Detection and Classification of ADHD from EEG Signals Using Tunable Q-Factor Wavelet Transform

Abstract

The automatic identification of Attention Deficit Hyperactivity Disorder (ADHD) is essential for developing ADHD diagnosis tools that assist healthcare professionals. Recently, there has been a lot of interest in ADHD detection from EEG signals because it seemed to be a rapid method for identifying and treating this disorder. This paper proposes a technique for detecting ADHD from EEG signals with the nonlinear features extracted using tunable Q-wavelet transform (TQWT). The 16 channels of EEG signal data are decomposed into the optimal amount of time-frequency sub-bands using the TQWT filter banks. The unique feature vectors are evaluated using Katz and Higuchi nonlinear fractal dimension methods at each decomposed levels. An Artificial Neural Network classifier with a 10-fold cross-validation method is found to be an effective classifier for discriminating ADHD and normal subjects. Different performance metrics reveal that the proposed technique could effectively classify the ADHD and normal subjects with the highest accuracy. The statistical analysis showed that the Katz and Higuchi nonlinear feature estimation methods provide potential features that can be classified with high accuracy, sensitivity, and specificity and is suitable for automatic detection of ADHD. The proposed system is capable of accurately distinguishing between ADHD and non-ADHD subjects with a maximum accuracy of 100%.