# Dose standardization for transcranial electrical stimulation: an accessible approach

## TL;DR

This paper introduces a new method to standardize transcranial electrical stimulation doses using accessible participant data, improving consistency across individuals.

## Contribution

A novel dose standardization method for tES that avoids structural imaging and uses regression models trained on demographic and morphological data.

## Key findings

- Regression models predicted peak E-field strengths using accessible parameters, reducing variability compared to fixed dosing.
- Montage-specific models explained 43% variability in conventional and 21% in HD montages, while montage-agnostic models explained 36% and 13%, respectively.
- The method showed robust performance on unseen data, offering a significant improvement over current approaches.

## Abstract

Transcranial electrical stimulation (tES) is a widely used non-invasive brain stimulation technique. However, due to high inter-individual variability in the induced electric fields (E-fields), a fixed stimulation current delivers an inconsistent dose. We developed a dose standardization method without the requirement of participant-specific structural imaging and E-field modeling. Robust multiple linear regression models were trained to predict peak E-field strengths across 10 electrode montages and 418 healthy adults. These regression models predicted peak E-field strengths in unseen participants from accessible demographic and morphological parameters. Estimated peak E-field strength values were subsequently used to standardize tES dosages across our population. Additionally, we developed montage-agnostic models which incorporated inter-electrode distances for each participant. Compared to fixed dosing, our approach significantly reduced peak E-field strength variation for conventional montages, though results were inconsistent for high-definition (HD) montages. Models trained on specific montages accounted for 43% of peak E-field strength variability in conventional montages and 21% in HD montages on average. Our montage-agnostic models accounted for 36% and 13% of the average peak E-field strength variability for conventional and HD montages, respectively. These results have been validated across a large dataset, demonstrating robust performance against unseen data, a significant advancement over current approaches.