Multifractal detrended fluctuation analysis based novel feature extraction technique for automated detection of focal and non‐focal electroencephalogram signals

Abstract

In this contribution, a method to segregate electroencephalogram (EEG) signals into focal (F) and non-focal (NF) groups has been proposed, employing a novel multifractal detrended fluctuation analysis (MFDFA)-based feature sets. Manifestations in the fractal behaviour occurring due to the subtle morphological changes in F and NF EEG signals, can serve as an essential presurgical intervention for automated detection of structural epileptogenic area within the human brain. Considering the above-said fact, in the present approach, EEG signals acquired from a publicly available database, are analysed using multifractal parameters to investigate the complex, non-linear and stochastic fluctuations. Based on MFDFA of EEG signals, four statistically significant, new set of features have been extracted, which are eventually being used as inputs to a support vector machines and k-nearest-neighbour classifiers for the purpose of classification of EEG signals. It has been observed that the proposed MFDFA aided feature extraction method delivers quite commensurable and even better results in discriminating F and NF EEG signals, compared with the existing methods studied on the similar database.