Reliability of NIRS-based BCIs

Abstract

Previously, we contributed to the development of a brain-computer interface (BCI), Brainput, using functional near infrared spectroscopy (NIRS). Initially Brainput was found to improve performance on a human-robot team task by adapting a robot's autonomy using NIRS-based classifications of the user's multitasking states [15, 16]. However, the failure to find any performance improvements in a follow-up study prompted reinvestigation of the original system via a reanalysis of Brainput's signal processing on a larger NIRS dataset and a placebo-controlled replication using random (instead of NIRS-based) state classifications. This reinvestigation revealed confounds in the original study responsible for the initial performance improvements, thus indicating that further work in signal processing is necessary to achieve reliable NIRS-based BCIs.