Дисертації з теми "Electroencephalographic (EEG)"

Since Csikszentmihalyi identified the psychological experience of flow over 40 years ago, the experiences have been heralded as the optimum human function and prescriptive to high levels of well-being and quality of life. Csikszentmihalyi theorized that flow represented an autonomous reality that represented an altered state unlike any other human experience. Flow states emerged from intrinsically motivated behavior that represented a fragile balance between the level of enjoyment from novel task stimulation and a sense of self-efficacy required to meet the specific task demands. However, flow is not well understood and research is skewed toward to phenomenological investigations that described the nature of the experience and many of the significant variables of interest across a diverse range of activities. The lack of experimental exploration of flow has created fundamental research gaps. The general problem is that flow is predictive and related to positive psychological outcomes; however, current assessment methodologies and research have not provided the functional neuroanatomy involved. The purpose of this quantitative experimental study was to examine the hypofrontality theory that a flow state occurs concurrently with decreased cognitive activation in the frontal cortex (hypofrontality) during the flow phenomena. Participants consisted of expert piano players that were assessed for changes in alpha activity in the frontal cortex during a flow and non-flow condition. Results from the paired samples paired t-test conducted revealed there were statistically significant differences in alpha power in the experimental conditions (DV) versus the control conditions (IV; M = 93, SD = 105, N = 14), t(13) = 3.29, p = .006. These results supported the main hypothesis that there is increased alpha power in the frontal cortex during flow states. This finding provides the first empirically validated biomarker for a flow. These results will assist future research to understand flow experiences as a conceptually unambiguous variable.

The doctoral thesis deals with a novel methodology of looking and processing electroencephalographic (EEG) data. The first part deals with real-time brain stimulation in the form of a sonified neurofeedback therapy derived from a clinically comparable portable, 4-channel EEG system. The therapy aims to provide an effective management for symptoms of the Autism Spectrum Disorder (ASD). ASD is characterized with a high level of delta electroencephalographic waveform levels, while alpha and beta prove to be present at lower levels especially in the frontal-temporal regions. The treatment aims at lowering delta waves and promoting alpha and beta waveforms. The second part of the thesis focuses on using EEG data in the prediction of epileptic seizures. With the help of custom built algorithms and neural networks, an effective prediction of an epileptic seizure can be achieved.

L'elettroencefalogramma (EEG) fornisce registrazioni non-invasive delle attività e delle funzioni cerebrali sotto forma di serie temporali, a loro volta caratterizzate da una risoluzione temporale e spaziale (dipendente dai sensori), e da bande di frequenza specifiche per alcuni tipi di condizioni cerebrali. Tuttavia, i segnali EEG risultanti sono non-stazionari, cambiano nel tempo e sono eterogenei, essendo prodotti da differenti soggetti e venendo influenzati da specifici paradigmi sperimentali, condizioni ambientali e dispositivi. Inoltre, questi segnali sono facilmente soggetti a rumore e possono venire acquisiti per un tempo limitato, fornendo un numero ristretto di condizioni cerebrali sulle quali poter lavorare. Pertanto, in questa tesi viene proposta una pipeline flessibile per l'elaborazione e la gestione dei segnali EEG, affinchè possano essere più facilmente comprensibili e quindi più facilmente sfruttabili in diversi tipi di applicazioni. Inoltre, la pipeline flessibile proposta è divisa in quattro moduli riguardanti la pre-elaborazione del segnale, la sua normalizzazione, l'estrazione e la gestione di feature e la classificazione dei dati EEG. La pre-elaborazione del segnale EEG sfrutta la multivariate empirical mode decomposition (MEMD) per scomporre il segnale nelle sue modalità oscillatorie, chiamate intrinsic mode function (IMF), ed usa un criterio basato sull'entropia per selezionare le IMF più relevanti. Queste IMF dovrebbero mantenere le naturali dinamiche cerebrali e rimuovere componenti non-informative. Le risultati IMF rilevanti sono in seguito sfruttate per sostituire il segnale o aumentare la numerosità dei dati. Nonostante MEMD sia adatto alla non-stazionarietà del segnale EEG, ulteriori passi computazionali dovrebbero essere svolti per mitigare la caratteristica eterogeneità di questi dati. Pertanto, un passo di normalizzazione viene introdotto per ottenere dati comparabili per uno stesso soggetto o più soggetti e tra differenti condizioni sperimentali, quindi permettendo di estrarre feature nel dominio temporale, frequenziale e tempo-frequenziale per meglio caratterizzare il segnale EEG. Nonostante l'uso di un insieme di feature differenti fornisca la possibilità di trovare nuovi pattern nei dati, può altresì presentare alcune ridondanze ed incrementare il rischio di incorrere nella curse of dimensionality o nell'overfitting durante la classificazione. Pertanto, viene proposta una selezione delle feature basata sugli algoritmi evolutivi con un approccio completamente guidato dai dati. Inoltre, viene proposto l'utilizzo di modelli di apprendimento non o supervisionati e di nuovi criteri di stop per un algoritmo genetico modificato. Oltretutto, l'uso di diversi modelli di apprendimento automatico può influenzare il riconoscimento di differenti condizioni cerebrali. L'introduzione di modelli di deep learning potrebbe fornire una strategia in grado di apprendere informazioni direttamente dai dati disponibili, senza ulteriori elaborazioni. Fornendo una formulazione dell'input appropriata, le informazioni temporali, frequenziali e spaziali caratterizzanti i dati EEG potrebbero essere mantenute, evitando l'introduzione di architetture troppo complesse. Pertato, l'utilizzo di differenti processi ed approcci di elaborazione potrebbe fornire strategie più generiche o più legate a specifici esperimenti per gestire il segnale EEG, mantenendone le sue naturali caratteristiche.
The electroencephalogram (EEG) provides the non-invasive recording of brain activities and functions as time-series, characterized by a temporal and spatial (sensor-dependent) resolution, and by brain condition-bounded frequency bands. Moreover, it presents some cost-effective device solutions. However, the resulting EEG signals are non-stationary, time-varying, and heterogeneous, being recorded from different subjects and being influenced by specific experimental paradigms, environmental conditions, and devices. Moreover, they are easily affected by noise and they can be recorded for a limited time, thus they provide a restricted number of brain conditions to work with. Therefore, in this thesis a flexible pipeline for signal processing and management is proposed to have a better understanding of the EEG signals and exploit them for a variety of applications. Moreover, the proposed flexible pipeline is divided in 4 modules concerning signal pre-processing, normalization, feature computation and management, and EEG data classification. The EEG signal pre-processing exploits the multivariate empirical mode decomposition (MEMD) to decompose the signal in oscillatory modes, called intrinsic mode functions (IMFs), and uses an entropy criterion to select the most relevant IMFs that should maintain the natural brain dynamics, while discarding uninformative components. The resulting relevant IMFs are then exploited for signal substitution and data augmentation. Even though MEMD is adapt to the EEG signal non-stationarity, further processing steps should be undertaken to mitigate these data heterogeneity. Therefore, a normalization step is introduced to obtain comparable data inter- and intra-subject and between different experimental conditions, allowing the extraction of general features in the time, frequency, and time-frequency domain for EEG signal characterization. Even though the use of a variety of feature types may provide new data patterns, they may also present some redundancies and increase the risk of incurring in classification problems like curse of dimensionality and overfitting. Therefore, a feature selection based on evolutionary algorithms is proposed to have a completely data-driven approach, exploiting both supervised and unsupervised learning models, and suggesting new stopping criteria for a modified genetic algorithm implementation. Moreover, the use of different learning models may affect the discrimination of different brain conditions. The introduction of deep learning models may provide a strategy to learn directly from the available data. By suggesting a proper input formulation it could be possible to maintain the EEG data time, frequency, and spatial information, while avoiding too complex architectures. Therefore, using different processing steps and approaches may provide general or experimental specific strategies to manage the EEG signal, while maintaining its natural characteristics.

Orientador: Prof. Dr. André Mascioli Cravo
Dissertação (mestrado) - Universidade Federal do ABC, Programa de Pós-Graduação em Neurociência e Cognição, 2016.
We constantly learn and update our predictions about when events we cause will occur. This flexibility is important to program motor actions and to estimate when errors have been made. However, the mechanisms that govern learning and updating in temporal domain are largely unknown. In order to clarify these mechanisms we had three mains objectives: 1. To describe how we learn a new temporal relation between two events and how expectation is updated based on new information; 2. To describe the neural correlates underlying temporal learning and temporal updating; 3. To investigate temporal learning in two different sensory modalities: vision and audition, in order to verify whether such processes occur independently of sensory modality. In order to achieve the objectives, we developed two different experiments with electroencephalography recordings. In the first experiment, we aimed to answer the first two objectives by developing a behavioral task in which participants had to monitor whether a temporal error had been made. Results evidenced a rapid temporal adjustment by the participants to a new temporal relation. Temporal errors evoked electrophysiological markers classically related to error coding as frontal theta oscillations and feedback-related negativity. Delta phase was modulated by behavioral adjustments, suggesting its importance in temporal prediction updating. In conclusion, low frequency oscillations appear to be modulated in error coding and temporal learning. The second experiment investigated temporal learning in two different sensory modalities. Results indicated that time perception is biased differently depending on temporal marker sensory modality. Besides, we found that intertrial phase coherence of theta oscillations was modulated by expectation on both sensory conditions. However, such result occurs on central electrodes analysis, but not on sensory electrodes analysis, indicating a supramodal mechanism of temporal prediction.

