DETECTION OF SEIZURE TYPES FROM THE WAVELET ENERGY OF SCALP EEG

Epilepsy is a disabling and devastating neurological disorder, characterized by recurrent seizures. These seizures are caused by the abrupt disturbance of the brain and are categorized into various types based on the clinical manifestations and localization. Seizures with clinical manifestations require immediate medical attention. In this work, an attempt has been made to differentiate the seizures with and without clinical manifestations using wavelet energy of scalp EEG signals. For this purpose, scalp EEG records from the publically available Temple University Hospital (TUH) database are considered in this work. The first four seconds of scalp EEG during seizure is subjected to seven-level Daubechies (db4) wavelet decomposition and energy is extracted from the resultant coefficients. These features are used to develop k-Nearest Neighbor (k-NN) classification model for the detection. The results show that the energy associated with most of the sub-bands exhibits significant difference (p<0.05) in these two types of seizures. It is found that the machine learning model based on k-NN achieves an accuracy of 87.6% and precision of 87.3%. Therefore, it appears that the proposed approach could aid in detecting life-threatening seizures in clinical settings.