Bayesian optimization of peripheral intraneural stimulation protocols to evoke distal limb movements

Area of Science:

• Biomedical Engineering
• Neuroscience
• Robotics

Background:

• Developing effective motor neuroprostheses necessitates identifying optimal stimulation protocols for desired movements.
• Manual search for these protocols is time-consuming and often yields suboptimal results due to the need for personalized parameters.
• Efficient methods are required to automate the tuning of stimulation parameters for various motor functions.

Purpose of the Study:

• To present an algorithm that efficiently tunes peripheral intraneural stimulation protocols for eliciting functionally relevant distal limb movements.
• To demonstrate the algorithm's capability in automating the configuration of motor neuroprostheses.
• To establish a translational framework for clinical applications of neuroprosthetics.

Main Methods:

• Developed an algorithm utilizing Bayesian optimization (BO) with multi-output Gaussian Processes (GPs).
• Defined objective functions based on coordinated muscle recruitment for movement control.
• Applied the algorithm offline to rat (walking) and monkey (grasping) data, followed by a preliminary online test in a monkey.

Main Results:

• Offline, optimal intraneural stimulation protocols were rapidly identified for diverse motor functions in both rat and monkey models.
• The algorithm converged to stimuli evoking functionally consistent movements efficiently, using approximately 20% of the search space.
• Online validation showed the algorithm quickly guided to effective stimuli for hand gestures, with potential for further refinement.

Conclusions:

• Bayesian optimization reliably and efficiently automates the tuning of peripheral neurostimulation protocols.
• The proposed method provides a translational framework for configuring peripheral motor neuroprostheses in clinical settings.
• This approach has potential applications in optimizing motor functions across various stimulation modalities.