This is a study of quantitative electroencephalographic (QEEG) comodulation analysis, which is used to assist in identifying regional brain patterns associated with Chronic Fatigue Syndrome (CFS) compared to an EEG normative database. Further, this study investigates EEG patterns in depression which is found to be a highly comorbid condition to CFS. The QEEG comodulation analysis examines spatial-temporal cross-correlation of spectral estimates in the individual resting dominant frequency band. A pattern shown by Sterman and Kaiser (2001) and referred to as the Anterior Posterior Dissociation (APD) discloses a significant reduction in shared functional modulation between frontal and centro-parietal areas of the cortex. Conversely, depressed patients have not shown this pattern of activity but have disclosed a pattern of frontal Hypercomodulation localized to bilateral pre-frontal and frontal cortex. This research investigates these comodulation patterns to determine whether they exist reliably in these populations of interest and whether a clear distinction between two highly comorbid conditions can be made using this metric. Sixteen CFS sufferers and 16 depressed participants, diagnosed by physicians and a psychiatrist respectively were involved in QEEG data collection procedures. Nineteen-channel cap recordings were collected in five conditions: eyes-closed, eyes open, reading task-one, math computations task-two, and a second eyes-closed baseline. Five of the 16 CFS patients showed a clear Anterior Posterior Dissociation pattern for the eyes-closed resting dominant frequency. However, 11 participants did not show this pattern of dysregulation. Examination of the mean 8-12 Hz band spectral magnitudes across three cortical regions (frontal, central and parietal) indicated a trend of higher overall alpha levels in the parietal region in CFS patients who showed the APD pattern compared to those who did not show this pattern. All participants who showed the APD pattern were free of medication, while the majority of those absent of this pattern were using antidepressant medications. For the depressed group, all of which were medication free, 100 % of the depressed group showed a frontal Hypercomodulation pattern. Furthermore, examination of the mean 8-12 Hz band spectral magnitudes across three cortical regions disclosed a trend of high frontal alpha and a left/right asymmetry of greater voltages in the left frontal cortex. Although these samples are small, it is suggested that this method of evaluating the disorder of CFS holds promise. The fact that this pattern is not consistently represented in the CFS sample could be explained by the possibility of subtypes of CFS, or perhaps comorbid conditions. Further, the use of antidepressant medications may mask the pattern by altering the temporal characteristics of the EEG. This study, however, was able to demonstrate that the QEEG was able to parse out the regional cerebral brain differences between CFS and depressed group.

Lorsque l'on enregistre l'activité cérébrale en électroencéphalographie (EEG) de surface, le signal d'intérêt est fréquemment bruité par des activités différentes provenant de différentes sources de bruit telles que l'activité musculaire. Le débruitage de l'EEG est donc une étape de pré-traitement important dans certaines applications, telles que la localisation de source. Dans cette thèse, nous proposons six méthodes permettant la suppression du bruit de signaux EEG dans le cas particulier des activités enregistrées chez les patients épileptiques soit en période intercritique (pointes) soit en période critique (décharges). Les deux premières méthodes, qui sont fondées sur la décomposition généralisée en valeurs propres (GEVD) et sur le débruitage par séparation de sources (DSS), sont utilisées pour débruiter des signaux EEG épileptiques intercritiques. Pour extraire l'information a priori requise par GEVD et DSS, nous proposons une série d'étapes de prétraitement, comprenant la détection de pointes, l'extraction du support des pointes et le regroupement des pointes impliquées dans chaque source d'intérêt. Deux autres méthodes, appelées Temps Fréquence (TF) -GEVD et TF-DSS, sont également proposées afin de débruiter les signaux EEG critiques. Dans ce cas on extrait la signature temps-fréquence de la décharge critique par la méthode d'analyse de corrélation canonique. Nous proposons également une méthode d'Analyse en Composantes Indépendantes (ICA), appelé JDICA, basée sur une stratégie d'optimisation de type Jacobi. De plus, nous proposons un nouvel algorithme direct de décomposition canonique polyadique (CP), appelé SSD-CP, pour calculer la décomposition CP de tableaux à valeurs complexes. L'algorithme proposé est basé sur la décomposition de Schur simultanée (SSD) de matrices particulières dérivées du tableau à traiter. Nous proposons également un nouvel algorithme pour calculer la SSD de plusieurs matrices à valeurs complexes. Les deux derniers algorithmes sont utilisés pour débruiter des données intercritiques et critiques. Nous évaluons la performance des méthodes proposées pour débruiter les signaux EEG (simulés ou réels) présentant des activités intercritiques et critiques épileptiques bruitées par des artéfacts musculaires. Dans le cas des données simulées, l'efficacité de chacune de ces méthodes est évaluée d'une part en calculant l'erreur quadratique moyenne normalisée entre les signaux originaux et débruités, et d'autre part en comparant les résultats de localisation de sources, obtenus à partir des signaux non bruités, bruités, et débruités. Pour les données intercritiques et critiques, nous présentons également quelques exemples sur données réelles enregistrées chez des patients souffrant d'épilepsie partielle
In the process of recording electrical activity of the brain, the signal of interest is usually contaminated with different activities arising from various sources of noise and artifact such as muscle activity. This renders denoising as an important preprocessing stage in some ElectroEncephaloGraphy (EEG) applications such as source localization. In this thesis, we propose six methods for noise cancelation of epileptic signals. The first two methods, which are based on Generalized EigenValue Decomposition (GEVD) and Denoising Source Separation (DSS) frameworks, are used to denoise interictal data. To extract a priori information required by GEVD and DSS, we propose a series of preprocessing stages including spike peak detection, extraction of exact time support of spikes and clustering of spikes involved in each source of interest. Two other methods, called Time Frequency (TF)-GEVD and TF-DSS, are also proposed in order to denoise ictal EEG signals for which the time-frequency signature is extracted using the Canonical Correlation Analysis method. We also propose a deflationary Independent Component Analysis (ICA) method, called JDICA, that is based on Jacobi-like iterations. Moreover, we propose a new direct algorithm, called SSD-CP, to compute the Canonical Polyadic (CP) decomposition of complex-valued multi-way arrays. The proposed algorithm is based on the Simultaneous Schur Decomposition (SSD) of particular matrices derived from the array to process. We also propose a new Jacobi-like algorithm to calculate the SSD of several complex-valued matrices. The last two algorithms are used to denoise both interictal and ictal data. We evaluate the performance of the proposed methods to denoise both simulated and real epileptic EEG data with interictal or ictal activity contaminated with muscular activity. In the case of simulated data, the effectiveness of the proposed algorithms is evaluated in terms of Relative Root Mean Square Error between the original noise-free signals and the denoised ones, number of required ops and the location of the original and denoised epileptic sources. For both interictal and ictal data, we present some examples on real data recorded in patients with a drug-resistant partial epilepsy

The purpose of this study was to explore the use of EEG biofeedback in a school setting to assist students who had attentional challenges. The equipment for implementing biofeedback was relatively inexpensive and was easily integrated into the school setting. Twenty students ranging in age from 7 to 17 were recruited for this study. Data was used from 14 subjects, 12 males (2 Hispanic, 1 African American, and 10 Caucasian) and 2 females (1 Hispanic, 1 Caucasian.) The subject pool was reduced due to non-compliance or the students. moving from the school district. Significant effect size was obtained in the treatment group in areas pertaining to visual perception and motor coordination. However, significant effect sizes in other areas were obtained when the control group scores worsened. The inclusion of student subjects who, perhaps, did not meet stringent criterion of attention deficit may have skewed the results. The small number of students in the study may have hindered accurate measures of statistical significance. Conversely, the information obtained from this study may offer insight to school districts in providing their students an alternate/adjunct to psychopharmacological medication and a non- invasive method of helping students with psycho-social challenges.

