Assessing Command-Following and Communication With Vibro-Tactile P300 Brain-Computer Interface Tools in Patients With Unresponsive Wakefulness Syndrome.

Christoph Guger,Rossella Spataro,Steven Laureys,Frederic Pellas,Vincenzo La Bella,Rupert Ortner,Alexander Heilinger,Woosang Cho,Jitka Annen,Ren Xu,Camille Chatelle,Günter Edlinger,Brendan Z Allison,Giorgio Mandalá

doi:10.3389/fnins.2018.00423

Abstract

Persons diagnosed with disorders of consciousness (DOC) typically suffer from motor disablities, and thus assessing their spared cognitive abilities can be difficult. Recent research from several groups has shown that non-invasive brain-computer interface (BCI) technology can provide assessments of these patients' cognitive function that can supplement information provided through conventional behavioral assessment methods. In rare cases, BCIs may provide a binary communication mechanism. Here, we present results from a vibrotactile BCI assessment aiming at detecting command-following and communication in 12 unresponsive wakefulness syndrome (UWS) patients. Two different paradigms were administered at least once for every patient: (i) VT2 with two vibro-tactile stimulators fixed on the patient's left and right wrists and (ii) VT3 with three vibro-tactile stimulators fixed on both wrists and on the back. The patients were instructed to mentally count either the stimuli on the left or right wrist, which may elicit a robust P300 for the target wrist only. The EEG data from −100 to +600 ms around each stimulus were extracted and sub-divided into 8 data segments. This data was classified with linear discriminant analysis (using a 10 × 10 cross validation) and used to calibrate a BCI to assess command following and YES/NO communication abilities. The grand average VT2 accuracy across all patients was 38.3%, and the VT3 accuracy was 26.3%. Two patients achieved VT3 accuracy ≥80% and went through communication testing. One of these patients answered 4 out of 5 questions correctly in session 1, whereas the other patient answered 6/10 and 7/10 questions correctly in sessions 2 and 4. In 6 other patients, the VT2 or VT3 accuracy was above the significance threshold of 23% for at least one run, while in 4 patients, the accuracy was always below this threshold. The study highlights the importance of repeating EEG assessments to increase the chance of detecting command-following in patients with severe brain injury. Furthermore, the study shows that BCI technology can test command following in chronic UWS patients and can allow some of these patients to answer YES/NO questions.

Highlights

Assessing consciousness and communication in persons with disorders of consciousness (DOC) is difficult
Brain-computer interfaces (BCIs) were originally developed to establish a communication channel with locked in syndrome (LIS) patients via brain activity alone, usually by measuring and analyzing the electroencephalographic (EEG) response for applications such as selecting letters (Wolpaw et al, 2002; Wolpaw and Wolpaw, 2012; Nam et al, 2018). Such BCIs have been validated with different types of EEG paradigms, including motor imagery (MI) (Guger et al, 2003; Acqualagna et al, 2016), steady-state visual evoked potentials (SSVEPs; Bin et al, 2009; Ahn et al, 2016) or P300 event-related potentials (ERPs; Guger et al, 2009, 2016; Lugo et al, 2014)
A vibro-tactile P300 study with LIS patients showed that the BCI system can still extract information from the EEG, even if visual inspection of the averaged ERPs suggests this is impossible

Summary

Introduction

Assessing consciousness and communication in persons with disorders of consciousness (DOC) is difficult. Brain-computer interfaces (BCIs) were originally developed to establish a communication channel with LIS patients via brain activity alone, usually by measuring and analyzing the electroencephalographic (EEG) response for applications such as selecting letters (Wolpaw et al, 2002; Wolpaw and Wolpaw, 2012; Nam et al, 2018) Such BCIs have been validated with different types of EEG paradigms, including motor imagery (MI) (Guger et al, 2003; Acqualagna et al, 2016), steady-state visual evoked potentials (SSVEPs; Bin et al, 2009; Ahn et al, 2016) or P300 event-related potentials (ERPs; Guger et al, 2009, 2016; Lugo et al, 2014). This is because the EEG data from each single trial was analyzed using linear discriminate analysis (LDA), in contrast visual inspection of averaged ERPs (Lugo et al, 2014)

Methods

Results

Discussion

Conclusion