Neurological Disorder Detection of Brain Abnormal Activities Using a New Enhanced Computer-aided Model

Abstract

In the field of neurodevelopmental disorders, Autism Spectrum Disorders (ASD) are recognized as one of the dramatically increased etiologically and clinically heterogeneous diseases. Although, increasing the number of children who have difficulties in communication or suffer from sudden malfunction of the brain, the current diagnostic approaches for those kind of disease are time-consuming and are mainly based on clinical interviews. In this paper, a new enhanced diagnostic model is introduced addressing many of the challenges which threats the analysis of Electroencephalography (EEG) signals. A preprocessing stage is proposed to choose the key segment of EEG channel and remove the artifacts in the EEG signals to enhance their quality. The proposed model uses a set of discriminative features based on discrete wavelet transform (DWT) such as skewness, standard division, shannon entropy and relative wave energy. Also, extracting cross correction between brain regions to detect abnormal connectivity and synchronization. Two EEG datasets are used to verify the accuracy of the proposed model. The fusion of two EEG dataset helps in building a more generalized mode to diagnose epilepsy and ASD. In the fused dataset, the combination of the mentioned features and Random Forest have produced a very promising diagnosis result with minimum diagnostic time, with an average accuracy equals to 96.78%. The proposed model obtained better classification accuracy compared to the existing methods.