Gaussian Process-Based Surrogate Models for Optimizing Electrode Configurations in HD-tDCS.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computational Neuroscience

Background:

• High-definition transcranial direct current stimulation (HD-tDCS) uses multiple scalp electrodes for targeted brain modulation.
• Optimizing electrode placement in HD-tDCS is complex due to intricate brain anatomy and numerous configurations.
• Current simulation methods for electrode optimization are often computationally intensive and restrict placement options.

Purpose of the Study:

• To develop a computationally efficient framework for optimizing HD-tDCS electrode placement.
• To enable continuous prediction of electric field distributions for enhanced stimulation precision.
• To expand the search space for electrode configurations beyond traditional leadfield-based limitations.

Main Methods:

• Proposed a Gaussian Process (GP)-based framework for continuous electric field prediction in HD-tDCS.
• Employed a Sparse Gaussian Process (SGP) approximation for efficient handling of large datasets.
• Utilized Block-Coordinate Descent and Subset of Data techniques for SGP optimization.

Main Results:

• The SGP-based model significantly improved focality for superficial and mid-brain stimulation targets.
• Achieved performance comparable to leadfield-based methods for deep brain targets.
• Demonstrated enhanced precision and flexibility in HD-tDCS electrode placement optimization.

Conclusions:

• The proposed GP-based framework offers a flexible and precise method for optimizing HD-tDCS electrode configurations.
• This approach overcomes computational limitations of traditional methods, expanding possibilities for electrode placement.
• Supports the advancement of HD-tDCS in both research and clinical neurostimulation applications.