Channel selection in chronic stroke rehabilitation

Wing Kin Tam, Xiaoling Hu, Kai Yu Tong

Research output: Journal article publicationConference articleAcademic researchpeer-review

2 Citations (Scopus)

Abstract

Brain-computer interface (BCI) can utilize signal directly from the brain for communication and control. Lately it has received increasing attention as a potential tool for stroke rehabilitation because it does not rely on the residue motor function of stroke patients. BCI for stroke rehabilitation still faces a major hurdle in wide adoption due to the long electrode preparation time before a training session. It is feasible to use channel selection techniques to reduce the number of channel in a BCI, but the effect of channel selection is not well-studied in stroke subjects. Specifically, it is still not clear how many channels are needed, how many calibration sessions are sufficient and which channel selection method can produce a better result. In this study, a 20-session dataset of 5 chronic stroke patients who have undergone BCI -functional electrical stimulation training was used to investigate the optimal configuration of channel selection in chronic stroke. A performance index has been proposed to provide a common ground of comparison between different configurations. SVM-RFE using 12 channels and 1 calibration session was revealed to obtain the best balance between convenience and accuracy. The average accuracy attained can be as high as 91.5±2.6 %.
Original languageEnglish
Pages (from-to)339-344
Number of pages6
JournalIFAC Proceedings Volumes (IFAC-PapersOnline)
Volume3
Issue numberPART 1
DOIs
Publication statusPublished - 1 Jan 2013
Event3rd IFAC Conference on Intelligent Control and Automation Science, ICONS 2013 - Chengdu, China
Duration: 2 Sep 20134 Sep 2013

Keywords

  • Brain-computer interface
  • Channel selection
  • Stroke

ASJC Scopus subject areas

  • Control and Systems Engineering

Cite this