Single-Trial Functional Connectivity Dynamics of Event-Related Desynchronization for Motor Imagery EEG-Based Brain-Computer Interfaces

Author(s)

Rodrigues, P. G. ; Fim-Neto, A. ; Sato, J. R. ; Soriano, D. C. ; Nasuto, S. J. ; Bastos-Filho, TF ; Caldeira, EMD ; Frizera-Neto, A

• Document type: Journal article
• Source: XXVII BRAZILIAN CONGRESS ON BIOMEDICAL ENGINEERING, CBEB 2020; v. N/A, p. 7-pg., 2022-01-01.

Abstract

Functional connectivity (FC) analysis has been widely applied to the study of the brain functional organization under different conditions and pathologies providing compelling results. Recently, the investigation of FC in motor tasks has drawn the attention of researchers devoted to post-stroke rehabilitation and those seeking robust features for the design of brain-computer interfaces (BCIs). In particular, concerning this application, it is crucial to understand: (1) how motor imagery (MI) networks dynamics evolve over time; (2) how it can be suitably characterized by its topological quantifiers (graph metrics); (3) what is the discrimination capability of graph metrics for BCI purposes. This work aims to investigate the MI single-trial time-course of functional connectivity defined in terms of event-related desynchronization/ synchronization (ERD/S) similarity. Both ERD/S and clustering coefficient (CC) underlying FC were used as features for characterizing rest, right-hand MI, and left-hand MI for 21 subjects. Our results showed that MI can be associated with a higher CC when compared to rest, while right- and left-hand MI present a similar CC time-course evolution. From the classification standpoint, ERD/S, CC and their combination provided moderate to substantial single electrode peak performances (in terms of Cohen's kappa) for discriminating rest and movement, i.e. for identifying alpha rhythm suppression. Weak peak classification performances were achieved for these features for right- and left-hand discrimination, but the combination of FC-based features and ERD/S provided significantly better results, suggesting complementary information. These results illustrate the symmetrical nature of brain activity relative power dynamics, as reflected in dynamics of functional connectivity during single trial MI and motivate need for further exploration of such measures for BCI applications. (AU)

• FAPESP's process: 19/09512-0 - Nonlinear dynamic functional connectivity analysis via recurrence quantification and its application to brain computer-interfaces
• Grantee: Diogo Coutinho Soriano
• Support Opportunities: Scholarships abroad - Research
• FAPESP's process: 13/07559-3 - BRAINN - The Brazilian Institute of Neuroscience and Neurotechnology
• Grantee: Fernando Cendes
• Support Opportunities: Research Grants - Research, Innovation and Dissemination Centers - RIDC