Dissertação apresentada na Faculdade de Ciências e Tecnologiea da Universidade Nova de Lisboa, para obtenção do Grau de Mestre em Engenharia Biomédica
The present study aims at examining the correlation among different brain areas while the subjects performed an arithmetical task, and how these differ from the mental relations in the same subjects during a resting state. In order to this, both linear and nonlinear methods were used, i.e., both algorithms capable of detecting linear relations and algorithms capable of detecting correlations without assuming any type of parametric relationship between the signals were implemented. The first algorithm that was implemented was the cross-correlation function, which gives an estimate of how much two signals are linearly correlated, and estimates the delay between them, thus permitting to make inferences on causality. Furthermore, this algorithm was validated using the statistic method called surrogation, in order to test for the applicability of the algorithm on the signals that were to be processed. The next part of the study consisted on implementing two analogous algorithms, the coefficient of determination and the nonlinear regression coefficient. These coefficients both measure the fraction of reduction of variance that can be obtained by estimating the relationship between two signals according to a fitted line, the difference being that the former assumes a linear relation between both sets of samples and the latter doesn‟t previously assume any type of relationship between the signals. The main differences in correlation that were observed between the state of mental rest and between the arithmetic task performance were that in the former more brain sites were correlated, whereas during the task this synchrony was mainly verified between frontal and parietal areas, showing a decrease in the other locations. Furthermore, the estimates provided by the linear and nonlinear algorithms were very similar, suggesting that in this case the relationships among different neural networks were mainly linear, and thus validating the application of linear methods in this type of analysis in particular cases. Regarding the estimation of delays between signals and inferences on causality, no conclusive results were attained.

The use of nonlinear signal analysis measures to characterize electroencephalographic (EEG) recordings can be key for a better understanding of the underlying brain dynamics. In neurological disorders such as epilepsy, these dynamics are altered as result of a disturbed coordination between neuronal populations. The aim of this thesis is to characterize the seizure-free interval of EEG recordings from epilepsy patients by means of nonlinear signal analysis techniques to investigate whether this type of analysis can contribute to the localization of the seizure onset zone, the brain region from which initial seizure discharges can be recorded. For this purpose, we used a surrogate-corrected nonlinear predictability score and a surrogatecorrected nonlinear interdependence measure to analyze all-night EEG recordings from epilepsy patients implanted with hybrid depth electrodes equipped with macro contacts and micro wires. Our results show that the combined analysis of macro and micro EEG recordings may help to further increase the degree to which quantitative EEG analysis can contribute to the diagnostics in epilepsy patients.
El uso de medidas de análisis no lineales de señales para caracterizar registros electroencefalográficos (EEG) puede ser clave para una mejor comprensión de las dinámicas cerebrales subyacentes. En trastornos neurológicos como la epilepsia, estas dinámicas están alteradas a consecuencia de una coordinación perturbada entrepoblaciones neuronales. El objetivo de esta tesis es caracterizarel intervalo de registros de EEG libre de crisis epilépticas de pacientes con epilepsia mediante técnicas de análisis no lineales de señales para investigar si este tipo de análisis puede contribuir ala localización del SOZ (en inglés, Seizure onset zone), la región del cerebro donde se pueden registrar las descargas iniciales de las crisis epilépticas. Con este propósito, utilizamos una puntuación de predictibilidad no lineal corregida por sustitutos y una medida de interdependencia no lineal corregida por sustitutos para analizar registros EEG de pacientes con epilepsia grabados durante noches completas implantados con electrodos híbridos equipados con macro- y microcontactos. Nuestros resultados demuestran que el análisis combinado de macro- y micro-registros de EEG puede ayudar a aumentar el grado en el que el análisis cuantitativo de EEG puede contribuir al diagnóstico de pacientes con epilepsia.
L’ús de mesures d’anàlisi de senyals no lineals per la caracterització de registres encefalogràfics (EEG) pot ser clau per una millor comprensió de les dinàmiques cerebrals subjacents. En trastorns neurològics com l’epilèpsia, aquestes dinàmiques estan alterades a conseqüència d’una coordinació pertorbada entre poblacions neuronals. L’objectiu d’aquesta tesi doctoral és caracteritzar l’interval de registres EEG lliures de crisis epilèptiques en pacients amb epilèpsia mitjançant tècniques d’anàlisi de senyals no lineals, per tal d’investigar si aquest tipus d’anàlisi pot contribuir a la localització de la SOZ (en anglès, Seizure onset zone), la regió del cervell on es poden registrar les primeres descàrregues de la crisi. Amb aquesta finalitat, utilitzem una puntuació de previsibilitat no lineal corregida mitjançant substituts i una mesura d’interdependència no lineal corregida per substituts per analitzar registres EEG de pacients amb epilèpsia. Aquests han sigut enregistrats durant nits completes amb elèctrodes híbrids equipats amb macro- i microcontactes. Els resultats obtinguts demostren que l’anàlisi combinat de macro- i microregistres en l’EEG pot ajudar a augmentar el grau de contribució de l’anàlisi quantitatiu de l’EEG dins el diagnòstic de pacients amb epilèpsia.

Previous research suggests that defensiveness involves decreased perceptual sensitivity for intense or threatening stimuli, which is reflected in various perceptual and electroencephalographic changes. Expanding on this theme, Kline, Schwartz, and Dikman (1992) found that high defensive subjects evidence decreased perceptual acuity for a putative human pheromone, androstenone (AND), proposing an olfactory "perceptual defense" effect. Thus, the present study explored relationships between performance and electroencephalographic parameters in a visual "perceptual defense" paradigm, AND perception, and repressive-defensiveness. Except for in the spring, high-defensive subjects were in general less perceptually sensitive for AND. Excluding subjects run during the spring, detection accuracy for AND correlated negatively with identification thresholds for unpleasant words. AND perception tended to correlated positively with identification thresholds for sexual-taboo words. In general, highest identification thresholds obtained for neutral and sexual-taboo words, and lowest for pleasant and unpleasant words. Mean hit rates in a word detection task were 0.19, 0.17, 0.24, 0.44, 0.65, and 0.80 for six ascending durations (chance = 0.17). For the shortest three durations (≤ 50.1 msec), low defensive subjects had higher hit rates for neutral (NEU) versus unpleasant (UPLS), and sexual-taboo (SEX) words. In contrast, HD showed lowest hit rates for NEU and highest hit rates for SEX. At the longest three durations (≥ 66.8 msec), LD showed lowest hit rates for NEU and UPLS, and HD had lower hit rates for pleasant words (PLS) and SEX than did LD. Confidence increased with duration, but no significant defensiveness or word category differences emerged. HD showed less alpha (8-13 Hz) in response to 50.1 msec masked words at posterior leads for SEX, VPLS, and especially PLS relative to NEV, where LD showed essentially the opposite pattern. Alpha decreases for SEX correlated significantly with ≤ 50.1 msec hit rates for SEX at 02, and correlated with ≤ 50.1 hit rates for PLS and VPLS at 01. In response to 100.2 msec duration masked words, all subjects showed less alpha during SEX than during NEV, especially posteriorly, which was somewhat right lateralized for HD. The results suggest that defensiveness may involve unconscious supersensitivity to emotional content that facilitates conscious subsensitivity to emotional content.

This work proposes the quantitative analysis of electroencephalographic (EEG) signals, comparing normal subjects to patients in coma state under brain death protocol, by means of a software platform developed in MATLAB. This analysis involved peak frequencies, median frequencies, average, standard deviation and modular index. This last one relates to the bilaterality, i.e. how equidistant electrodes are on the same frequency. Results outlined that patients in coma present lower average median frequency (11.89 Hz), when compared to normal subjects (30.69 Hz). Moreover, with regard the analysis of bilaterality based on modular index, patients in coma also presented lower average standard deviation of the average median frequency (4.72 Hz) than the normal individuals (13 , 67 Hz), e.g., comatose patients presented higher bilaterality. Therefore, results suggests the possibility of applying these quantifications as biomarkers of brain death.
Este trabalho apresenta uma análise quantitativa dos sinais de exames de Eletroencefalografia comparando pacientes normais e pacientes em estado de coma, a partir de análises visuais das amplitudes pelo Neurologista. O estudo comparou dois grupos distintos, sendo um grupo formado por pacientes em estado de coma sob protocolo de morte encefálica e outro composto por pacientes normais. Foi analisado e comparado dados gerados por um software desenvolvido na plataforma MATLAB. Estes dados foram Frequências de Pico, Frequências Medianas, Média, Desvio Padrão e Índice Modular, que se relaciona à bilateralidade, ou seja, o quanto os pontos equidistantes no couro cabeludo representado pela posição dos eletrodos estão em mesma frequência. Pode-se comprovar que pacientes em estado de coma apresentam médias de frequências medianas (11,89 Hz) menores tanto em relação aos canais quanto às épocas, quando comparado aos pacientes normais (30,69 Hz). Além disso, com relação à análise de bilateralidade a partir do cálculo do Índice Modular derivado da média da frequência mediana e da media do Desvio padrão, os pacientes em estado de coma (4.72 Hz) também apresentaram média menor do que os pacientes normais (13,67 Hz), ou seja, os pacientes em coma apresentaram maior bilateralidade. Portanto, pode-se cogitar a aplicação desta ferramenta para auxiliar no diagnóstico de morte encefálica, sendo assim uma ferramenta a mais de auxílio aos profissionais de saúde.
Mestre em Ciências

