A P300 Brain-Computer Interface With a Reduced Visual Field

Luiza Kirasirova; Vladimir Bulanov; Alexei Ossadtchi; Alexander Kolsanov; Vasily Pyatin; Mikhail Lebedev

doi:10.3389/fnins.2020.604629

A P300 Brain-Computer Interface With a Reduced Visual Field

Front Neurosci. 2020 Dec 3:14:604629. doi: 10.3389/fnins.2020.604629. eCollection 2020.

Authors

Luiza Kirasirova¹, Vladimir Bulanov², Alexei Ossadtchi³, Alexander Kolsanov¹, Vasily Pyatin¹, Mikhail Lebedev^{3

4

5}

Affiliations

¹ Samara State Medical University, Samara, Russia.
² Laboratory of Mathematical Processing of Biological Information, IT Universe Ltd, Samara, Russia.
³ Center for Bioelectric Interfaces of the Institute for Cognitive Neuroscience, National Research University Higher School of Economics, Moscow, Russia.
⁴ Department of Information and Internet Technologies of Digital Health Institute, I.M. Sechenov First Moscow State Medical University, Moscow, Russia.
⁵ Center For Neurobiology and Brain Restoration, Skolkovo Institute of Science and Technology, Moscow, Russia.

Abstract

A P300 brain-computer interface (BCI) is a paradigm, where text characters are decoded from event-related potentials (ERPs). In a popular implementation, called P300 speller, a subject looks at a display where characters are flashing and selects one character by attending to it. The selection is recognized as the item with the strongest ERP. The speller performs well when cortical responses to target and non-target stimuli are sufficiently different. Although many strategies have been proposed for improving the BCI spelling, a relatively simple one received insufficient attention in the literature: reduction of the visual field to diminish the contribution from non-target stimuli. Previously, this idea was implemented in a single-stimulus switch that issued an urgent command like stopping a robot. To tackle this approach further, we ran a pilot experiment where ten subjects operated a traditional P300 speller or wore a binocular aperture that confined their sight to the central visual field. As intended, visual field restriction resulted in a replacement of non-target ERPs with EEG rhythms asynchronous to stimulus periodicity. Changes in target ERPs were found in half of the subjects and were individually variable. While classification accuracy was slightly better for the aperture condition (84.3 ± 2.9%, mean ± standard error) than the no-aperture condition (81.0 ± 2.6%), this difference was not statistically significant for the entire sample of subjects (N = 10). For both the aperture and no-aperture conditions, classification accuracy improved over 4 days of training, more so for the aperture condition (from 72.0 ± 6.3% to 87.0 ± 3.9% and from 72.0 ± 5.6% to 97.0 ± 2.2% for the no-aperture and aperture conditions, respectively). Although in this study BCI performance was not substantially altered, we suggest that with further refinement this approach could speed up BCI operations and reduce user fatigue. Additionally, instead of wearing an aperture, non-targets could be removed algorithmically or with a hybrid interface that utilizes an eye tracker. We further discuss how a P300 speller could be improved by taking advantage of the different physiological properties of the central and peripheral vision. Finally, we suggest that the proposed experimental approach could be used in basic research on the mechanisms of visual processing.

Keywords: ERP; P300 BCI; aperture; central vision; visual attention; visual fatigue.