In silicodevelopment and validation of Bayesian methods for optimizing deep brain stimulation to enhance cognitive...

Area of Science:

• Neuroscience
• Computational Psychiatry
• Medical Engineering

Background:

• Deep brain stimulation (DBS) of the ventral internal capsule/striatum (VCVS) shows promise for treatment-resistant mental health disorders.
• Current VCVS programming is lengthy, relying on subjective patient feedback and taking weeks for adjustments.
• Objective measures are needed to efficiently optimize DBS settings and confirm circuit engagement.

Purpose of the Study:

• To develop and evaluate an objective method for measuring and optimizing VCVS DBS circuit engagement.
• To reduce the time and improve the confidence in selecting effective DBS parameters.

Main Methods:

• Developed a simulation framework based on prior findings of VCVS DBS engaging cognitive control circuitry.
• Integrated an adaptive optimizer to explore electrode contacts and identify those maximizing cognitive control.
• Compared optimization algorithms, input numbers, and stimulation parameter change speeds on simulated tasks.

Main Results:

• An upper confidence bound algorithm, in a majority-vote ensemble, demonstrated an approximately 80% probability of converging to the global optimum.
• Optimization successfully converged despite a lag between stimulation and observed effects.
• A complete optimization was achievable within a clinically practical timeframe of a few hours.

Conclusions:

• This approach offers a scalable method for optimizing DBS settings without specialized hardware.
• It can significantly accelerate the process of finding effective DBS parameters for psychiatric and non-motor applications.
• The method enhances confidence in DBS programming, potentially expanding its clinical utility.