Une Interface Cerveau-Ordinateur (ICO) est un système de communication qui permet à ses utilisateurs d'envoyer des commandes à un ordinateur via leur activité cérébrale, cette activité étant mesurée, généralement par ÉlectroEncéphaloGraphie (EEG), et traitée par le système. Dans la première partie de cette thèse, dédiée au traitement et à la classification des signaux EEG, nous avons cherché à concevoir des ICOs interprétables et plus efficaces. Pour ce faire, nous avons tout d'abord proposé FuRIA, un algorithme d'extraction de caractéris- tiques utilisant les solutions inverses. Nous avons également proposé et étudié l'utilisation des Systèmes d'Inférences Flous (SIF) pour la classification. Nos évaluations ont montré que FuRIA et les SIF pouvaient obtenir de très bonnes performances de classification. De plus, nous avons proposé une méthode utilisant ces deux algorithmes afin de concevoir une ICO complétement interprétable. Enfin, nous avons proposé de considérer la conception d'ICOs asynchrones comme un problème de rejet de motifs. Notre étude a introduit de nouvelles techniques et a permis d'identifier les classifieurs et les techniques de rejet les plus appropriés pour ce problème. Dans la deuxième partie de cette thèse, nous avons cherché à concevoir des applications de Réalité Virtuelle (RV) controlées par une ICO. Nous avons tout d'abord étudié les performances et les préférences de participants qui interagissaient avec une application ludique de RV à l'aide d'une ICO asynchrone. Nos résultats ont mis en évidence le besoin d'utiliser des ICO adaptées à l'utilisateur ainsi que l'importance du retour visuel. Enfin, nous avons développé une application de RV permettant à un utilisateur d'explorer un musée virtuel par la pensée. Dans ce but, nous avons conçu une ICO asynchrone et proposé une nouvelle technique d'interaction permettant à l'utilisateur d'envoyer des commandes de haut niveau. Une première évaluation semble montrer que l'utilisateur peut explorer le musée plus rapidement avec cette technique qu'avec les techniques actuelles.

In my PhD research project, I wanted to investigate infants’ representation of the peripersonal space, which is the portion of environment between the self and the others. In the last three decades research provided evidence on newborns’ and infants’ perception of their own bodies and of other individuals, whereas not many studies investigated infants’ perception of the portion of space where they can interact with both others and objects, namely the peripersonal space. Considering the importance of the peripersonal space, especially in light of its defensive and interactive functions, I decided to investigate the development of its representation focusing on two aspects. On one side, I wanted to study how newborns and infants processed the space around them, if they differentiated between near and far space, possibly perceiving and integrating depth cues across sensory modalities and when and how they started to respond to different movements occurring in the space surrounding their bodies. On the other side, I was interested in understanding whether already at birth the peripersonal space could be considered as a delimited portion of space with special characteristics and, relatedly, if its boundaries could be determined. In order to respond to my first question, I investigated newborns’ and infants’ looking behaviour in response to visual and audio-visual stimuli depicting different trajectories taking place in the space immediately surrounding their body. Taken together, the results of these studies demonstrated that humans show, since the earliest stages of their development, a rudimentary processing of the space surrounding them. Newborns seemed, in fact, to already differentiate the space around them, through an efficient discrimination of different moving trajectories and a visual preference for those directed towards their own body, possibly due to their higher adaptive relevance. They also seemed to integrate multimodal, audio-visual information about stimuli moving in the near space, showing a facilitated processing of congruent audio-visual approaching stimuli. Furthermore, the results of these studies could help understand the development of the integration of multimodal stimuli with an adaptive valence during infancy. When newborns’ and infants were presented with unimodal, visual stimuli, they all directed their visual preferences to the stimuli moving towards their bodies. Conversely, their pattern of looking times was more complex when they were presented with congruent and incongruent audiovisual stimuli. Right after birth infants showed a spontaneous visual preference for congruent audio-visual stimuli, which was challenged by a similarly strong visual preference for adaptively important visual stimuli moving towards their bodies. The looking behaviours of 5-month-old infants, instead, seemed to be driven only by a spontaneous preference for multimodal congruent stimuli, i.e. depicting motion along the same trajectory, irrespective of the adaptive value of the information conveyed by either of the two sensory components of the stimulus. Nine-month-old infants, finally, seemed to flexibly integrate multisensory integration principles with the necessity of directing their attention to ethologically salient stimuli, as shown by the fact that their visual preference for unexpected, incongruent audio-visual stimuli was challenged by the simultaneous presence of adaptively relevant stimuli. Similarly to what happened with newborns, presenting 9-month-old infants with the two categories of preferred stimuli simultaneously led to the absence of a visual preference. Within my project I also investigated the electroencephalographic correlates of the processing of unimodal, visual and auditory, stimuli depicting different trajectories in a sample of 5-month-old infants. The results seemed to provide evidence in support of the role of the primary sensory cortices in the processing of crossmodal stimuli. Furthermore, they seemed to support the possibility that infants’ brain could allocate, already during the earliest stages of processing, different amounts of attention to stimuli with different adaptive valence. Two further studies addressed my second question, namely whether already at birth the peripersonal space could be considered as a delimited portion of space with special characteristics and if its boundaries could be determined. In these studies I measured newborns’ saccadic reaction times (RTs) to tactile stimuli presented simultaneously to a sound perceived at different distances from their body. The results showed that newborns’ RTs were modulated by the perceived position of the sound and that their modulation was very similar to that shown by adults, suggesting that the boundary of newborns’ peripersonal space could be identified in the perceived sound position in whose correspondence the drop of RTs happened. This suggested that at birth the space immediately surrounding the body seems to be already invested of a special salience and characterised by a more efficient integration of multimodal stimuli. As a consequence, it might be considered as a rudimentary representation of the peripersonal space, possibly serving, as a working space representation, early interactions between newly born humans and their environment. Overall, these findings provide a first understanding of how humans start to process the space surrounding them, which, importantly, is the space linking them with others and the space where their first interactions will take place.
Il mio progetto di Dottorato è nato con l’obiettivo di investigare la rappresentazione dello spazio peripersonale, cioè la porzione di spazio tra noi stessi e gli altri, durante l’infanzia. Nel corso degli ultimi trent’anni diversi studi hanno dimostrato la capacità di neonati ed infanti di percepire il proprio corpo, così come gli altri individui. Al contrario, non molti studi si sono interessati alla loro percezione della porzione di spazio dove essi possono interagire con gli oggetti e con gli altri, definita “spazio peripersonale”. Vista l’importanza dello spazio peripersonale, specialmente alla luce delle sue funzioni difensiva da un lato ed interattiva dall’altro, ho deciso di investigarne la rappresentazione concentrandomi su due aspetti. Da un lato, ho studiato come i neonati e gli infanti elaborino lo spazio intorno a loro, se differenzino tra spazio vicino e lontano, se percepiscano ed integrino gli indicatori di profondità provenienti da diverse modalità sensoriali, nonché come e quando inizino a rispondere ai diversi movimenti che hanno luogo nello spazio che circonda il loro corpo. Dall’altro lato, ero interessata a capire se già alla nascita lo spazio peripersonale potesse essere considerato come una porzione delimitata di spazio, contraddistinta da caratteristiche specifiche, e se i suoi confini potessero già essere stimati. Per rispondere alla mia prima domanda, ho analizzato il comportamento visivo di neonati ed infanti in risposta a stimoli visivi e audio-visivi raffiguranti diverse traiettorie che avevano luogo nello spazio immediatamente circostante il corpo. I risultati di questi studi, complessivamente, dimostrano che gli esseri umani mostrano, fin dai primi stadi dello sviluppo, una rudimentale capacità di elaborare lo spazio che circonda il loro corpo. I neonati sembrano, infatti, poter già differenziare lo spazio che li circonda, attraverso un’efficiente discriminazione di diverse traiettorie di movimento ed una preferenza visiva per quelle dirette verso il loro corpo, forse a causa della loro maggiore importanza adattiva. Inoltre, essi sembrano capaci di integrare informazioni multimodali rispetto al movimento di stimoli nello spazio circostante, mostrando un’elaborazione facilitata di stimoli in avvicinamento segnalati, al tempo stesso, da componenti visive ed uditive congruenti. Inoltre, i risultati di questi studi hanno permesso di aumentare la comprensione dello sviluppo della capacità di integrare stimoli multimodali caratterizzati da un’alta valenza adattiva durante l’infanzia. Quando ai neonati ed agli infanti sono stati presentati stimoli visivi (unimodali), essi hanno sempre rivolto la loro preferenza visiva agli stimoli che mostravano un movimento diretto verso il loro corpo. Diversamente, il loro comportamento visivo si è dimostrato più complesso quando sono stati presentati loro stimoli audiovisivi congruenti o incongruenti. Subito dopo la nascita, i neonati hanno mostrato una spontanea preferenza visiva per gli stimoli multimodali caratterizzati da una direzione di movimento congruente, a sua volta contrastata da un’altrettanta forte preferenza visiva per quegli stimoli che, muovendosi verso il loro corpo, erano caratterizzati da una grande salienza adattiva. Il comportamento visivo degli infanti di cinque mesi di età, invece, è sembrato essere guidato solamente da una spontanea preferenza per gli stimoli multimodali congruenti, cioè quelli che rappresentavano movimenti lungo la stessa traiettoria, indipendentemente dal valore adattivo delle informazioni trasmesse da ognuna delle due componenti sensoriali degli stimoli. Gli infanti di nove mesi di età, infine, sono sembrati capaci di integrare con flessibilità i principi dell’integrazione multisensoriale con la necessità di dirigere la loro attenzione verso gli stimoli etologicamente rilevanti, come dimostrato dal fatto che la loro preferenza visiva per gli stimoli audiovisivi incongruenti ed inaspettati è stata contrastata dalla simultanea presenza di stimoli importanti a livello adattivo. Come successo per i neonati, quando agli infanti di questa età venivano presentati contemporaneamente stimoli facenti parte delle due categorie preferite, essi non hanno mostrato alcuna preferenza visiva. All’interno del mio progetto ho anche investigato i correlati elettroencefalografici dell’elaborazione di stimoli unimodali, visivi ed uditivi, raffiguranti diverse traiettorie in un campione di infanti di cinque mesi di età. I risultati sembrano supportare il ruolo delle cortecce sensoriali primarie nell’elaborazione di stimoli provenienti da diverse modalità sensoriali, così come la possibilità che il cervello degli infanti possa assegnare diversi quantitativi di attenzione a stimoli di diversa importanza adattiva, già durante i primissimi stadi dell’elaborazione. Due ulteriori studi hanno indirizzato la mia seconda domanda, ovvero se già alla nascita lo spazio peripersonale possa essere considerato quale una porzione delimitata di spazio contraddistinta da particolari caratteristiche e se i suoi confini possano essere determinati. In questi studi ho misurato i tempi di reazione saccadici ad una stimolazione tattile accompagnata da un suono percepito a diverse distanze dal corpo. I risultati hanno mostrato che i tempi di reazione dei neonati sono stati modulati dalla distanza percepita del suono dal corpo. Inoltre, la modulazione dei tempi di reazione nei neonati è risultata molto simile a quella mostrata dagli adulti, suggerendo che i confini dello spazio peripersonale dei neonati possono essere identificati nella posizione in corrispondenza della quale i tempi di reazione sono drasticamente diminuiti. Questo dato suggerisce che alla nascita lo spazio immediatamente circostante il corpo sembra possedere già un’importanza particolare e sembra essere caratterizzato da una più efficace integrazione di stimoli multimodali. Di conseguenza, potrebbe essere considerato come una rudimentale rappresentazione dello spazio peripersonale, che può essere considerata al servizio delle interazioni precoci tra i neonati ed il loro ambiente. Complessivamente, questi risultati forniscono una prima comprensione di come gli esseri umani inizino a processare lo spazio che li circonda, cioè è lo spazio che li unisce agli altri, nonché lo spazio nel quale le loro prime interazioni avranno luogo.

