Recurrence Quantification Analysis of EEG signals for Children with ASD

Abstract

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by impairment in sensory modulation, repetitive behavior etc. It would lead to difficulties in adaptive behavior and intellectual functioning. Subjective scales as childhood autism rating scale, 3Di, etc. are available to assess the symptoms of Autism. Currently there are no reliable objective diagnostic methods available for assessment of Autism. Also, Early diagnosis of will help in designing customized training and putting those kids in regular stream. The purpose of this research is to observe the response of the brain for auditory/visual stimuli in typically Developing (TD) and children with autism through electroencephalography (EEG). Application of nonlinear methods for EEG signal analysis may help in characterization of brain activity to describe the neurophysiological commonalities and differences between typically developing and autism children. Among the various non-linear methods, the underlying dynamics can be analyzed well with Recurrent Quantification Analysis (RQA). But, the performance of RQA based classification depends on the choice of parameters like embedding dimension, time delay, neighborhood selection and distance metric. Different experiments were conducted by varying methods for neighborhood selection and distance metric. In this research, for better information retrieval cosine distance metric is additionally considered for analysis and compared      with other distance metrics in RQA. Each computational combination of RQA measures and the responding channels were analyzed and discussed. It was observed that FAN neighborhood with cosine distance parameters were able to discriminate between ASD and TD.