Multilayer Autoencoders and EM - PCA with Genetic Algorithm for Epilepsy Classification from EEG

The structural components and the activities are marvelous for the human brain but it is prone to more neurological disorders and one such disorder is epilepsy. The term epilepsy is referred to a particular condition when a person has recurrent seizures. When abnormal discharges occur in the electrical activities of the brain cells, seizures occur and it gives rise to many abnormal behaviors like unusual perceptions, involuntary muscle movement and a high disturbed level of consciousness. Electroencephalography (EEG) serves as an indispensable tool for the analysis and diagnosis of epilepsy. In this work, Multilayer Autoencoders and Expectation - Maximization Based Principal Component Analysis (EM-PCA) are used to reduce the dimensions of the data and then it is classified with the help of Genetic Algorithm (GA). Results show that an average classification accuracy of 93.78% is obtained when Autoencoders are employed and when EM-PCA is utilized an average classification accuracy of 93.92% is obtained.