Objective: Brain-computer interface (BCI) technologies have been the subject of study for the past decades to help restore functions for people with severe motor disabilities and to improve their quality of life. BCI research can be generally categorized by control signals (invasive/non-invasive) or applications (e.g. neuroprosthetics/brain-actuated wheelchairs), and efforts have been devoted to better understand the characteristics and possible uses of brain signals. The purpose of this research is to explore the feasibility of a non-invasive BCI system with the combination of unique sensorimotor-rhythm (SMR) features. Specifically, a 2D virtual wheelchair control BCI is implemented to extend the application of previously designed 2D cursor control BCI, and the feasibility of the prototype is tested in electroencephalography (EEG) experiments; guidance on enhancing system performance is provided by a simulation incorporating intelligent control approaches under different EEG decoding accuracies; pattern recognition methods are explored to provide optimized classification results; and a hybrid BCI system is built to enhance the usability of the wheelchair BCI system. Methods: In the virtual wheelchair control study, a creative and user friendly control strategy was proposed, and a paradigm was designed in Matlab, providing a virtual environment for control experiments; five subjects performed physical/imagined left/right hand movements or non-control tasks to control the virtual wheelchair to move forward, turn left/right or stop; 2-step classification methods were employed and the performance was evaluated by hit rate and control time. Feature analysis and time-frequency analysis were conducted to examine the spatial, temporal and frequency properties of the utilized SMR features, i.e. event-related desynchronization (ERD) and post-movement event-related synchronization (ERS). The simulation incorporated intelligent control methods, and evaluated navigation and positioning performance with/without obstacles under different EEG decoding accuracies, to better guide optimization. Classification methods were explored considering different feature sets, tuned classifier parameters and the simulation results, and a recommendation was provided to the proposed system. In the steady state visual evoked potential (SSVEP) system for hybrid BCI study, a paradigm was designed, and an electric circuit system was built to provide visual stimulus, involving SSVEP as another type of signal being used to drive the EEG BCI system. Experiments were conducted and classification methods were explored to evaluate the system performance. Results: ERD was observed on both hemispheres during hand's movement or motor imagery; ERS was observed on the contralateral hemisphere after movement or motor imagery stopped; five subjects participated in the continuous 2D virtual wheelchair control study and 4 of them hit the target with 100% hit rate in their best set with motor imagery. The simulation results indicated that the average hit rate with 10 obstacles can get above 95% for pass-door tests and above 70% for positioning tests, with EEG decoding accuracies of 70% for Non-Idle signals and 80% for idle signals. Classification methods showed that with properly tuned parameters, an average of about 70%-80% decoding accuracy for all the classifiers could be reached, which reached the requirements set by the simulation test. Initial test on the SSVEP BCI system exhibited high classification accuracy, which may extend the usability of the wheelchair system to a larger population when finally combined with ERD/ERS BCI system. Conclusion: This research investigated the feasibility of using both ERD and ERS associated with natural hand's motor imagery, aiming to implement practical BCI systems for the end users in the rehabilitation stage. The simulation with intelligent controls provided guides and requirements for EEG decoding accuracies, based on which pattern recognition methods were explored; properly selected features and adjusted parameters enabled the classifiers to exhibit optimal performance, suitable for the proposed system. Finally, to enlarge the population for which the wheelchair BCI system could benefit for, a SSVEP system for hybrid BCI was designed and tested. These systems provide a non-invasive, practical approach for BCI users in controlling assistive devices such as a virtual wheelchair, in terms of ease of use, adequate speed, and sufficient control accuracy.

The goal of this thesis work was to study the characteristics of the EEG signal and then, based on the insights gained from these studies, pursue an initial investigation into a processing method that would extract useful event related information from single trial EEG. The fundamental tool used to study the EEG signal characteristics was autoregressive modeling. Early investigations pointed to the need to employ robust techniques in both model parameter estimation and signal estimation applications. Pursuing robust techniques ultimately led to the development of a single trial processing method which was based on a simple neurological model that assumed an additive outlier nature of event related potentials to the ongoing EEG process. When event related potentials, such as motor related potentials, are generated by a unique additional process they are "added" into the ongoing process and hence, will appear as additive outlier content when considered from the point of view of the ongoing process. By modeling the EEG with AR models with robustly estimated (GM-estimates) parameters and by using those models in a robust signal estimator, a "cleaned" EEG signal is obtained. The outlier content, data that is extracted from the EEG during cleaning, is then processed to yield event related information. The EEG from four subjects formed the basis of the initial investigation into the viability of this single trial processing scheme. The EEG was collected under two conditions: an active task in which subjects performed a skilled thumb movement and an idle task in which subjects remained alert but did not carry out any motor activity. The outlier content was processed which provided single trial outlier waveforms. In the active case these waveforms possessed consistent features which were found to be related to events in the individual thumb movements. In the idle case the waveforms did not contain consistent features. Bayesian classification of active trials versus idle trials was carried out using a cost statistic resulting from the application of dynamic time warping to the outlier waveforms. Across the four subjects, when the decision boundary was set with the cost of misclassification equal, 93% of the active trials were classified correctly and 18% of the idle trials were incorrectly classified as active. When the cost of misclassifying an idle trial was set to be five times greater, 80% of the active trials were classified correctly and only 1.7% of the idle trials were incorrectly classified as active.
Applied Science, Faculty of
Electrical and Computer Engineering, Department of
Graduate

