Non-Gaussianity Detection of EEG Signals Based on a Multivariate Scale Mixture Model for Diagnosis of Epileptic Seizures

Objective: The detection of epileptic seizures from scalp electroencephalogram (EEG) signals can facilitate early diagnosis and treatment. Previous studies suggested that the Gaussianity EEG distributions changes depending on presence or absence seizures; however, no general signal models explain such in within a unified scheme. Methods: This paper describes formulation stochastic model based multivariate scale mixture distribution represent non-Gaussianity caused by fluctuations EEG. In addition, we propose an analysis method combining with filter bank introduce feature representing latent each frequency band. Results: We applied proposed to multichannel data twenty patients focal epilepsy. results showed significant increase during seizures, particularly high-frequency calculated band allowed highly accurate classification seizure non-seizure segments [area under receiver operating characteristic curve (AUC) = 0.881] using only simple threshold. Conclusion: distribution-based capable associated seizures. Experiments simulated real demonstrated validity its applicability detection. Significance: quantified help detect high accuracy.