Visuospatial attention based brain-computer interfacing using real-time fMRI

Abstract

Despite all the literature showing that brain-computer interfacing (BCI) is achievable in a research setting, reports of paralyzed people utilizing the proposed systems are few. The performances of the BCI systems must most likely improve before they will move from research and become an assistive tool used by patients in their daily life. The majority of BCIs presented to date have been based on EEG. With the sensors placed outside the skull, this technique suffers from poor spatial resolution and low signal power. These limiting factors can be overcome by instead placing the electrodes on the cortical surface (ECoG). The opportunities for evaluating new ideas with invasive techniques are however very limited since the need for surgery excludes testing on healthy volunteers. We wished to investigate whether covert shifts of visuospatial attention could be decoded in real time and used for BCI control. With this in mind we wanted to create an environment that makes it possible to test this and other new control paradigms, to practice them with real-time performance feedback and to identify the target brain areas. The tool should be non-invasive, but still give results that are relevant for implanted technologies. Based on convincing evidence of fMRI correlating with ECoG in localization of brain functions we hypothesized that real-time fMRI (RT-fMRI) could meet our needs. We implemented our RT-fMRI system on a ultra-high field (7T) MRI scanner since it could provide us with high resolution images while maintaining the high contrast-to-noise ratio that is needed to classify single trials. Using the developed system we evaluated our BCI control task, based on covert shifts of visuospatial attention, that is completely independent of any stimulation or eye movements. Our work shows that visuospatial attention can provide a novel means of controlling a BCI with multiple degrees of freedom. This control does not depend on the ability to move the eyes, and in contrast to BCI based on P300 and SSVEP it is independent of any form of visual stimulation. The RT-fMRI setup described in our work could also be used to evaluate other control strategies and to practice them in a BCI environment.