Online Bayesian optimization of vagus nerve stimulation.

Area of Science:

• Bioelectronic Medicine
• Neuroscience
• Machine Learning

Background:

• Neuromodulation therapies use electrical waveforms to alter neural signals for therapeutic effects.
• Optimizing stimulation parameters for individual patients is complex and time-consuming.
• Current methods risk suboptimal results or adverse effects due to large parameter spaces.

Purpose of the Study:

• To develop a machine learning framework for efficient optimization of neural stimulation parameters.
• To reduce the number of stimulation steps required for waveform optimization.
• To enable personalized precision dosing in neuromodulation therapies.

Main Methods:

• Utilized Bayesian optimization for a flexible, data-driven approach to parameter searching.
• Developed a framework for acquiring and processing data to model neural and physiological responses.
• Applied the method to vagus nerve stimulation in porcine subjects to control heart rate and neural activation.

Main Results:

• The Bayesian optimization framework rapidly converged on optimal parameters for target heart rates and neural B-fiber activation.
• Demonstrated effectiveness despite significant inter-subject variability in porcine models.
• Achieved optimized stimulation waveforms in real-time with substantially fewer stimulations than traditional methods.

Conclusions:

• The proposed machine learning approach significantly accelerates the optimization of complex neural stimulation parameters.
• Uncertainty estimates within the framework enhance safety by avoiding stimulations outside therapeutic ranges.
• This method facilitates real-time, personalized neuromodulation for improved patient outcomes in bioelectronic medicine.