Nonlinear analysis of EEG for seizure prediction

Epilepsy is one of the most common neurological disorders of the human population. Seizure prediction is an important research area which aims to warn the patient and the epileptologist about an impending seizure so that the necessary preventive steps can be taken well in advance. In this paper an algorithm is developed to predict seizures by using nonlinear features namely entropy and approximate entropy (ApEn). Analysis is carried out on scalp electroencephalogram (EEG) data and the seizure prediction time varies from 5 to 60 minutes. The false detection rate varies from 0.38/hour to 1.00/hour. With all the tested seizures being detected and low false detection rate per hour obtained, this algorithm shows that seizure prediction can be done by using scalp EEG which paves the way for future development of portable and safe seizure prediction systems using scalp EEG as input data.