Back to search

National Technical University of Athens (NTUA)

Ανάπτυξη μεθοδολογιών επεξεργασίας και αξιολόγησης εγκεφαλικών σημάτων από πειραματικές μετρήσεις

Abstract

dc:description

The present study aims to the development of processing methodologies and theevaluation of brain signals recorded during experimental measurements of EEG.The proposed methodology is applied to Electroencephalographic (EEG) signals,Event Related Potentials (ERP) and in one case to facial surface Electromyographic(EMG) signals. At the same time, an attempt to study and evaluate the effect ofElectromagnetic Field (EMF) radiation of different characteristics and parameters onthe electrical signals of the brain is being made.In this context, data acquired by previous experiments were used. Such datawere brain signals recorded under the influence of EMF radiation and EMG andEEG measurements executed in an uncontaminated by EMF radiation environment.This “pool” of data was enriched by new experiments conducted during this study.The latter were designed and executed in collaboration with the University MentalHealth Research Institute and their initial objective was to investigate the effect ofbroadband signals radiation, corresponding to the latest communicationtechnologies, on the brain signals.The data collected from the aforementioned experimental measurementswere processed and analyzed by various techniques.As a first approach, established and standard statistical methods were usedto analyze the P600 component of ERPs recorded during an auditory memory task,in the presence and absence of monochromatic RF radiation, similar to that emittedby mobile phones, operating in 900 and 1800 MHz. The aim was to examine thepatterns of activation of the P600 waveform of the ERPs, applying principalcomponent analysis (PCA) and repeated measures ANOVA, and whether thesepatterns were RF and gender dependent. The results showed that the P600component follows distinct patterns of activation in the anterior, central and posteriorbrain areas, while gender differences are observed simultaneously at severalelectrodes within these areas. Finally, the gender-related functional architecture withregard to the P600 component appears to be RF sensitive. In conclusion, theapplication of the PCA procedure provided an adequate model of the spatiallydistributed event-related dynamics that correspond to the P600 waveform.The same data, as described in the previous paragraph, were used toinvestigate the co-variance among ERP signals, using methods applied to timeseries and specifically the cointegration method. Within this framework,cointegration factor was defined as the ability of an ERP signal to co-vary with otherERP signals. Additionally, its dependance on different EMF conditions and gender,as well as the locations of the electrodes on the scalp, was investigated. Thefindings revealed that women have significantly higher cointegration factor thanmen, while all subjects have increased cointegration factors in the presence of EMF.What is more, the cointegration factor is location dependent, creating a distinctcluster of high cointegration capacity at the central and lateral electrodes of thescalp, in contrast to clusters of low cointegration capacity at the anterior andposterior electrodes. These findings seemed to have distinct similarities with thosederived from standard methodologies of the ERPs.In the course of this thesis, a new experiment was conducted measuringbrain signals of healthy volunteers, while performing an auditory memory task in thepresence and absence of RF radiation, similar to that emitted by a WiFi access pointoperating at 2.4 GHz. For each subject, radiation condition and electrode, theamplitude in the frequency domain of the EEG signal was calculated, using the Fast Fourier transform. The presence of radiation had no effect on the energies of alphaand beta band of male subjects, while it reduced these energies of female subjects,resulting in significantly lower energies, as compared to those of males. Delta andtheta band energies did not experience any noteworthy effect from gender, radiationcondition and their interaction. Conversely, there was a significant interaction effect(gender x radiation) on the energies of alpha and beta rhythms. Interestingly, thispattern was observed for a number of electrodes, which formed two distinct clusters,one located at right- anterior and the second at occipital brain areas. These findingssupported the idea that Wi-Fi signal may influence normal physiology throughchanges in gender related cortical excitability, as reflected by alpha and beta EEGfrequencies.The last part of the present study implicates the development of a tool, whichintends to improve the de- noising of brain signals’ measurements, by intervening atan early stage in them. This tool was based on the theory of Genetic Algorithms. Infact, a method of weighting data, which were recorded during experiments of manyrepetitions, was developed, so as to achieve the normality of their distribution.In this context, EEG and EMG signals, obtained from an experiment conducted inan environment clear of EMF were used. Initially, the distribution of the energyvalues of each repetition was checked in terms of normality. Subsequently, thesevalues were weighted using the GAs and their distribution was checked once again.In both cases the test was performed using three different methods, while theresults obtained suggest both to the need for such a methodology and to theeffectiveness of the proposed procedure.Thereafter, the data acquired from an experiment that was implementedunder the influence of non-monochromatic W-CDMA radiation, were processed.Common weights were calculated and applied to the entire EEG signal. This ideawas extended to ERPs, by applying the same weights, calculated from the restingEEG signal, to the signal that followed the stimulus. The aim again was to transformthe distribution of the values through iterations in each moment, into normal. Thevalidity of the method was tested by means of the values of skewness and kurtosiscoefficient. Generally the method appeared to have an effect, however not in allcases of subjects and electrodes, and sometimes only in one of the coefficients ofskewness and kurtosis.More specifically, it was shown that regarding the EEG signal, the distributionof the values resulting from many repetitions at each moment, is possible toapproach normality, upon weighting of all the data with common weight factors. Onthe other hand, weighting ERP signal with the factors calculated from the restingEEG signal showed a so be it small-scale effect. There have been many cases ofsubjects and electrodes where skewness and kurtosis coefficients found improvedover those corresponding to the original data, although the improvement was notsimilar to that occurred in the EEG signal.The last chapter presents the overall conclusions of the thesis and its scientificcontribution. Finally, references to the future perspectives of this thesis are beingmade.

Degree

thesis:*
Grantor dc:publisher
National Technical University of Athens (NTUA)
Year dc:date
2013

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Maganioti, Argyro
  • Μαγκανιώτη, Αργυρώ

Subjects

dc:subject × 18

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/37276

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Maganioti, Argyro; Μαγκανιώτη, Αργυρώ. Ανάπτυξη μεθοδολογιών επεξεργασίας και αξιολόγησης εγκεφαλικών σημάτων από πειραματικές μετρήσεις. National Technical University of Athens (NTUA), 2013. http://hdl.handle.net/10442/hedi/37276