Real-time Bayesian optimization of deep brain stimulation for personalized cognitive control enhancement

Area of Science:

• Neuroscience
• Computational Psychiatry
• Biomarkers

Background:

• Deep brain stimulation (DBS) parameter optimization for psychiatric disorders is challenging due to the lack of objective readouts for target engagement.
• Cognitive control shows potential as a biomarker for DBS treatment efficacy, but reliable optimization for individual patients remains unproven.

Purpose of the Study:

• To investigate the efficacy of optimization algorithms in identifying effective DBS amplitudes for enhancing cognitive control in a rat model.
• To determine if state-of-the-art optimization can consistently improve cognition through precise stimulation parameter selection.

Main Methods:

• Rats performed a Set-Shifting task, a stimulation-sensitive cognitive control measure.
• Active and inactive DBS-like stimulation were delivered at variable parameters.
• Bayesian Optimization was used to personalize stimulation amplitudes, comparing task performance (reaction time, accuracy) to predefined and traditional settings.

Main Results:

• Acute DBS stimulation reduced reaction times without affecting accuracy in 15 rats, confirming previous findings.
• In a separate cohort of 6 rats, Bayesian Optimization successfully identified stimulation amplitudes that reduced reaction times in all subjects.

Conclusions:

• Optimization techniques, particularly Bayesian Optimization, can effectively enhance cognitive markers relevant to psychiatric and cognitive disorders.
• These findings support the feasibility of personalized, quantitatively-driven neuromodulation for improved target engagement and treatment efficacy.