Advanced Electroencephalogram Processing: Automatic Clustering of EEG Components

Advanced Electroencephalogram Processing: Automatic Clustering of EEG Components

Diana Rashidovna Golomolzina (Laboratory of Intel-NSU, Novosibirsk State University, Novosibirsk, Russia), Maxim Alexandrovich Gorodnichev (Institute of Computational Mathematics and Mathematical Geophysics SB RAS, Laboratory of Intel-NSU, Novosibirsk State University, Novosibirsk, Russia), Evgeny Andreevich Levin (Novosibirsk Research Institute of Circulation Pathology, Novosibirsk, Russia & Institute of Physiology and Fundamental Medicine, Novosibirsk, Russia), Alexander Nikolaevich Savostyanov (Institute of Physiology and Fundamental Medicine, Novosibirsk State University, Novosibirsk, Russia & Tomsk State University, Tomsk, Russia), Ekaterina Pavlovna Yablokova (Novosibirsk State University, Novosibirsk, Russia), Arthur C. Tsai (Institute of Statistical Science, Academia Sinica, Taipei, Taiwan), Mikhail Sergeevich Zaleshin (Tomsk State University, Tomsk, Russia), Anna Vasil'evna Budakova (Tomsk State University, Tomsk, Russia), Alexander Evgenyevich Saprygin (Novosibirsk State University, Novosibirsk, Russia), Mikhail Anatolyevich Remnev (Novosibirsk State University, Novosibirsk, Russia) and Nikolay Vladimirovich Smirnov (Novosibirsk State University, Novosibirsk, Russia)
Copyright: © 2014 |Pages: 21
DOI: 10.4018/ijehmc.2014040103
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

The study of electroencephalography (EEG) data can involve independent component analysis and further clustering of the components according to relation of the components to certain processes in a brain or to external sources of electricity such as muscular motion impulses, electrical fields inducted by power mains, electrostatic discharges, etc. At present, known methods for clustering of components are costly because require additional measurements with magnetic-resonance imaging (MRI), for example, or have accuracy restrictions if only EEG data is analyzed. A new method and algorithm for automatic clustering of physiologically similar but statistically independent EEG components is described in this paper. Developed clustering algorithm has been compared with algorithms implemented in the EEGLab toolbox. The paper contains results of algorithms testing on real EEG data obtained under two experimental tasks: voluntary movement control under conditions of stop-signal paradigm and syntactical error recognition in written sentences. The experimental evaluation demonstrated more than 90% correspondence between the results of automatic clustering and clustering made by an expert physiologist.
Article Preview

Introduction

Electroencephalography (EEG) is a method of noninvasive exploration of functional brain activity. It records electric potentials generated by cortical and, in lesser extent, subcortical neurons by reading the signal from electrodes placed over the head (see Figure 1). EEG plays an important role in investigation of neurocognitive processes, in diagnosis, treatment and prognosis of some diseases.

Figure 1.

The subject wearing the EEG cap

Different brain processes can take place simultaneously. Some of these processes are spatially separated, other processes take place in the same areas of the brain, and EEG recordings reflect superposition of all the signals originating from many brain processes and noise. For example, non-brain noise can occur from eye movement or from external electrical sources such as power mains. Brain activity can be connected with sensory, motor, emotion processes, etc. In order to understand brain functioning, a researcher should be able to distinguish the contributions of physiologically different sources to the measured EEG signals.

One of the methods for EEG analysis is the decomposition of initial signal recorded from head surface into several linearly independent signals. Independent component analysis (ICA) is a powerful signal processing technique widely applied in biomedicine, telecommunication, finance and machine vision, etc. And ICA is applied for isolating artifacts and/or cortical processes from EEG data (Makeig et al., 1996; Laubach et al., 1999; Hyvärinen and Oja, 2000; Vigario and Oja, 2000). The ICA method is included in many modern EEG processing software packages, such as EEGlab toolbox, BESA, CURRY Scan 7, WinEEG, BrainVision Analyzer2 and others.

The use of independent components analysis (ICA) for processing of the EEG is based on the model presuming that the EEG signal on each electrode is the linear combination of signals from several mutually independent sources of activity, which could be as of brain origin, as of non-brain (e.g. cranial muscles activity, electrooculogram, electrocardiogram). The ICA was proposed as a method for solving the blind source separation problem, i.e. for decomposing the N source signals from the N recorded linear combinations of these sources, while assuming as little as possible about the nature of the source signals, and, therefore it fits with this model almost ideally.

In the result of ICA, the recorded EEG signals are transformed into the set of pairwise statistically independent components which are linear combinations of initial signals and have the same dimension. However, among these statistically independent components some could be physiologically dependent. It means that one physiological process can be represented by several statistically independent components. Also, some of the ICA components can represent noise (non-brain, external) signals arising from different sources. In medical and diagnostic investigations, it is important to get rid of the noise and be able to select components that reflect one or several brain processes of interest. This requires development of a tool for automatic selection of defined physiological processes. The first step to the selection of the components according to their nature is to group them into clusters based on the criteria of functional similarity.

Various neurophysiologic processes could differ by their locations at cortex surface, by timing and by characteristic frequency ranges (Basar, 1999; Klimesch 1999; Knyazev, 2007; Li et al., 2012). Consequently, the criteria for clustering of the independent components will be coincidence by one, some, or all three characteristics: location, time and/or frequency. Researcher evaluates the components as similar or non-similar and basing on his/her own decision combines them into one or a few groups (clusters). However, such clustering “by hand” is subjective and often gives irreproducible results. Besides, it is usually highly time consuming for the researcher.

Therefore there is a need in developing the software for the independent components clustering which would be “as automated as possible” and, consequently, the objectives of the present work were:

Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 9: 4 Issues (2018): 1 Released, 3 Forthcoming
Volume 8: 4 Issues (2017)
Volume 7: 4 Issues (2016)
Volume 6: 4 Issues (2015)
Volume 5: 4 Issues (2014)
Volume 4: 4 Issues (2013)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing