Recalibration of neuromodulation parameters in neural implants with adaptive Bayesian optimization

Area of Science:

• Biomedical Engineering
• Neurotechnology
• Artificial Intelligence

Background:

• Neuromodulation technology offers potential for treating neural activity disorders.
• Personalized neurostimulation protocols are crucial for treatment efficiency and device effectiveness.
• Current neural interfaces require frequent, expert-mediated recalibration of stimulation parameters, leading to increased costs, reduced usability, and patient discomfort.

Purpose of the Study:

• To develop an adaptable, AI-based system for autonomous recalibration of neuromodulation parameters.
• To address the time-variability challenges in neural interfaces for neuroprosthetics.
• To improve the usability and efficacy of neuromodulation devices through intelligent adaptation.

Main Methods:

• Utilized Gaussian process-based Bayesian optimization (GPBO) to re-adjust neurostimulation parameters.
• Integrated temporal information into GPBO to manage time variability in neural interfaces.
• Developed predictive models to optimize active electrode sites and adapt injected charge for neural activation, validated on human clinical trial data.

Main Results:

• The AI-driven algorithm successfully adapted neurostimulation parameters to changes in evoked-sensation location and perceptual threshold over time.
• Demonstrated a rapid, automatic method for managing neurostimulation parameter variability.
• Quantitative assessment in an offline setting confirmed the effectiveness of the GPBO-based approach.

Conclusions:

• AI-based methods, specifically GPBO, can autonomously adapt neuromodulation parameters, enhancing device usability and reducing treatment burden.
• This technology has the potential for widespread adoption in future 'smart' neuromodulation devices.
• The findings support a significant advancement in personalized and efficient neuromodulation therapies.