ИСТИНА

Aims This study aimed to investigate gaze and attention behavior of P300 based brain-computer interface user to reveal the possible effects on system performance. Methods In this study, we employed two experimental modes: the “classic” mode, in which the gaze and attention should be directed to the same stimulus (target), and the “distraction” mode, where the gaze and attention should be directed to 2 different closely located stimuli (target and distractor). Eye-movements data was acquired from 22 healthy subjects with the help of SMI RED 500 eye tracker and processed with PyGaze library. Results We found that the accuracy of P300 BCI reduced drastically with the gaze shifts from target stimulus – median accuracy was only 5% comparing to 90% in “classic” mode (Mood median test p < 0.001). Attention shifts affected median accuracy as well, reducing it to 80%. We also found that the target gaze dispersion was significantly higher (paired t-test p = 0.004) and the target gaze duration was lower (paired t-test p = 0.003) in “distraction” mode, which was probably a result of high cognitive demand. Conclusion This study revealed a dramatic negative effect of gaze shifts on P300 BCI performance, which should be taken into consideration for future studies and applications. Observed gaze behavior changes can be used as predictors of attention shifts and high cognitive demand in P300 BCI.