A novel Brain Computer Interface for classification of social joint attention in Autism and comparison of 3 experimental setups: a feasibility study

Publication date: Available online 29 July 2017
Source:Journal of Neuroscience Methods
Author(s): Carlos P. Amaral, Marco A. Simões, Susana Mouga, João Andrade, Miguel Castelo-Branco
BackgroundWe present a novel virtual-reality P300-based Brain Computer Interface (BCI) paradigm using social cues to direct the focus of attention. We combined interactive immersive virtual-reality (VR) technology with the properties of P300 signals in a training tool which can be used in social attention disorders such as autism spectrum disorder (ASD).New methodWe tested the novel social attention training paradigm (P300-based BCI paradigm for rehabilitation of joint-attention skills) in 13 healthy participants, in 3 EEG systems. The more suitable setup was tested online with 4 ASD subjects. Statistical accuracy was assessed based on the detection of P300, using spatial filtering and a Naïve-Bayes classifier.ResultsWe compared: 1 − g.Mobilab+ (active dry-electrodes, wireless transmission); 2 − g.Nautilus (active electrodes, wireless transmission); 3 − V-Amp with actiCAP Xpress dry-electrodes. Significant statistical classification was achieved in all systems. g.Nautilus proved to be the best performing system in terms of accuracy in the detection of P300, preparation time, speed and reported comfort. Proof of concept tests in ASD participants proved that this setup is feasible for training joint attention skills in ASD.Comparison with Existing methodsThis work provides a unique combination of 'easy-to-use' BCI systems with new technologies such as VR to train joint-attention skills in autism.ConclusionsOur P300 BCI paradigm is feasible for future Phase I/II clinical trials to train joint-attention skills, with successful classification within few trials, online in ASD participants. The g.Nautilus system is the best performing one to use with the developed BCI setup.



http://ift.tt/2tSndft