S-129 | Are you self-modulating enough? Measuring motor imagery skills in brain-computer interfaces

Theoretical and Computational Neuroscience
Author: Victoria Peterson | Email: vpeterson@santafe-conicet.gov.ar

Victoria Peterson^1°, Valeria Spagnolo^1°, Catalina M. Galván^1°,  Nicolás Nieto^1°2°, Rubén Spies^1°, Diego H. Milone^2°

^1° Instituto de Matemática Aplicada del Litoral, IMAL, CONICET-UNL
^2° Instituto de Investigaciones en Señales, Sistemas e Inteligencia Computacional, sinc(i), FICH-UNL, CONICET

Mastering a brain-computer interface based on motor imagery (MI-BCI) can be a long and frustrating journey for users. While users learn how to self-regulate their brain activity and generate distinguishable electroencephalography (EEG) patterns, the machine learning model that analyzes this brain data must adapt to signal changes that occur over multiple MI-BCI sessions. To address these challenges, we developed BOTDA -a backward formulation of optimal transport for domain adaptation- that adapts data distribution shifts between sessions in real time. In this study, we investigate the relationship between effective adaptation and the user’s ability to modulate their brain activity. We show that the effort BOTDA needs to put in performing the adaptation is correlated with the discriminability of EEG patterns, providing insights into the user MI capabilities in real-time. In this way, BOTDA not only facilitates seamless adaptation between sessions but also offers meaningful feedback that can help the user to enhance their MI-BCI control capabilities. By integrating BOTDA into MI-BCI systems, we aim to increase user confidence and engagement, ultimately contributing to improved clinical outcomes in motor rehabilitation.