Related Experiment Video
Updated: Jun 14, 2026
09:41
Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping
Published on: April 21, 2023
1.6K
A Deep Learning Approach to Classify Surgical Skill in Microsurgery Using Force Data from a Novel Sensorised Surgical
Jialang Xu1,2, Dimitrios Anastasiou1,2, James Booker1,3
1Wellcome/EPSRC Centre for Interventional and Surgical Sciences, University College London, London W1W 7TY, UK.
Sensors (Basel, Switzerland)
|November 14, 2023
Summary
This study introduces deep learning models to assess surgical skill using force data from a sensorized glove. Convolutional Deep Neural Network (CLDNN) and Temporal Convolutional Network (TCN) models achieved high accuracy, demonstrating the value of force data in skill evaluation.
Area of Science:
- Surgical Skill Assessment
- Medical Robotics and Technology
- Machine Learning in Healthcare
Background:
- Microsurgery is a technically demanding field requiring objective surgical skill assessment.
- Interaction forces between surgical tools and tissues are critical indicators of surgical proficiency.
- Existing methods for skill assessment may lack objective, quantitative metrics.
Purpose of the Study:
- To evaluate the effectiveness of deep learning architectures for classifying surgical skill levels.
- To investigate the utility of force data captured by a novel sensorized surgical glove.
- To propose and validate data augmentation techniques to improve model performance.
Main Methods:
- Six deep learning architectures (LSTM, GRU, Bi-LSTM, CLDNN, TCN, Transformer) were employed.
- Force data from a sensorized surgical glove during a microsurgical task was utilized.
- Six data augmentation techniques were applied to enhance model training.
- Quantitative and qualitative analyses, including cross-validation and interpretable visualizations, were performed.
Main Results:
- The Convolutional Deep Neural Network (CLDNN) achieved 96.16% accuracy.
- The Temporal Convolutional Network (TCN) achieved 97.45% accuracy.
- Both CLDNN and TCN demonstrated superior performance in surgical skill classification.
Conclusions:
- Deep learning models, particularly CLDNN and TCN, are effective in classifying surgical skill levels.
- Force data acquired via a sensorized surgical glove provides valuable insights into surgical skill.
- The findings support the integration of sensorized tools and AI for objective surgical training and evaluation.
