Autism Spectrum Disorder Indicative Decision Support System using Psychophysiological Data

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder which affects social interaction, communication, and leads to restricted and repetitive behavior. The phenotypic heterogeneity of ASD makes timely and accurate diagnosis challenging, requiring highly trained clinical practitioners. The development of automated approaches to ASD classification, based on integrated psychophysiological measures, may one day help expedite the diagnostic process.

The main objective of this research is to design and develop a machine learning based decision support system to classify ASD using Electroencephalography (EEG) and facial thermographic data indicating the severity range of the disorder. The research was carried out in multiple phases.

In phase one, the possibilities to use the minimum number of EEG channels were explored using statistical features extracted from noise filtered EEG data before and after Discrete Wavelet Transform and several learning models. Using Random Forest and Correlation-based Feature Selection, an accuracy level of 93% was achieved.

In the second phase, as an extension to phase one, the possibility of improving the classification with facial thermographic data was evaluated. The methodology used in this phase extracts a variety of feature sets and evaluates the possibility of using several learning models. Mean, standard deviation, and entropy values of the EEG signals and mean temperature values of regions of interest (ROIs) in facial thermograms were extracted as features. The integration of EEG and thermographic features achieved an accuracy of 94% with both Logistic Regression and Multi-Layer Perceptron classifiers.

In the third phase, a multi-class classification was employed to identify the severity level of ASD.

Finally, a web-based decision support system named "ASDGenus" for ASD classification was developed based on the third phase.

Publications