Is breathing rate a confounding variable in brain-computer interfaces (BCIs) based on EEG spectral power?

Abstract

Brain-computer interfaces (BCIs) enable paralyzed patients to interact with the world by directly decoding brain activity. We investigated if systematic changes in breathing rate affect EEG bandpower features that are commonly used in BCIs. This is of particular interest for the development of cognitive BCIs for patients with artificial ventilation, e.g. for those in late stages of amyotrophic lateral sclerosis (ALS). If subjects can alter the spectrum of the EEG by changing their breathing rate, decoding results obtained with healthy subjects may not generalize to this patient population. We recorded a high-density EEG from twelve healthy subjects, who were instructed to alternate between fast and slow breathing. We do not find any statistically significant modulation of EEG bandpower. As such, changes in breathing rate are unlikely to substantially bias the performance of BCIs based on EEG bandpower features.