Design of electrode layout for motor imagery based brain–computer interface

Introduction: A motor imagery based brain–computer interface (BCI) translates a subject’s motor intention into a command signal through real-time detection of motor imagery states, e.g. imagination of left and right hand movement. During motor imagery, electroencephalogram (EEG) signals accompany power changes in movement related m (8–12 Hz) and b (18–26 Hz) rhythms, representing a power increase or decrease named event-related desynchronisation and sychronisation (ERD=ERS) in specific motor cortex areas [1]. The first motor imagery based BCI was developed by Pfurtscheller et al. [2] and was based upon the detection of EEG power changes caused by ERD=ERS of m and b rhythms during imagination of left and right movements. Another motor imagery based approach was to train users to regulate the amplitude of m and b rhythms to realise 2-D control of cursor movement [3]. Functional neuroimaging studies indicated that motor imagery also activates the supplementary motor cortex area (SMA) [4]. However, existing algorithms for classifying motor imagery states only focus on the ERD=ERS over the primary sensory-motor cortex. How to explore the value of the SMA for motor imagery classification is a challenge because SMA may be activated under all motor imagery states, i.e. no obvious power difference exists. In recent years, measurement of brain synchrony with EEG signals has been applied for exploring the dynamics of brain networks [5]. In our previous study, we investigated the phase synchronisation of m rhythms between the SMA and the primary motor cortex (M1) and observed a contralateral increased synchronisation similar to the ERD distribution [6]. This phenomenon makes it possible to utilise the signal over the SMA to enhance the significance of power difference between M1 areas, through considering SMA as the reference. Here we propose a novel electrode layout inspired by the synchronisation between the SMA and M1. The layout of two bipolar leads, i.e. C3-FCz and C4-FCz, is demonstrated to be optimal for recognising motor imagery states, which thus can satisfy the necessity of a practical BCI.