challenge in robotic-assisted UBE surgery.
Purpose of the Study:
- To develop a real-time penetration recognition method for ultrasonic drilling in robotic-assisted UBE surgery.
- To enhance the safety and efficiency of UBE procedures by providing reliable feedback on drill penetration.
Main Methods:
- A novel method combining force signals and a Variational Autoencoder-Multilayer Perceptron (VAE-MLP) model was proposed.
- Force signals were collected during ultrasonic drilling, denoised using Kalman filtering, and processed by VAE-MLP for feature extraction and classification.
- The system achieved real-time penetration recognition through this integrated approach.
Main Results:
- The VAE-MLP method demonstrated superior accuracy (99.32%) compared to traditional time-series classification algorithms (95.90%).
- The proposed method achieved faster inference speeds (17 ms) than existing algorithms (33 ms).
- Robotic ex vivo bone experiments confirmed the method's effectiveness and reliability.
Conclusions:
- The developed force and VAE-MLP framework provides fast and accurate penetration detection capabilities.
- This technology offers a reliable solution for minimizing nerve damage during robotic-assisted UBE surgery.
- The method represents a significant advancement in surgical robotics and intraoperative safety.
