Brain-machine interface (BMI) or brain-computer interface (BCI) is an interface technology that utilizes neurophysiological signals from the brain to control external machines or computers. We have developed a P300-based BMI system to support daily activities of persons with disabilities (Komatsu et al., 2008; Kansaku, 2011), and found that green/blue chromatic and luminance flicker matrices improved the BMI performance for able-bodied subjects and spinal cord injury patients (Takano et al, 2009; Ikegami et al., 2011). We also created a region-based 2-step speller, which has a larger flashing area than the conventional visual array, and reported that the two-step procedure provided significantly increased accuracy compared with a conventional row/column speller for the amyotrophic lateral sclerosis patients (Ikegami et al., 2014). In this study, we investigated the performance of the region-based 2-step P300 BMI speller for spinocerebellar ataxia (SCA) patients and also for age- and sex-matched able-bodied control subjects.
Eight SCA patients and 8 age- and sex-matched control subjects who had not trained in BMI operation before (52-72 years old; 2 men) participated in this study. They were required to input Japanese hiragana characters using the in-house P300-based BMI system. Eight channel EEG data was recorded, and a linear discriminant analysis distinguished a region that the subject gazed at from other regions of the matrix. Online accuracy was evaluated.
The mean online accuracy was 70.6% for the SCA patients, and 69.5% for the able-bodied control subjects (the first step: 87.5% and 87.2%, the second step: 79.0% and 67.2%, respectively). No significant difference was observed between the SCA group and the control group (p = 0.928). The accuracy exceeded 70% in 5 out of 8 SCA patients and 6 out of 8 control subjects without significant training.
These results suggest that the region-based 2-step speller for P300-based BMI may be beneficial for SCA patients.