PCA and K-means clustering for classification of epilepsy risk levels from EEG signals — A comparitive study between them

Epilepsy is the most commonly occurring neurological disorder when compared to other neurological disorders like dementia and chronic headaches and is characterized by recurrent seizures. To understand the mechanism of a particular seizure, the prediction of epileptic seizure plays a significant role in it. The EEG signals are often relied upon to study and analyze brain's behavior during seizures. In this paper, Approximate Entropy (ApEn) is considered as a Feature Extraction Technique followed by K-means Clustering and Principal Component Analysis (PCA) as Post Classifiers for the Classification of Epilepsy Risk levels from EEG Signals. The benchmark parameters are analyzed in terms of Performance Index (PI), Quality Values (QV), Specificity, Sensitivity, Accuracy and Time Delay.