A novel 2-piece rose spiral curve model: Application in epileptic EEG classification

This paper aimed to provide an innovative 2D phase space model and evaluate its performance in categorizing electroencephalogram (EEG) signals of normal and epileptic patients. The main contributions of the current study are as follows. (1) For the first time, it was proposed a new 2D model based on a 2-piece Rose Spiral Curve (RSC) in EEG analysis. (2) The trajectory patterns of the model were examined for signals of different natures, including constant, periodic, random, and EEG. (3) It was presented some benchmarks for quantifying the trajectory patterns. (4) Applying these measures, support vector machine, Naïve Bayes, AdaBoost, and K-nearest neighbor were used in the epileptic EEG classification problem to estimate the method efficiency. Bonn database, which takes account of EEG signals of healthy, in the course of an epileptic seizure occurrence, and seizure-free cases, was assessed. The results indicated that the proposed framework provided a correct rate of 100% for recognizing healthy subjects and the EEGs with seizure activity. Additionally, seizure-free brain activity was classified with an accuracy of 96.7%. To conclude, the proposed RSC model can be suitable for serving as a computer-aided diagnosis tool for epileptic seizures.