EEG Based Decision Support System to Diagnose Adults with ADHD

EEG is considered as a useful tool for diagnosis of neurological disorders due to its capability to depict rapid transient cortical events. Attention Deficit Hyperactivity Disorder (ADHD) is estimated to be prevalent among 5% of children and 3.4% of adults worldwide. In this paper, an EEG based decision support system is proposed using Autoregressive (AR) modeling and eigenvector method for diagnosis of ADHD adults. For this, AR model parameters are extracted from rest state EEG signals of 30 ADHD and 30 control adults using burg, covariance and yule-walker methods. Absolute and relative powers in various frequency bands are calculated through power spectral density estimates obtained from eigenvector method. Correlation based Feature Selection (CFS) is applied to find most discriminative feature set. Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) algorithms are employed to classify brain signals of ADHD and control adults. Experimental results demonstrated that covariance method parameters under an eyes-open state have given highest accuracy of 85% in discrimination of two groups. Burg and covariance method parameters under an eyes-closed state have yielded next higher classification accuracy of 80%. Relative alpha power under an eyes-open state has provided highest accuracy of 80%. It has been observed that AR model parameters can better discriminate between ADHD and control adults as compared to spectral features.