Modified-Distribution Entropy as the Features for the Detection of Epileptic Seizures

Epilepsy is one of the most common chronic neurological disorders, and therefore, diagnosis and treatment methods are urgently needed for these patients. Many methods and algorithms that can detect seizures in epileptic patients have been proposed. Electroencephalogram (EEG) is one of helpful tools for investigating epilepsy forms in patients, however, an expert in the neurological field must perform a visual inspection to identify a seizure. Such analyses require longer time because of the huge dataset recorded from many electrodes which are put on the human scalp. With the non-stationary nature of EEG, especially during the abnormality periods, entropy measures gain more interest in the field. In this work, by exploring the advantages of both reliable state-of-the-art entropies, fuzzy entropy and distribution entropy, a modified-Distribution entropy (mDistEn) for epilepsy detection is proposed. As the results, the proposed mDistEn method can successfully achieve the same consistency and better accuracy than using the state-of-the-art entropies. The mDistEn corresponds to higher Area Under the Curve (AUC) values compared with the fuzzy entropy and the distribution entropy and yields 92% classification accuracy.