Online use of error-related potentials in healthy users and people with severe motor impairment increases performance of a P300-BCI

Abstract

Objective To investigate whether error-related potentials can be used to increase information transfer rate of a P3 brain-computer interface (BCI) in healthy and motor-impaired individuals. Methods Extraction and classification of the error-related potential was performed offline on data recorded from six amyotrophic lateral sclerosis (ALS) patients. An online study with 17 healthy and six motor impaired participants followed, using a modified P3 speller to provide explicit feedback of spelled letters. On recognition of error-related potentials, the interface informed users that the incorrect letter was automatically deleted. Results The offline cross-validation estimate of P3 speller data of six ALS patients increased bit rate by 0.44 bit/trial. During online copy spelling, the participants increased their bit rate by 0.52 bit/trial with the error correction system (ECS). Some participants performed free spelling and were able to increase their bit rate. Finally, we demonstrated that healthy participants could increase their bit rate by using a classifier pre-trained on other users' data. Conclusions Error-related potentials as a secondary source of information can be used to increase overall bit rate in a P3 BCI. Significance The method should be made available to any patient using the P3 BCI for communication.