Traditionally research into how we perceive our external world focused on the unisensory approach, examining how information is processed by one sense at a time. This produced a vast literature of results revealing how our brains process information from the different senses, from fields such as psychophysics, animal electrophysiology, and neuroimaging. However, we know from our own experiences that we use more than one sense at a time to understand our external world. Therefore to fully understand perception, we must understand not only how the brain processes information from individual sensory modalities, but also how and when this information interacts and combines with information from other modalities. In short, we need to understand the phenomenon of multisensory perception. The work in this thesis describes three experiments aimed to provide new insights into this topic. Specifically, the three experiments presented here focused on examining when and where effects related to multisensory perception emerged in neural signals, and whether or not these effects could be related to behaviour in a time-resolved way and on a trial-by-trial basis. These experiments were carried out using a novel combination of psychophysics, high density electroencephalography (EEG), and advanced computational methods (linear discriminant analysis and mutual information analysis). Experiment 1 (Chapter 3) investigated how behavioural and neural signals are modulated by the reliability of sensory information. Previous work has shown that subjects will weight sensory cues in proportion to their relative reliabilities; high reliability cues are assigned a higher weight and have more influence on the final perceptual estimate, while low reliability cues are assigned a lower weight and have less influence. Despite this widespread finding, it remains unclear when neural correlates of sensory reliability emerge during a trial, and whether or not modulations in neural signals due to reliability relate to modulations in behavioural reweighting. To investigate these questions we used a combination of psychophysics, EEG-based neuroimaging, single-trial decoding, and regression modelling. Subjects performed an audio-visual rate discrimination task where the modality (auditory, visual, audio-visual), stimulus stream rate (8 to 14 Hz), visual reliability (high/low), and congruency in rate between audio-visual stimuli (± 2 Hz) were systematically manipulated. For the behavioural and EEG components (derived using linear discriminant analysis), a set of perceptual and neural weights were calculated for each time point. The behavioural results revealed that participants weighted sensory information based on reliability: as visual reliability decreased, auditory weighting increased. These modulations in perceptual weights emerged early after stimulus onset (48 ms). The EEG data revealed that neural correlates of sensory reliability and perceptual weighting were also evident in decoding signals, and that these occurred surprisingly early in the trial (84 ms). Finally, source localisation suggested that these correlates originated in early sensory (occipital/temporal) and parietal regions respectively. Overall, these results provide the first insights into the temporal dynamics underlying human cue weighting in the brain, and suggest that it is an early, dynamic, and distributed process in the brain. Experiment 2 (Chapter 4) expanded on this work by investigating how oscillatory power was modulated by the reliability of sensory information. To this end, we used a time-frequency approach to analyse the data collected for the work in Chapter 3. Our results showed that significant effects in the theta and alpha bands over fronto-central regions occurred during the same early time windows as a shift in perceptual weighting (100 ms and 250 ms respectively). Specifically, we found that theta power (4 - 6 Hz) was lower and alpha power (10 – 12 Hz) was higher in audio-visual conditions where visual reliability was low, relative to conditions where visual reliability was high. These results suggest that changes in oscillatory power may underlie reliability based cue weighting in the brain, and that these changes occur early during the sensory integration process. Finally, Experiment 3 (Chapter 5) moved away from examining reliability based cue weighting and focused on investigating cases where spatially and temporally incongruent auditory and visual cues interact to affect behaviour. Known collectively as “cross-modal associations”, past work has shown that observers have preferred and non-preferred stimuli pairings. For example, subjects will frequently pair high pitched tones with small objects and low pitched tones with large objects. However it is still unclear when and where these associations are reflected in neural signals, and whether they emerge at an early perceptual level or later decisional level. To investigate these questions we used a modified version of the implicit association test (IAT) to examine the modulation of behavioural and neural signals underlying an auditory pitch – visual size cross modal association. Congruency was manipulated by assigning two stimuli (one auditory and one visual) to each of the left or right response keys and changing this assignment across blocks to create congruent (left key: high tone – small circle, right key: low tone – large circle) and incongruent (left key: low tone – small circle, right key: high tone – large circle) pairings of stimuli. On each trial, subjects were presented with only one of the four stimuli (auditory high tone, auditory low tone, visual small circle, visual large circle), and asked to respond which was presented as quickly and accurately as possible. The key assumption with such a design is that subjects should respond faster when associated (i.e. congruent) stimuli are assigned to the same response key than when two non-associated stimuli are. In line with this, our behavioural results demonstrated that subjects responded faster on blocks where congruent pairings of stimuli were assigned to the response keys (high pitch-small circle and low pitch large circle), than blocks where incongruent pairings were. The EEG results demonstrated that information about auditory pitch and visual size could be extracted from neural signals using two approaches to single-trial analysis (linear discriminant analysis and mutual information analysis) early during the trial (50ms), with the strongest information contained over posterior and temporal electrodes for auditory trials, and posterior electrodes for visual trials. EEG components related to auditory pitch were significantly modulated by cross-modal congruency over temporal and frontal regions early in the trial (~100ms), while EEG components related to visual size were modulated later (~220ms) over frontal and temporal electrodes. For the auditory trials, these EEG components were significantly predictive of single trial reaction times, yet for the visual trials the components were not. As a result, the data support an early and short-latency origin of cross-modal associations, and suggest that these may originate in a bottom-up manner during early sensory processing rather than from high-level inference processes. Importantly, the findings were consistent across both analysis methods, suggesting these effects are robust. To summarise, the results across all three experiments showed that it is possible to extract meaningful, single-trial information from the EEG signal and relate it to behaviour on a time resolved basis. As a result, the work presented here steps beyond previous studies to provide new insights into the temporal dynamics of audio-visual perception in the brain.
All experiments, although employing different paradigms and investigating different processes, showed early neural correlates related to audio-visual perception emerging in neural signals across early sensory, parietal, and frontal regions. Together, these results provide support for the prevailing modern view that the entire cortex is essentially multisensory and that multisensory effects can emerge at all stages during the perceptual process.

