Brain Computer Interface (BCI) As a Potential Tool for Characterizing the Cognition of Non-Verbal Individuals with Autism

Background:  Individuals with severe autism, who do not possess functional spoken language, are often called non-verbal, but this does not necessarily imply the lack of receptive language or impaired cognition. It has been suggested that for some of these individuals, the main barrier for communication is a severe difficulty in the initiation and control of reliable intentional actions, i.e.an "output problem", and they are therefore unable to show their cognitive skills.

Objectives:  Develop a brain computer interface (BCI) technique suitable for assessing the cognition of non-verbal individuals with autism, and demonstrate it at a first stage on a small group of non-autistic adults. 

Methods: We have chosen an EEG-based BCI method which could be suitable for non-communicating individuals with autism as follows: (1) Steady-state visual evoked potentials (SSVEP) that allows longer presentations (e.g. 5 sec) which are less sensitive to lapses of attention or fixation than event-driven methods, with high flicker-frequency tagging (>40 Hz) that minimizes the risk for epilepsy; (2) Wide screen (~40 deg field-of-view, using a projector) in a dim room that minimizes distractions; (3) Presenting two or four choices, either static or moving slowly and independently around the screen to minimizes spatial locking and biases typical of severe autism. We used an 8-electrode g.tec (Austria) EEG system, with a classifier first obtained by exposing the subject to simple stimuli using the minimum energy EEG method. Ten communicating adults were tested on a set of twenty 2 and 4-choice simple questions with real time feedback.

Results:  There was an overall >90% matching between the answers obtained by the BCI system and the answers reported separately by the observers.

Conclusions:  Our BCI method could be used to assess cognitive skills and knowledge of adult individuals, without needing explicit communication. We are currently exploring its use for characterizing the cognition of non-communicating individuals with severe autism.