Binary Semantic Classification Using Cortical Activation with Pavlovian-Conditioned Vestibular Responses in Healthy and Locked-In Individuals

Natsue Yoshimura; Kaito Umetsu; Alessandro Tonin; Yasuhisa Maruyama; Kyosuke Harada; Aygul Rana; Gowrishankar Ganesh; Ujwal Chaudhary; Yasuharu Koike; Niels Birbaumer

doi:10.1093/texcom/tgab046

Binary Semantic Classification Using Cortical Activation with Pavlovian-Conditioned Vestibular Responses in Healthy and Locked-In Individuals

Cereb Cortex Commun. 2021 Jul 23;2(3):tgab046. doi: 10.1093/texcom/tgab046. eCollection 2021.

Affiliations

¹ Institute of Innovative Research, Tokyo Institute of Technology, Yokohama 226-8503, Japan.
² Wyss-Center for Bio and NeuroEngineering, Geneva CH-1202, Switzerland.
³ Institute of Medical Psychology and Behavioral Neurobiology, University of Tübingen, 72076 Tübingen, Germany.
⁴ Laboratorie d'Informatique, de Robotique et de Microelectronique de Montpellier, U. Montpellier, CNRS, 34095 Montpellier, France.

Abstract

To develop a more reliable brain-computer interface (BCI) for patients in the completely locked-in state (CLIS), here we propose a Pavlovian conditioning paradigm using galvanic vestibular stimulation (GVS), which can induce a strong sensation of equilibrium distortion in individuals. We hypothesized that associating two different sensations caused by two-directional GVS with the thoughts of "yes" and "no" by individuals would enable us to emphasize the differences in brain activity associated with the thoughts of yes and no and hence help us better distinguish the two from electroencephalography (EEG). We tested this hypothesis with 11 healthy and 1 CLIS participant. Our results showed that, first, conditioning of GVS with the thoughts of yes and no is possible. And second, the classification of whether an individual is thinking "yes" or "no" is significantly improved after the conditioning, even in the absence of subsequent GVS stimulations. We observed average classification accuracy of 73.0% over 11 healthy individuals and 85.3% with the CLIS patient. These results suggest the establishment of GVS-based Pavlovian conditioning and its usability as a noninvasive BCI.

Keywords: Pavlovian conditioning; brain–computer interface; completely locked-in state; electroencephalography; galvanic vestibular stimulation.