From dawn till dusk: Time-adaptive bayesian optimization for neurostimulation.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computational Biology

Background:

• Neuromodulation therapies require precise stimulation settings for optimal efficacy.
• Current optimization methods often assume static parameters, neglecting time-dependent factors like disease progression or biological rhythms.
• These time-varying factors can significantly impact therapy outcomes and necessitate adaptive strategies.

Purpose of the Study:

• To develop and evaluate a novel time-varying Bayesian optimization (TV-BayesOpt) algorithm for dynamically tracking optimal neuromodulation parameters.
• To assess the algorithm's performance in handling gradual and periodic variations in optimal stimulation settings.
• To provide a flexible framework for optimizing neuromodulation therapies for neurological and psychiatric conditions.

Main Methods:

• Developed a time-varying Bayesian optimization (TV-BayesOpt) algorithm incorporating gradual forgetting and periodic covariance functions.
• Evaluated TV-BayesOpt using a computational model of phase-locked deep brain stimulation for oscillopathies.
• Compared TV-BayesOpt performance against standard time-invariant optimization techniques.

Main Results:

• TV-BayesOpt successfully tracked optimal stimulation settings that varied gradually and periodically over time.
• The algorithm demonstrated superior performance compared to time-invariant methods in dynamic parameter tracking.
• Robust performance was maintained even when initial assumptions about variations differed from observed data.

Conclusions:

• TV-BayesOpt offers a robust and adaptive approach for optimizing neuromodulation therapies where stimulation parameters change over time.
• This framework can be applied to various invasive and non-invasive neuromodulation strategies for conditions like Parkinson's disease and Essential Tremor.
• The algorithm's ability to track dynamic parameter changes enhances the potential for personalized and effective neuromodulation treatments.