Related Experiment Video
Updated: Jan 15, 2026
12:25
C-arm-Free Simultaneous OLIF51 and Percutaneous Pedicle Screw Fixation in a Single Lateral Position
Published on: September 16, 2022
4.5K
Multisensory Integration for Identifying the Milling States in Robot-Assisted Cervical Laminectomy
Chao Sun1, Yingjie Zheng1, Junfei Hu2
1Department of Orthopedic Surgery, Key Laboratory of Spine and Spinal Cord, Tianjin Medical University General Hospital, Tianjin, China.
Orthopaedic Surgery
|October 9, 2025
Summary
Integrating tactile and auditory perception significantly improves high-speed bur milling state detection accuracy in robot-assisted spinal surgery. The Long Short-Term Memory (LSTM) model achieved 99.32% accuracy, enhancing patient safety.
Area of Science:
- Robotics in Surgery
- Biomedical Engineering
- Surgical Navigation
Background:
- Precise identification of high-speed bur milling states is critical for patient safety in spinal surgery.
- Robot-assisted procedures require enhanced sensory feedback for accurate tool-tissue interaction.
- Current methods may lack the precision needed for complex bone milling tasks.
Purpose of the Study:
- To investigate the integration of tactile and auditory perception for improved milling state detection.
- To evaluate the accuracy of machine learning and deep learning models in identifying milling states during robot-assisted cervical laminectomy.
- To develop a multi-perception fusion framework for enhanced surgical feedback.
Main Methods:
- Vibration and sound signals were collected using acceleration sensors and microphones during sheep cervical laminectomy.
- Seven distinct milling states were defined, including varying depths in cortical and cancellous bone and boundary conditions.
- Fast Fourier Transform (FFT) was used to extract features, creating an 18-dimensional multi-perception spatial vector for analysis with machine learning (SVM, KNN, NB, LDA, DT) and deep learning (LSTM) models.
Main Results:
- The Long Short-Term Memory (LSTM) model, trained on 6600 data sets, demonstrated superior accuracy and stability compared to traditional machine learning algorithms.
- A single-layer LSTM with 12 memory units achieved an accuracy of 99.32% in identifying cervical lamina milling states.
- The multi-perception fusion approach significantly enhanced the precision of milling state detection.
Conclusions:
- A multi-perception fusion framework integrating tactile and auditory data was successfully developed.
- The LSTM model accurately identified cervical vertebra milling states, offering a promising sensory feedback mechanism.
- This approach can advance the capabilities of operational spinal surgery robots, improving surgical outcomes and safety.
