Neural mechanisms of brain–computer interface control

Abstract

Brain–computer interfaces (BCIs) enable people with paralysis to communicate with their environment. Motor imagery can be used to generate distinct patterns of cortical activation in the electroencephalogram (EEG) and thus control a BCI. To elucidate the cortical correlates of BCI control, users of a sensory motor rhythm (SMR)-BCI were classified according to their BCI control performance. In a second session these participants performed a motor imagery, motor observation and motor execution task in a functional magnetic resonance imaging (fMRI) scanner.

Group difference analysis between high and low aptitude BCI users revealed significantly higher activation of the supplementary motor areas (SMA) for the motor imagery and the motor observation tasks in high aptitude users. Low aptitude users showed no activation when observing movement. The number of activated voxels during motor observation was significantly correlated with accuracy in the EEG-BCI task (r = 0.53). Furthermore, the number of activated voxels in the right middle frontal gyrus, an area responsible for processing of movement observation, correlated (r = 0.72) with BCI-performance. This strong correlation highlights the importance of these areas for task monitoring and working memory as task goals have to be activated throughout the BCI session.

The ability to regulate behavior and the brain through learning mechanisms involving imagery such as required to control a BCI constitutes the consequence of ideo-motor co-activation of motor brain systems during observation of movements. The results demonstrate that acquisition of a sensorimotor program reflected in SMR-BCI-control is tightly related to the recall of such sensorimotor programs during observation of movements and unrelated to the actual execution of these movement sequences.

Introduction

Motor imagery uses the conscious effort of simulating movements without actually performing them in an overt fashion (Jeannerod, 1995, Decety, 1996). The concept of a functional equivalence between imagery of movement and the perception of performing goes back to the inception of the ideo-motor theory of thinking in the 19th century (Carpenter, 1874, Stock and Stock, 2004). Motor imagery is generally assumed to activate the same representations as the corresponding motor execution. For example, it has been shown that imagined and executed movements follow the same temporal constraints (Decety et al., 1989). Using motor imagery in a training program has even been shown to lead to measurable increases in the voluntary force production of one of the five digits (Yue and Cole, 1992). Consequently mental practice of movement sequences activates similar brain areas as physical practice and also leads to improved performance when learning a sequence of foot movements (Lafleur et al., 2002, Jackson et al., 2003). There is some evidence that mental rehearsal may have rehabilitative effects in stroke and spinal cord injury (Malouin and Richards, 2010).

Brain–computer interfaces (BCIs) use voluntary control of brain activity in many applications (Kübler and Neumann, 2005). They were developed for people with neurodegenerative diseases, brain injuries or stroke that can lead to severe or complete paralysis. Various BCI paradigms have been proposed and three in particular have been evaluated with patients (mostly affected by amyotrophic lateral sclerosis (ALS), a degenerative disease of the first and second motor neurons leading to paralysis). BCIs are controlled by components of the electroencephalogram (EEG) (Berger, 1929) such as slow cortical potentials (SCPs) (Birbaumer et al., 1999), event-related potentials (ERPs, mainly the P300) (Nijboer et al., 2008b, Silvoni et al., 2009) and sensorimotor-rhythm (SMR) or μ-rhythm (Kübler et al., 2005). Apart from restoring communication there is some evidence that motor imagery-based BCIs may have a rehabilitative effect in chronic stroke patients (Buch et al., 2008). It was shown that out of a sample of 8 patients, 6 were able to learn successfully to control an orthosis through a neuromagnetic BCI.

The SMR is a brain rhythm (8–15 Hz) located over sensorimotor areas and is insensitive to visual input (Kuhlman, 1978). Instead of one uniform rhythm the sensorimotor area generates a variety of rhythms that have specific functional and topographic properties. Distinct rhythms are generated by areas involving e.g. hand and foot movements over the postcentral somatosensory cortex with the strongest contributions originating from the somatosensory cortex of the hand (Hari and Salmelin, 1997, Pfurtscheller et al., 1997). Also associated with the sensorimotor system is a phase coupled second peak in the beta band (16–30 Hz) of the resting EEG over the precentral motor cortex anterior to the sources of the alpha component of the SMR (Hari and Salmelin, 1997). The alpha and the beta peak can become independent for example at the offset of a movement after which the beta band rebounds faster and with a higher amplitude than the alpha band (Pfurtscheller, 1981). Additionally, desynchronization of the beta band during motor tasks can occur in slightly different frequency bands than the subsequent synchronization (rebound) after the motor task (Müller-Putz et al., 2007). Both of these rhythms are desynchronized by overt and covert movement, planning and observation of movement and movement imagery. Specifically, the sensitivity to motor imagery renders the SMR a useful control signal for BCIs (Blankertz et al., 2008, Blankertz et al., 2010, Kübler et al., 2005, Neuper et al., 2003, Neuper et al., 2009, Nijboer et al., 2008a). Consequently SMR based BCIs are a useful tool for measuring the effectiveness of regulating the thalamocortical system responsible for SMR-modulation through motor imagery.

In the current study we aimed at elucidating differences in the cortical network between high and low aptitude users during BCI performance using motor imagery. To achieve this goal we identified 10 high and 10 low aptitude BCI users in an fMRI study. The 20 participants were recruited from the participant pool (n = 80) that participated in the study of Blankertz et al. (2010) and were assigned to the group of high or low aptitude users according to their performance in the EEG-BCI experiment (median split). Functional MRI was measured on average 2 weeks after the EEG experiment, never on the same day, and comprised a motor imagery, motor execution and motor observation task.

Motor imagery and motor observation have been extensively studied with fMRI and other brain imaging techniques such as positron emission tomography (PET) (see Lotze and Halsband (2006) for a review). The results indicate that neural networks similar to those of executed movements are activated by imagery and observation of movement (Decety et al., 1994). Differences in activations between motor imagery and motor execution tasks have been found in bilateral premotor, prefrontal, supplementary motor and left posterior parietal areas as well as the caudate nuclei (Gerardin et al., 2000). The individual capacity to imagine movements has also been studied previously with imaging techniques. Stronger activations in high aptitude than in low aptitudemotor imagery performers were found in parietal and ventrolateral premotor regions and stronger activations for low as compared to high aptitude performers in the cerebellum, orbito-frontal and posterior cingulate cortices indicating that low aptitude users unsuccessfully compensate a reduced activation of motor areas with activation of cortical areas not directly related to movement execution or imagery (Guillot et al., 2008). In the previously described study motor imagery ability was assessed using questionnaires and physiological measures based on peripheral autonomic responses such as skin resistance. In a comparison of professional and amateur musicians the participant's level of expertise coincided with focused activations in the SMA, the superior PMC, more anterior areas (Larsell's lobule HVI) in the left cerebellar hemisphere and bilateral superior parietal areas whereas amateurs showed less focused and thus more widespread activations while playing and not imaging of a well trained musical piece (Lotze et al., 2003).

Taking into account the positive correlation between the resting SMR power and later BCI performance found in Blankertz et al. (2010), we hypothesized that high aptitude BCI users would recruit and activate to a larger extent the motor network known to be involved in motor imagery, namely premotor and supplementary motor areas. We further predicted that we would find such differences to a lesser extent during the motor observation task. During observation participants were asked to image the observed movement. We assumed that this would be easier than motor imagery alone. Finally, we hypothesized that the recruitment and activation of motor areas as measured with fMRI would predict performance in the BCI controlled by electrocortical signals.