Comparison of tactile, auditory, and visual modality for brain-computer interface use: a case study with a patient in the locked-in state

Tobias Kaufmann,Elisa M Holz,Andrea Kübler

doi:10.3389/fnins.2013.00129

Abstract

This paper describes a case study with a patient in the classic locked-in state, who currently has no means of independent communication. Following a user-centered approach, we investigated event-related potentials (ERP) elicited in different modalities for use in brain-computer interface (BCI) systems. Such systems could provide her with an alternative communication channel. To investigate the most viable modality for achieving BCI based communication, classic oddball paradigms (1 rare and 1 frequent stimulus, ratio 1:5) in the visual, auditory and tactile modality were conducted (2 runs per modality). Classifiers were built on one run and tested offline on another run (and vice versa). In these paradigms, the tactile modality was clearly superior to other modalities, displaying high offline accuracy even when classification was performed on single trials only. Consequently, we tested the tactile paradigm online and the patient successfully selected targets without any error. Furthermore, we investigated use of the visual or tactile modality for different BCI systems with more than two selection options. In the visual modality, several BCI paradigms were tested offline. Neither matrix-based nor so-called gaze-independent paradigms constituted a means of control. These results may thus question the gaze-independence of current gaze-independent approaches to BCI. A tactile four-choice BCI resulted in high offline classification accuracies. Yet, online use raised various issues. Although performance was clearly above chance, practical daily life use appeared unlikely when compared to other communication approaches (e.g., partner scanning). Our results emphasize the need for user-centered design in BCI development including identification of the best stimulus modality for a particular user. Finally, the paper discusses feasibility of EEG-based BCI systems for patients in classic locked-in state and compares BCI to other AT solutions that we also tested during the study.

Highlights

Damages to neuromuscular pathways, e.g., due to a stroke in the brainstem, or neurodegenerative diseases such as amyotrophic lateral sclerosis (ALS) or spinal muscular atrophy (SMA) can entail a severe loss of voluntary muscular control
brain-computer interface (BCI) based on classification of event-related potentials (ERP) in the electroencephalogram (EEG) of a patient are most frequently used for communication purpose (Farwell and Donchin, 1988; for review, e.g., Kleih et al, 2011; Mak et al, 2011; Sellers et al, 2012)
In the tactile modality five or more trials led to 100% classification accuracy

Summary

Introduction

E.g., due to a stroke in the brainstem, or neurodegenerative diseases such as amyotrophic lateral sclerosis (ALS) or spinal muscular atrophy (SMA) can entail a severe loss of voluntary muscular control. These patients are summarized under the term locked-in syndrome (LIS; Plum and Posner, 1966) as they are locked into their own body despite often intact cognitive functioning. Several options (e.g., characters for typing words) are iteratively presented and users focus their attention on presentation of the one option they intend to select Such target stimuli will elicit more pronounced ERPs than all other, irrelevant non-target stimuli. The P300 can be elicited in www.frontiersin.org

Methods

Results

Discussion

Conclusion