Multistep prediction of physiological tremor based on machine learning for robotics assisted microsurgery

Area of Science:

• Robotics
• Biomedical Engineering
• Machine Learning

Background:

• Accurate tremor filtering is crucial for robotic-assisted hand-held devices.
• Time-varying phase delays from software and hardware impede device performance.

Purpose of the Study:

• To develop a method for multistep tremor prediction that overcomes time-varying delays.
• To enhance the performance of robotics-assisted hand-held devices.

Main Methods:

• Formulation of a moving window-based least squares support vector machine (LS-SVM) approach.
• Utilizing kernel-learning techniques for tremor prediction.
• No prior knowledge of prediction horizon is required.

Main Results:

• The proposed LS-SVM method effectively predicts tremor and compensates for time-varying delays.
• Evaluated through simulations and experiments using tremor data from surgeons and novice subjects.
• Demonstrated superior performance compared to state-of-the-art techniques.

Conclusions:

• The moving window LS-SVM approach is suitable for tremor compensation in robotic devices.
• The method offers improved performance and overcomes limitations of existing techniques.