A Methodical Approach to Epileptic Classification with Multi-Scale Patterns

The life cycle of epileptic patients alternates between inter-ictal and ictal stage, visual inspection and analysis is often used in seizure detection and diagnosis clinically, which is time consuming and error-prone. Modern computer technology shows that different stages of epileptic seizure can be classified by machine learning and pattern recognition. The main work of this paper proposed an automatic seizure detection approach methodically. First the EEG signals were decomposed into respective brain rhythms waves, then the Hurst index were used as features to design discriminative classifiers to classify inter-ictal and ictal EEG segments. The performance of the model was evaluated on the public and private datasets. The results revealed the good accuracy. Thus, this paper may serve as a benchmark in seizure detection procedure and advance the classification accuracy of a seizure.