Parametric modeling of EEG signals with real patient data for simulating seizures and pre-seizures

Abstract

Numerous theories and models have been developed to associate various findings or in relating EEG patterns to develop a software simulators. In this paper, a Dynamic model for simulating the EEG signal has been developed with empirical reference to real EEG signals from patients suffering from Seizure and Partial Seizure. Real EEG data set can be obtained in either .edf or .tdms or .txt formats from any clinical patient tests or database repository. The proposed model for the EEG signal has led to the development of a simulator which can be used to obtain any number of samples of data of a specific type (Normal, Pre-Seizure, and Seizure) and can be used by researchers for algorithmic testing. The presented simulator has a core of 22 patient's data with a variety of ages and gender selection options with possible connectivity to hardware based modules to generate the real EEG signal for external use as well. One can simulate, validate and test the detection algorithms beforehand, before actual clinical testing of the algorithms. Further, one can also develop pre-prediction algorithms for Seizure and pre-seizure states of a patient to take appropriate precautions just before the actual occurrence of the seizure. The model is based on the conventional ARX structure with subset frequencies from the real EEG signal used as excitation input. When plotted together, the resemblance between the original and simulated signals was very significant thus providing with a means to keep simulating with those frequencies to whatever length needed, with whatever variability in terms of amplitude and patient specific parameters.