Computer-Aided Diagnosis Framework for ADHD Detection Using Quantitative EEG

Abstract

Attention Deficit Hyperactivity Disorder (ADHD) is a mental disorder that is marked by abnormally high levels of impulsivity, hyperactivity and inattention. One of the methods to detect and diagnose brain disorders is Electroencephalogram (EEG). This paper proposes a framework that uses Quantitative Electroencephalogram (QEEG) features to diagnose ADHD in children. A 19-channel EEG signal is used to extract the spectral, amplitude, functional connectivity and Range EEG (rEEG) features from five frequency bands to diagnose ADHD children. Four feature selection methods: ANOVA, Chi-square, Gini Index and Information Gain are used to rank the QEEG features based on their relative importance to the classification task. The feature ranks are then averaged and the top-600 most discriminative features are passed as the input to an array of classifiers. We carried out experiments on a benchmark ADHD dataset and proved that our proposed framework gives better accuracy as compared to the state of the art. The highest accuracy of 81.82% is obtained with the Random Forest classifier, while the KNN, SVM and ANN classifiers yield accuracies of 78.51%, 76.86% and 76.93%, respectively.