Epilepsy detection method based on magnitude squared coherence and machine learning

Epilepsy, a chronic non-communicable condition characterized by paroxysmal aberrant supersynchronous electrical activity of brain neurons, is one of the most prevalent complicated brain diseases. In this study, we suggested a real-time method for detecting epileptic seizures using electroencephalography (EEG). Various techniques, including the discrete wavelet transform and the methodology for calculating the eight eigenvalues, are used to extract characteristics at various sub-frequency levels. Then, a support vector machine is used to classify the user into one of three states (healthy, seizure-free, or seizure-active), and lastly, the RUS Boosted tree Ensemble approach is used to identify the start of a seizure in real time. An approach known as magnitude-squared coherence (MSC) fusion was created in order to identify EEG data. This algorithm was built on the basis of several statistical factors and crossing frequency properties. An additional perk is that the important features were first chosen with the assistance of a genetic algorithm. It has been determined whether or not the proposed approach is able to differentiate between normal and EEG patterns by using data obtained from the University of Bonn's EEG recordings. Experiments have shown that the MSC fusion that has been proposed might potentially improve the EEG categorization ability. Using the MSCs fusion method as recommended also makes it feasible to diagnose epilepsy from EEG data with a precision of one hundred percent.