Introduction Combined electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) studies enables to non invasively investigate human brain function and to find the direct correlation of these two important measures of brain activity. The combination of these technologies provides informations and details on the spatio-temporal aspects of human brain processing. fMRI has an excellent spatial resolution and allows the localization of brain regions in which there is a change in the level of neuronal activity during an experimental condition compared to a control condition. In contrast, EEG measures neuronal currents directly from the subject’s scalp with a high temporal resolution in the range of milliseconds. Combined recording wants to overcome the spatial limitations of EEG and the temporal limitations of fMRI, using their complementary features. For instance, combined EEG-fMRI technique can be used to identify the neural correlates of clinically or behaviourally important spontaneous EEG activity, such as interictal spikes, the alpha rhythm and sleep waves. The presurgical evaluation of patients with epilepsy is one of the areas where combining EEG and fMRI has considerable clinical relevance for localizing the brain regions generating interictal epileptiform activity. fMRI is mostly used in the study of sensory, motor and cognitive functions, where there is a difference between experimental condition and control condition. In the context of epilepsy, one can consider the control condition to occur when the EEG is at baseline and experimental condition to correspond to the presence of an epileptic discharge. The conventional analysis of EEG-fMRI data is based on the visual identification of the interictal epileptiform discharges (IEDs) on scalp EEG which are used in conjunction with a General Linear Model (GLM) approach to analyze fMRI data. A model is obtained by the convolution of the EEG events, which are represented as stick functions of unitary amplitude, with a model of the event-related fMRI response, represents by the haemodynamic response function (HRF); maps showing regions of significant IED-related change are obtained through voxel-wise fitting of the model and application of appropriate statistical thresholds. In this thesis we present an easy to use approach for combined EEG-fMRI analysis developed to improve the identification of the IEDs. The novel automatic method is based on Independent Component Analysis (ICA) and allows to detect IED activity in order to use it as a parametric modulator in fMRI analysis. The Novel Method Data quality is a crucial issue in multimodal functional imaging and data integration. Both fMRI and EEG data acquisition processes can severely affect the other’s performance through electromagnetic interactions, therefore the pre-processing is necessary for both EEG and fMRI data. While for fMRI data the pre-processing is generally standard, apart from the choice of spatial smoothing; the EEG pre-processing requires a complex and not one-way procedure to remove the artifacts. In literature different methods have been developed to remove gradient and pulse artifacts, considering both hardware and software solutions. The gradient EEG artifact removal method implemented in our EEG system acquisition did not give completely satisfactory results; so we decided to developed a novel method. Since the project regarding the gradient filter started together with the novel EEG-fMRI integration method and the analysis on patients with partial epilepsy are still in progress to avoid the introduction of a further variable in the validation of the method we decided to use the algorithms implemented in the SystemPlus software. After a pre-processing applied on EEG data and composed by a re-reference and filtering, a method based on ICA decomposition was applied. In the field of biomedical signal processing, Blind Source Separation (BSS) methods are generally used to separate multi-channel recordings into their constituent components; ICA is a subset of such techniques used to separate statistically independent components from a mixture of data. ICA decomposition of the data was performed using FastICA algorithm implemented in EEGLAB. The novel method consists in four fundamental steps: • Selection of components • Reconstruction of EEG signal • Selection of channel and FFT analysis • Construction of EEG regressor The crucial point is the selection of components. To select the components related to IED activity, we used a time-frequency representation obtained by using wavelet-based analysis. We computed the wavelet power for all the components in the epochs of interest and then, for each component, we selected from the frequency bins the one with the maximal power over total recording session. Finally power was averaged along time, obtaining one value for each component. Components that exceeded mean value ± standard deviation were chosen for further analysis. After the components of interest have been selected, they were back projected to obtain a new EEG signal (reconstructed EEG). A Fast Fourier Transform (FFT) analysis was applied on the time series of the selected channel (where the IED activity is clearly visible) for epochs acquired during each fMRI volume. Then the power time course created for all volumes was used to form the EEG regressor used in GLM analysis. Discussion The aim of the research project here described is the development of an innovative procedure for integrating neurophysiological and functional neuroimaging data. In fMRI processing the selection of the experimental paradigm as difference between task and rest conditions is of great importance, in fact the information related to the experimental events and to the rest condition are to be used as input in GLM analysis. Regressors of interest are typically obtained by convolving impulses or boxcar functions, which are representations of the events or conditions of interest, with a model of the BOLD response (HRF). In the study of spontaneous EEG activity without a task condition we can use the EEG signal to derive the input for GLM. In literature several methods for the analysis of simultaneous acquired EEG-fMRI data are proposed. The aim is to find regions of BOLD change linked to the discharges. In the conventional approach each event is marked by visual inspection of the EEG data recorded in the scanner, then a series of identical impulses functions (delta functions) are created and convolved with a canonical HRF, obtaining the regressor for a GLM. The methods presented in Formaggio et al., 2008 and Manganotti et al., 2008 are two attempts of EEG and fMRI integration. However in the first study signals were recorded simultaneous but their correlation analysis was as whether they were recorded in separate sessions, while in the second one we used a conventional approach based on the creation of the regressor as a set of stick functions representing the timing of IED activity. Hence the necessity to developed a new method of integration. The new method aimed to improve upon existing methods since the epileptiform activity, recorded from a scalp EEG, is used to modulate changes in BOLD signal. ICA decomposition is used to identify signals representing activity of interest but one of the major difficulties is their identification. We proposed an automatic selection based on wavelet analysis, because typically IEDs activity is higher in amplitude than background activity and its power increases. The reconstructed EEG signal is obtained with the only contribution of the selected components, method used in many studies to remove artifact from EEG traces. Like in the resting state studies, where alpha rhythm or its spectrum is used as a regressor in GLM analysis, the power time series of EEG signal is used as GLM input. Using conventional approach each event is treated as equal, although epileptic spikes may vary in amplitude, duration and also in appearance. They ignore the fact that IED activity is continuous and contains also fluctuating subthreshold epileptic activity, not clearly seen on surface EEG recordings. In contrast, such meaningful information is contained in the ICA factors employed in our method. Analysis of in silico data validates the method, since demonstrates the reliability of reconstructed IED regressor. All five patients with partial epilepsy we enrolled in this study had frequent interictal focal slow wave activity on routine EEG. In all continuous EEG-fMRI recording sessions, after fMRI artifact removal, we obtained a good quality EEG that allowed us to detect spontaneous IEDs and analyze the related BOLD activation. In their focal distribution, these BOLD activations resembled the focal IEDs seen on routine scalp EEG and EEG recorded during EEG-fMRI sessions; and they are in agreement with the clinical history of the patients. We plan to increase the number of patients and also test this method on EEG with various patterns other than the epileptiform discharges, for example in resting state analysis where, like in the context of epilepsy, the activation task used to drive GLM analysis is missing. For this reason EEG signal is necessary to evaluate hemodynamic changes in fMRI and its analysis is fundamental to derive informations on the electrical activity. Even if it is believed that the HRF to epileptic spikes does not vary significantly from that to external stimuli, HRF could shows different peak times or even non canonical shape in the epileptogenic zone. This observation may be advanced as a working hypothesis for further investigating the choice of HRF in patients with epilepsy; future developments possibly involve a study of BOLD signal in this category of patients, and its relation with the electrical activity. In this way the sensitivity of EEG-fMRI studies in epilepsy could be improved with the use of different HRFs. Moreover, in the future, we will test the integration method to data filtered with the new algorithm in order to conclude this project.
Introduzione La registrazione simultanea fra l’elettroencefalogramma (EEG) e la risonanza magnetica funzionale (fMRI) è un importante strumento nel campo del neuroimaging funzionale che unisce l’alta risoluzione spaziale delle immagini fMRI (1-2 mm) con l’alta risoluzione temporale dell’EEG (ms). Registrare il segnale EEG durante l’acquisizione di immagini fMRI permette di identificare l’attività cerebrale e di ottenere informazioni localizzatorie sui generatori di tale attività. Nonostante i numerosi problemi legati alla presenza di artefatti sul segnale e sulle immagini, dovuti all’interazione fra le due apparecchiature, tale metodica si sta affermando e rafforzando all’interno delle neuroscienze. I campi di applicazioni sono diversi e in particolare la coregistrazione EEG-fMRI può essere utilizzata per studiare e descrivere l’attività elettrica spontanea durante una condizione di riposo (resting state), durante il sonno o causata da forme di epilessia. Molti pazienti con una forma di epilessia farmaco-resistente non possono sottoporsi ad un intervento chirurgico, in quanto la semplice risonanza magnetica non permette l’individuazione della sorgente epilettogena. In questo senso la registrazione simultanea dell’EEG e della fMRI permetterebbe l’identificazione di una possibile sorgente, legata direttamente all’attività elettrica del paziente. Il cambiamento dell’attività neuronale, infatti, è associato ad un cambiamento del rapporto di concentrazione nel sangue fra l’emoglobina ossigenata e quella deossigenata e tale cambiamento può essere misurato attraverso l’effetto BOLD (Blood Oxygen Level Dependent). Le attivazioni cerebrali, infatti, sono date da alterazioni coordinate dell’attività elettrica regionale e del flusso sanguigno cerebrale. La tecnica di coregistrazione EEG-fMRI permette di evidenziare, nel momento in cui si verifica un evento elettrico, un’area di alterato contenuto di desossiemoglobina dovuta ad un aumentato afflusso ematico nella zona cerebrale che genera tale segnale EEG. In genere l’fMRI è usata in studi in cui è presente una condizione sperimentale che differisce da una condizione di riposo, entrambe controllate da un operatore. Il principio base dell’analisi fMRI è il confronto tra un’attività basale cerebrale ed un’attività dovuta ad un evento da studiare (spontaneo o evocato), al fine di ottenere una variazione relativa di flusso ematico. Nello studio dell’epilessia si può considerare l’EEG a riposo come condizione di controllo mentre come condizione sperimentale può essere usato il segnale EEG caratterizzato dalla presenza di eventi parossistici (crisi o attività intercritica). L’analisi convenzionale applicata ai dati EEG-fMRI consiste nell’individuazione visiva da parte del neurologo degli intervalli temporali di interesse, che caratterizzano l’attività intercritica del paziente. Dalla convoluzione degli eventi, rappresentati matematicamente da impulsi, con un modello di risposta emodinamica (haemodynamic response function: HRF), si ottiene il regressore utilizzato nell’analisi General Linear Model (GLM). Si producono così mappe di elevata risoluzione spaziale delle aree cerebrali che generano l’evento patologico osservato. Inoltre l’EEG-fMRI associata ad altre metodiche come video-EEG, risonanza magnetica nucleare (RMN) convenzionale, tomografia computerizzata ad emissione di fotoni singoli (SPECT), tomografia ad emissione di positroni (PET), spettroscopia ecc. contribuisce allo studio di pazienti epilettici candidati alla terapia chirurgica. Lo scopo della presente tesi è quello di sviluppare un metodo automatico, basato sull’analisi delle componenti indipendenti (ICA), per individuare l’attività intercritica in esame, al fine di utilizzare il segnale EEG in toto per la generazione di mappe di attivazione fMRI. Il Nuovo Metodo La qualità dei dati è molto importante nel processo di integrazione; pertanto è necessario applicare un pre-processing ad entrambe le tipologie di dati. Mentre tale elaborazione è standard per i dati fMRI, non lo è per i dati EEG. In letteratura sono stati sviluppati diversi metodi per rimuovere l’artefatto da gradiente di campo magnetico e quello da pulsazione cardiaca. Il metodo per la rimozione dell’artefatto da gradiente implementato nel nostro sistema di acquisizione EEG non ha dato dei risultati completamente soddisfacenti in alcune situazioni. Pertanto è stato necessario implementare un nuovo metodo. Tuttavia l’implementazione di questo nuovo filtro è iniziata contemporaneamente all’implementazione del nuovo metodo di integrazione EEG-fMRI e la sua applicazione su segnali di pazienti epilettici è ancora in atto. Per questi motivi e per non introdurre ulteriori variabili nella validazione del metodo di integrazione, è stato deciso di utilizzare l’algoritmo implementato nel software di acquisizione EEG. In seguito ad un pre-processamento dei dati, caratterizzato da un cambio di referenza e da opportuni filtraggi, è stato applicato il metodo delle componenti indipendenti. L’ICA è una tecnica statistica che permette di individuare le componenti che stanno alla base di una serie multidimensionale di dati, assumendo che le sorgenti siano statisticamente indipendenti e la loro distribuzione non sia gaussiana. Tale analisi è stata effettuata utilizzando l’algoritmo FastICA implementato in EEGLAB ed ha prodotto un numero di componenti per ciascun tracciato pari al numero dei canali EEG. Il nuovo metodo può essere suddiviso in 4 passaggi: • Selezione delle componenti • Ricostruzione del segnale EEG • Selezione del canale ed analisi FFT • Costruzione del regressore EEG Il punto cruciale è la scelta delle componenti che descrivono l’attività intercritica in esame. Per ogni componente si è calcolata la trasformata wavelet continua negli intervalli di interesse che fornisce i valori di potenza nel tempo in funzione della frequenza. Selezionando la frequenza massima si è ottenuto un segnale dipendente esclusivamente dal tempo. Successivamente è stato calcolato il valore medio nell’intervallo temporale e sono state scelte le componenti con più elevata potenza. In seguito si è ricostruito il segnale EEG utilizzando solo il contributo delle componenti scelte. E’ stata applicata un’analisi in frequenza utilizzando la Fast Fourier Transform (FFT) ad epoche di durata pari al tempo di acquisizione di un volume di fMRI; la potenza ottenuta è stata convoluta con la risposta emodinamica scelta ottenendo un modello chiamato ‘regressore’ usato successivamente nella stima GLM dell’analisi fMRI. Questo metodo è stato validato utilizzando dati simulati, ed in seguito applicato a due datasets: il primo composto da due soggetti sani a cui è stata fatta la coregistrazione EEG-fMRI durante apertura e chiusura degli occhi, il secondo composto da 5 pazienti con epilessia parziale a cui è stata fatta la registrazione simultanea in condizione di riposo. L’applicazione del metodo ai dati simulati ha portato alla sua validazione. In tutte e tre le simulazioni si sono ottenute delle forme d’onda, rappresentanti i regressori, molto simili ai regressori assunti come “veri”. Nei due soggetti sani, che hanno svolto un task di apertura e chiusura degli occhi, l’analisi ha prodotto un’attivazione degli occhi ed una deattivazione occipitale, in accordo con i networks ormai noti dalla letteratura. Per quanto riguarda i pazienti, l’integrazione dei due segnali ha portato ad attivazioni concordi con l’attività elettrica e con il loro quadro clinico in 4 pazienti su 5. Le componenti scelte in base al metodo rispecchiano visivamente l’attività parossistica visibile nel tracciato EEG registrato durante acquisizione fMRI e confrontato con l’EEG standard acquisito di routine. Discussione In questo lavoro è stato presentato un nuovo metodo di integrazione fra un segnale neurofisiologico (EEG) e dati di neuroimaging funzionale (fMRI), basato sull’analisi delle componenti indipendenti. Il paradigma sperimentale (protocollo) è un dato molto importante per l’analisi fMRI, infatti le informazioni legate al task e alla condizione di riposo sono utilizzate come ingresso nell’analisi GLM. In assenza di un task, come nello studio dell’epilessia, è necessario utilizzare il segnale EEG per pilotare l’analisi GLM. In letteratura sono stati proposti diversi metodi di integrazione. Nell’approccio convenzionale il protocollo, formato dagli intervalli temporali degli eventi di interesse individuati in seguito ad ispezione visiva, viene convoluto con un modello di risposta emodinamica, ottenendo il regressore per l’analisi GLM. I metodi presentati in Formaggio et al., 2008 e in Manganotti et al., 2008 rappresentano due primi tentativi di integrazione. Tuttavia nel primo studio i segnali vengono analizzati come se fossero stati acquisiti in due sessioni separate, mentre nel secondo studio viene utilizzato l’approccio convenzionale. Da qui la necessità di sviluppare un nuovo metodo di integrazione. Il nuovo metodo ha lo scopo di migliorare quelli già esistenti sfruttando l’informazione derivante da tutto il segnale EEG e non tenendo conto dei soli intervalli temporali di interesse. Il punto cruciale è l’identificazione del segnale legato all’attività di interesse. E’ stato proposto un metodo automatico per facilitare tale scelta, basato sulle trasformate wavelet e valorizzando il contenuto energetico del segnale. Il segnale EEG ricostruito è ottenuto con il solo contributo delle componenti scelte ed in fine la sua potenza spettrale viene utilizzata come ingresso nell’analisi GLM. Uno degli scopi futuri sarà quello di aumentare il numero dei pazienti e di testare il metodo anche su altre tipologie di EEG, come ad esempio quello legato alla condizione di resting state. Anche in questo caso, infatti, manca la presenza di un task che possa pilotare l’analisi GLM, e l’EEG risulta l’unico strumento di informazione per poter arrivare a delle mappe di attivazione. Un ulteriore progetto futuro è legato alla scelta della risposta emodinamica HRF. Tale risposta potrebbe non essere identica a quella ottenuta in seguito ad un task o ad uno stimolo esterno; il suo picco e la sua forma potrebbero infatti essere diversi nella zona epilettogena. In questo senso la sensibilità degli studi EEG-fMRI nell’epilessia potrebbe migliorare utilizzando diverse HRF. In fine verrà applicato il nuovo metodo di integrazione a dati EEG filtrati con il nuovo algoritmo sviluppato.

