The Effects of Sensory Threshold Somatosensory Electrical Stimulation on Users With Different MI-BCI Performance

Abstract

Motor imagery-based brain-computer interface (MI-BCI) has been largely studied to improve motor learning and promote motor recovery. However, the difficulty in performing MI limits the widespread application of MI-BCI. It has been suggested that the usage of sensory threshold somatosensory electrical stimulation (st-SES) is a promising way to guide participants on MI tasks, but it is still unclear whether st-SES is effective for all users. In the present study, we aimed to examine the effects of st-SES on the MI-BCI performance in two BCI groups (High Performers and Low Performers). Twenty healthy participants were recruited to perform MI and resting tasks with EEG recordings. These tasks were modulated with or without st-SES. We demonstrated that st-SES improved the performance of MI-BCI in the Low Performers, but led to a decrease in the accuracy of MI-BCI in the High Performers. Furthermore, for the Low Performers, the combination of st-SES and MI resulted in significantly greater event-related desynchronization (ERD) and sample entropy of sensorimotor rhythm than MI alone. However, the ERD and sample entropy values of MI did not change significantly during the st-SES intervention in the High Performers. Moreover, we found that st-SES had an effect on the functional connectivity of the fronto-parietal network in the alpha band of Low Performers and the beta band of High Performers, respectively. Our results demonstrated that somatosensory input based on st-SES was only beneficial for sensorimotor cortical activation and MI-BCI performance in the Low Performers, but not in the High Performers. These findings help to optimize guidance strategies to adapt to different categories of users in the practical application of MI-BCI.

Keywords: EEG; brain-computer interface; functional connectivity; motor imagery; sensory threshold somatosensory electrical stimulation.

Figure 1

The task procedure of experimental condition. (A) MI condition. In each trial of MI condition, the subjects were required to perform MI task according to the cue. (B) SES condition. In each trial of SES condition, the subjects were asked to remain at rest and received st-SES intervention. (C) MI-SES condition. In each trial of MI-SES condition, the subjects were instructed to perform MI task according to the cue and st-SES was applied to subjects during MI. (D) Rest condition. In each trial of Rest condition, the subjects were asked to remain relax.

Figure 2

Offline accuracies of brain-switch BCI with or without st-SES modulated. (A) Classification accuracy results in High Performers group. (B) Classification accuracy results in Low Performers group. Error bars represent standard error of mean. “*” indicates p < 0.05, “**” indicates p < 0.01.

Figure 3

Offline accuracies of distinguishing three conditions (MI-SES, MI, SES) from Rest condition. (A) Classification accuracy results in High Performers group. (B) Classification accuracy results in Low Performers group. Error bars represent standard error of mean. “*” indicates p < 0.05, “**” indicates p < 0.01.

Figure 4

The average time-frequency maps over C3 channel for all High Performers and all Low Performers. The period [1 5] s indicated the MI or rest task. The period [3 5] s corresponds to st-SES in MI-SES and SES conditions.

Figure 5

Averaged ERSP values of alpha (8–13 Hz) and beta (14–28 Hz) rhythms during period [3 5] s for MI-SES and MI conditions. (A) High Performers group. (B) Low Performers group. Error bars represent the standard deviation. “*” indicates p < 0.05, “**” indicates p < 0.01.

Figure 6

Averaged topographical distributions of r² during period [3 5] s for all High Performers and all Low Performers.

Figure 7

Averaged relative sample entropy values of C3 channel in MI-SES and MI conditions. Error bars represent standard error of mean. “**” indicates p < 0.01.

Figure 8

The connectivity networks of alpha and beta rhythms for MI-SES and MI conditions. (A) High Performers group. (B) Low Performers group. The line between channels represents functional connectivity with significantly different wPLI values between MI and resting states.