Sequential Convolutional Neural Networks for classification of cognitive tasks from EEG signals

Cognitive abilities encompass all aspects of mental functioning ranging from simple to complex tasks. These skills have a tremendous effect on our day-to-day routine. The electroencephalogram (EEG) is a powerful tool to analyze brain activities while performing different cognitive tasks. In this paper, we consider four cognitive tasks (Symbol digit modality test, Stroop test, Benton's visual retention test, and Hopkins verbal learning test) along with a baseline task, carried out by healthy subjects and record their EEG. We perform phase-amplitude coupling to extract the features of classification, and segregate them into the tasks, through deep learning algorithms. The Sequential Convolutional Network (SCN) is designed to classify these features. Multi-branch Convolutional Network (MBCN) is also proposed, which is inspired by the ResNeXt architecture and the inception module. The performance of the proposed model is evaluated using the metrics such as accuracy, F1-score, precision, and specificity using the EEG signals collected from the PAC dataset and real-time recording. The performance evaluation reveals that MBCN outperforms SCN by achieving higher accuracy, Fl-score, precision, and sensitivity of 88.33%, 87.9%, 89.18%, and 88.23% respectively. Also, the computational complexity of the MBCN architecture is found to be less than the SCN model. Evaluation results show that the proposed MBCN model outperforms the traditional methods. (C) 2021 Elsevier B.V. All rights reserved.