2023 Volume 19 Issue 1 Pages 55-62
Introduction: Maintaining means of communication is an extremely important issue in occupational therapy for patients with intractable neurological disease. A brain-computer interface (BCI) is a device control technology for utilizing physiological signals, and is eagerly awaited for application in occupational therapy as a communication aid for patients with severe motor dysfunction. As an exploratory study for the development of BCI, we examined the accuracy of discriminating brain activation state using near-infrared spectroscopy (NIRS) signals and the effects of cognitive tasks on subjective burden through case study.
Methods: One 21-year-old male with Duchenne muscular dystrophy participated in the experiment. We measured NIRS signals with16 channels in the prefrontal cortex during three cognitive tasks: mental arithmetic (MA), figure rotation imagery (FRI), and Japanese word chain (JWC). Using support vector machine, a supervised machine learning, we constructed a classification model that discriminate type of cognitive task from features in the NIRS signals. In addition, the state-trait anxiety inventory and visual analogue scale were conducted to assess subjective burden associated with performing cognitive tasks.
Results: The classification accuracy of the MA versus FRI tasks was 73.1% (p = .03). The subjective evaluation scores showed the JWC had the lowest burden, and the MA and FRI were also rated as low anxiety tasks.
Conclusion: In this participant, we could identify whether the MA or FRI was being performed with 73.1% accuracy. These results suggest that brain activation state may be applicable as a means of communicating intentions.
