Title Metoda analize funkcionalne povezanosti mozga korištenjem koeficijenata kompleksne Pearsonove korelacije i adaptivne širine prozora Title (english) A method for brain functional connection analysis using complex pearson correlation coefficients and adaptive window width Author Zoran Šverko Mentor Saša Vlahinić (mentor)Mentor Miroslav Vrankić (komentor)Mentor Peter Rogelj Committee member Nino Stojković (predsjednik povjerenstva)Committee member Ivan Volarić (član povjerenstva)Committee member Željka Tomasović (član povjerenstva)Granter University of Rijeka Defense date and country 2023-10-30, Croatia Scientific / art field, TECHNICAL SCIENCES Universal decimal classificationUDC ) 62 - Engineering. Technology in general Abstract Skupovi veza izmedu različitih neurona ili grupa neurona su u pozadini cjelokupnog ljudskog razmišljanja, tj. djelovanja i reagiranja. Na temelju elektroencefalografskih (engl. electroencephalography - EEG) signala kroz ovaj rad proučavaju se i procjenjuju razine tih veza. U tu svrhu razvijena je mjera na temelju koeficijenata kompleksne Pearsonove korelacije (engl. complex Pearson correlation coefficient - CPCC) koja ima svojstvo identificiranja razine moždane povezanosti s i bez utjecaja volumne provodljivosti. Iako je Pearsonov koeficijent korelacije opće prihvaćena mjera statističkih odnosa između promatranih slučajnih varijabli i signala, kao takva ne koristi se pri analizi EEG podataka. Značenje Pearsonovog koeficijenta korelacije pri analizi EEG-a nije jednostavno i lako razumljivo. U ovom radu uspoređena je razvijena mjera s najčešće korištenim neusmjerenim mjerama statičke funkcionalne povezanosti, a to su mjera zaključavanja faze ( ˇ engl. phase locking value - PLV) te ponderirani indeks faznog kašnjenja (engl. weighted phase lag index - wPLI). Dan je analitički prikaz odnosa između mjera. Zatim je praktičnom usporedbom koristeći sintetičke i stvarne EEG signale dokazan odnos između mjera. Nadalje, odnosi između promatranih mjera povezanosti opisani su koristeći korelacijske vrijednosti između njih, koje su za usporedbu apsolutne vrijednosti CPCC-a (engl. absolute value of complex Pearson correlation coefficient - absCPCC) i PLV-a ne niže od 0,97, a za imaginarnu komponentu CPCC-a (engl. imaginary component of complex Pearson correlation coefficient - imCPCC) i wPLI-a ne niže od 0,92, za sve promatrane frekvencijske pojaseve. Time je pokazano da razvijena CPCC mjera objedinjuje informacije sadržane u PLV i wPLI mjeri. Takoder u ovom radu razvijena je metoda procjene dinamičke funkcionalne povezanosti mozga. Razvijena metoda relativnog presjecišta intervala pouzdanosti za imaginarnu komponentu koeficijenta kompleksne Pearsonove korelacije (engl. relativ intersection of confidence intervals for imaginary component of complex Pearson correlation coefficient - RICI-imCPCC) temelji se na prilagodljivoj veličini prozora promatranja za svaki vremenski uzorak i imaginarnoj komponenti razvijene CPCC mjere statičke funkcionalne povezanosti. Ova metoda nadilazi nedostatke najcešće korištene metode analize pomičnim prozorom konstantne širine, kao što su niska vremenska razlučivost i niska pouzdanost za široke prozore te velika osjetljivost na šum za uske prozore, što dovodi do niske točnosti procjene. Razvijena metoda nadilazi nedostatke dinamičkim podešavanjem širine prozora pomoću algoritma relativnog presjecišta intervala pouzdanosti ( ´ engl. relative intersection of confidence intervals - RICI), koji se temelji na statističkim svojstvima područja oko promatranog vremenskog uzorka. Osim usporedbe s najcešće korištenom metodom analize pomičnim prozorom konstantne širine, usporedba je provedena i s jednoskaliranim vremenski-ovisnim algoritmom (engl. single-scale time-dependent - SSTD) odabira prozora za procjenu indeksa funkcionalne povezanosti. Krivulje procjene dobivene korištenjem tih algoritama daju uvid u neovisnost o utjecajima šuma pri korištenju RICI algoritma te ujedno i veću točnost procjene u odnosu na korištenje SSTD algoritma.
Abstract (english) The connections between different neurons or groups of neurons underlie all human thinking, action, and reaction. This paper examines and evaluates the levels of these connections based on electroencephalographic (EEG) signals. This is achieved by developing a measure based on the complex Pearson correlation coefficients (CPCC) that can identify the level of brain connectivity with and without the influence of volume conduction. While the Pearson correlation coefficient is a generally accepted measure of statistical relationships between observed random variables, it is not typically used in EEG data analysis. The meaning of Pearson’s correlation coefficient in EEG analysis is not straightforward and easily understood. This study compares the developed CPCC measure with the most commonly used undirected measures of static functional connectivity, namely the phase locking value (PLV) and the weighted phase lag index (wPLI). An analytical relationship between the measures has been provided, and an experimental comparison using synthetic and real-life EEG signals is used to demonstrate the relationship between the measures. Additionally, the relationships between the observed connectivity measures have been described using correlation values between them, which are not lower than 0.97 between the absolute value of CPCC (absCPCC) and the PLV, and not lower than 0.92 between the imaginary component of CPCC (imCPCC) and the wPLI, for all observed frequency bands. The study shows that the developed CPCC measure incorporates information from both PLV and wPLI in one complex index. Furthermore, this study develops a method for assessing dynamic functional connectivity. The method for relative intersection of confidence intervals for the imaginary
component of the complex Pearson correlation coefficient (RICI-imCPCC) is based on an adaptive window size for each time sample and the imaginary component of the developed CPCC measure of static functional connectivity. This method overcomes the limitations of the most commonly used method of analysis with a sliding window of constant size, such as low temporal resolution and low reliability for wide windows and high sensitivity to noise for narrow windows, which lead to low accuracy in estimation.The developed method overcomes these limitations by dynamically adjusting the window size using the relative intersection of confidence intervals (RICI) algorithm, which is based on the statistical properties of the area around the observed time sample. In addition, the comparison has been performed with the most commonly used
method of analysis with a constant sliding window, the study also compares the RICI algorithm for selecting the optimal window width with the single-scale time-dependent (SSTD) algorithm for window selection. The estimation curves obtained using these algorithms provide insight into the independence of noise influences when using the RICI algorithm and also greater accuracy in estimation compared to using the SSTD algorithm.
Keywords
statička funkcionalna povezanost
dinamička funkcionalna povezanost
koeficijent kompleksne Pearsonove korelacije
obrada EEG signala
Keywords (english)
static functional connectivity
dynamic functional connectivity
complex Pearson correlation coefficient
EEG signal processing
Language croatian URN:NBN urn:nbn:hr:190:136095 Promotion 2023 Study programme Title: Postgraduate university doctoral study in the area of Engineering sciences, in the field of Electrical engineering; specializations in: Electrical Engineering Type of resource Text Extent VIII, 143 str.; 30 cm File origin Born digital Access conditions Open access Terms of use Created on 2023-10-30 13:35:24
