An adaptive neuro-fuzzy inference system for the physiological presentation of seizure disorder

Abstract

Seizure is the clinical manifestation of an excessive, hypersynchronous discharge of a population of cortical neurons accompanied by indescribable "pins-and needles-like" bodily sensations, smells or sounds, fear or depression, hallucinations, momentary jerks or head nods, staring with loss of awareness, and convulsive movements (i.e., involuntary muscle contractions) lasting for some seconds to a few minutes. In this work, an attempt is made to promote a better understanding of seizure disorder by proposing an adaptive neuro-fuzzy simulation model as a tool for capturing the physiological presentation of the disorder. Decision making was performed in two stages, namely the feature extractions using Microsoft Excel for corresponding digital value of the waveform of the EEG recordings of a seizure and those of a non-seizure patient directly from the EEG machine, and the transient features are accurately captured and localized in both time and amplitude. This extracted data were used for our Adaptive Neuro-Fuzzy Inference System (ANFIS) training and the ANFIS was trained with the backpropagation gradient descent method in combination with the least squares method to establish the validity of our ANFIS. The result shows an accuracy of 90.7% of predictions as the number of epochs increase.