Identification of Neural Markers for Autism Spectral Disorder

Khuntia, Adyasha Tejaswi (2018) Identification of Neural Markers for Autism Spectral Disorder. Masters thesis, Indian Institute of Science Education and Research Kolkata.

PDF (MS disertation of Adyasha Tejaswi Khuntia (13MS020)) 13MS020.pdf - Submitted Version Restricted to Repository staff only Download (2MB)

Abstract

Autism Spectral Disorder(ASD) is a neurobiological disorder in which the affected individuals are deficit in social interaction, communication and social reciprocity. Emotion processing impairment is a prominent characteristic feature of ASD. Hence cognitive tasks pertaining to emotion processing are often selected as tasks to distinguish the ASD children (non-typical) from the typically developing (TD) ones. Our objective was to correctly identify children diagnosed with ASD and typically developed children from their neural data while they were performing a cognitive task. We used data from a previously conducted experiment where high-density EEG system was used to record neural data while subjects were shown different images. Five types of images were shown in random order to both typical and non-typical children: Neutral, Happy, Fear, Cartoon and Tree. Machine learning algorithms were used to classify the two data sets. Two types of pattern classifiers were used: Class-wise Principle Component Analysis (CPCA) and Support Vector Machine (SVM). Both the classifiers could correctly predict children with ASD with relatively high accuracy (~65%). The highest activity was observed between 50 ms to 170 ms after the stimulus which was confirmed by plotting the scalp topoplots. Short time Fourier transformation was done to obtain the spectrograms for each condition. Different brain waves were characterized: alpha (8-12 Hz), beta (12-38 Hz), delta (0.5-3 Hz), gamma (38-42 Hz) and theta (3-8 Hz) for both ASD and TD groups. The Power Spectra Density obtained was used for plotting topoplots to understand the discriminatory activity arising due to the specific waves. Source Reconstruction was used to localize the electrical activity of the brain and has been done to elucidate the distinguished active regions in the brain for the two classes. In this study we could correctly predict ASD. We hypothesize that the differences attribute to visual and attention processing. This study will be beneficial to get the neural markers specific for ASD.

Actions (login required)

View Item