Spatial abilities play a major role in BCI performance

Jeunet, C, Lotte, F, Hachet, M, Subramanian, S and N'Kaoua, B N (2016) Spatial abilities play a major role in BCI performance. In: 6th International BCI Meeting, 30-May to 3rd June, California, USA.

[img] PDF - Accepted Version
Download (635kB)


Introduction: Despite their promising potential impact for many applications, Mental-Imagery based BCIs (MI-BCIs) remain barely used outside laboratories. One reason is that 15% to 30% of naïve users seem unable to control them [1] and only a few reach high control abilities. Although different predictors of BCI performance (i.e., command classification accuracy) have been investigated to explain this huge inter-user variability [2, 3], no strong predictive model has yet been determined. This could be due to (a) the often small samples used (N=5 or 6) and (b) the fact that these predictors have been mostly determined based on one-session experiments. Yet there is no evidence that performance obtained at the first session is predictive of users' MI-BCI control ability. Material, Methods and Results: In [4], we investigated the impact of the user's personality and cognitive profile on MI-BCI performance based on a 6-session experiment. Averaging performances over these sessions reduced the intra-subject variability (e.g., due to fatigue or external factors), and thus led to a better estimation of participants' MI-BCI control ability. Each session comprised 5 runs during which the participants (N=18) had to learn to perform 3 MI tasks: left-hand motor imagery, mental rotation and mental calculation. The results stressed the impact of mental rotation scores (measured using questionnaires), and which reflect Spatial Abilities (SA), on mean MI-BCI performance [r=0.696, p<0.05] (see Fig. 1[A]). SA are the mental capacities which enable the construction, transformation and interpretation of mental images. In a more recent study (to be published), we trained 20 participants to control a 2-class MI-BCI by performing motor-imagery of their left-and right-hands, within 1 session of 5 runs. Results confirmed the role of SA: mental rotation scores were correlated with peak MI-BCI performance [r=0.464, p<0.05]. This suggests that SA are a generic predictor of MI-BCI performances. Figure 1. [A] Diagram representing the mean classification accuracy for the different subjects as a function of their mental rotation score; [B] One item per exercise included in the Spatial Ability training:the shape on top is the target, and the participant must identify the two shapes that are identical to the target among the four below.

Item Type: Conference or Workshop Item (Other)
Schools and Departments: School of Engineering and Informatics > Informatics
Related URLs:
Depositing User: Sriram Subramanian
Date Deposited: 01 Apr 2016 12:18
Last Modified: 30 Aug 2017 13:14

View download statistics for this item

📧 Request an update