Tracking Epileptogenesis Progressions with Layered Fuzzy K-means and K-medoid Clustering

This paper illustrates a method that identifies abnormal neurological events associated with acute brain injuries and seizures. The data is derived from intracortical electrodes that transmit EEG recordings that contain substantial amounts of noise. The output for the identification is specifically designed to be fed back into nan artificial intelligence unit the authors have designed that predict seizures in rats within 6 seconds of onset. This data will verify whether or not the prediction as accurate or not. The methodology set forth in this paper refiects the results of an ongoing research program that has investigated multiple techniques that have computational simplicity and are able to identify events with high precision. The paper presents our most recent in-house algorithm for seizure detection. The authors present this novel approach through the analysis of 3 channel EEG recordings from 3 rats that have 10 seizures each. The algorithm uses a Riemann sum analysis that compares the time to zero crossings with the absolute amplitude of the intervening signal. Herein, the authors present a system that incorporates clustering analysis using fuzzy K-means (FCM) and K-medoid clustering that identifies and separates artifact and normal states of animal with states seizure.