Exploring Sex-Related Biases in Deep Learning Models for Motor Imagery Brain-Computer Interfaces

Motor imagery (MI) brain-computer interfaces (BCIs) are promising technologies for neurorehabilitation. In this context, deep learning (DL) models are increasingly being used for MI-BCI. However, countless studies have shown that DL models are susceptible to bias, which can lead to disparate performance across subpopulations in terms of protected attributes, such as sex. The reported presence of sex-related information in electroencephalography (EEG) signals, widely used for MI-BCI, raises further concerns. For these reason, we conducted an in-depth analysis of underlaying factors driving model performance, identifying the discriminability of EEG patterns themselves as the principal contributor. Moreover, DL models actually improve overall performance as well as per-group performance, particularly boosting performance in subjects with less discriminable EEG patterns. Our work highlights the benefits of DL methods for MI-BCI and showcases the need for in-depth studies of model behavior beyond standard performance metrics to ensure the development of fair BCI systems.