Efficiently searching through large tACS parameter spaces using closed-loop Bayesian optimization

Romy Lorenz,Laura E Simmons,Ricardo P Monti,Joy L Arthur,Severin Limal,Ilkka Laakso,Robert Leech,Ines R Violante

doi:10.1016/j.brs.2019.07.003

Abstract

BackgroundSelecting optimal stimulation parameters from numerous possibilities is a major obstacle for assessing the efficacy of non-invasive brain stimulation. ObjectiveWe demonstrate that Bayesian optimization can rapidly search through large parameter spaces and identify subject-level stimulation parameters in real-time. MethodsTo validate the method, Bayesian optimization was employed using participants’ binary judgements about the intensity of phosphenes elicited through tACS. ResultsWe demonstrate the efficiency of Bayesian optimization in identifying parameters that maximize phosphene intensity in a short timeframe (5 min for >190 possibilities). Our results replicate frequency-dependent effects across three montages and show phase-dependent effects of phosphene perception. Computational modelling explains that these phase effects result from constructive/destructive interference of the current reaching the retinas. Simulation analyses demonstrate the method's versatility for complex response functions, even when accounting for noisy observations. ConclusionAlongside subjective ratings, this method can be used to optimize tACS parameters based on behavioral and neural measures and has the potential to be used for tailoring stimulation protocols to individuals.

Highlights

A challenge when selecting optimal parameters for studies involving transcranial alternating current stimulation is that it rapidly results in a combinatorial explosion of experimental conditions, as physiologically plausible frequencies (0.1e100 Hz) and relative phase differences between electrodes (0e359) spanR
Group-level Bayesian mean model for montage Cz-Oz (Fig. 1g; for group-level variance see Supplementary Fig. 1; subject-level results are available in Supplementary Fig. 2) implicate that the strongest phosphene perception was predicted at 19 Hz, in close agreement with previous studies [12]
We demonstrate that Bayesian optimization based on binary preference ratings provides a feasible and efficient method for searching through large transcranial alternating current stimulation (tACS) parameter spaces (312/192 possible combinations in Study 1/Study 2)

Summary

Introduction

A challenge when selecting optimal parameters for studies involving transcranial alternating current stimulation (tACS) is that it rapidly results in a combinatorial explosion of experimental conditions, as physiologically plausible frequencies (0.1e100 Hz) and relative phase differences between electrodes (0e359) span. R. Lorenz et al / Brain Stimulation 12 (2019) 1484e1489 accommodating many different types of functions. To validate nonparametric Bayesian optimization in the context of non-invasive brain stimulation, we looked at phosphenes: illusory flash-like visual percepts that can be reliably induced by tACS. Two tACS-related stimulation parameters have been shown to affect phosphene perception: frequency and intensity. Depending on the montage employed surface electrodes can generate superimposition of currents injected at different relative phases, which the retinas might be sensitive to, providing a compelling model to test our nonparametric Bayesian optimization approach. Conclusion: Alongside subjective ratings, this method can be used to optimize tACS parameters based on behavioral and neural measures and has the potential to be used for tailoring stimulation protocols to individuals

Objectives

Methods

Results

Conclusion