A new method for the estimation of the EEG evoked potential (EP) is presented in this thesis. This method is based on a new model of the EEG response which is assumed to be the sum of the EP and independent correlated Gaussian noise representing the spontaneous EEG activity. The EP is assumed to vary in both shape and latency, with the shape variation represented by correlated Gaussian noise which is modulated by the EP. The latency of the EP is also assumed to vary over the ensemble of responses in a random manner governed by some unspecified probability density. No assumption on stationarity is needed for the noise.
With the model described in state-space form, a Kalman filter is constructed, and the variance of the innovation process of the response measurements is derived. A maximum likelihood solution to the EP estimation problem is then obtained via this innovation process.
Tests using simulated responses show that the method is effective in estimating the EP signal at signal-to-noise ratio as low as -6db. Other tests using real normal visual response data yield reasonably consistent EP estimates whose main components are narrower and larger than the ensemble average. In addition, the likelihood function obtained by our method can be used as a discriminant between normal and abnormal responses, and it requires smaller ensembles than other methods.

Thesis advisor: Robert Signorile
It has long been known that as neurons fire within the brain they produce measurable electrical activity. Electroencephalography (EEG) is the measurement and recording of these electrical signals using sensors arrayed across the scalp. Though there is copious research in using EEG technology in the fields of neuroscience and cognitive psychology, it is only recently that the possibility of utilizing EEG measurements as inputs in the control of computers has emerged. The idea of Brain-Computer Interfaces (BCIs) which allow the control of devices using brain signals evolved from the realm of science fiction to simple devices that currently exist. BCIs naturally present themselves to many extremely useful applications including prosthetic devices, restoring or aiding in communication and hearing, military applications, video gaming and virtual reality, and robotic control, and have the possibility of significantly improving the quality of life of many disabled individuals. However, current BCIs suffer from many problems including inaccuracies, delays between thought, detection, and action, exorbitant costs, and invasive surgeries. The purpose of this research is to examine the Emotiv EPOC© System as a cost-effective gateway to non-invasive portable EEG measurements and utilize it to build a thought-based BCI to control the Parallax Scribbler® robot. This research furthers the analysis of the current pros and cons of EEG technology as it pertains to BCIs and offers a glimpse of the future potential capabilities of BCI systems
Thesis (BA) — Boston College, 2010
Submitted to: Boston College. College of Arts and Sciences
Discipline: Computer Science Honors Program
Discipline: Computer Science

This thesis addresses some theoretical issues generated by the results of recent analysis of EEG time series proving the brain dynamics are driven by abrupt changes making them depart from the ordinary Poisson condition. These changes are renewal, unpredictable and non-ergodic. We refer to them as crucial events. How is it possible that this form of randomness be compatible with the generation of waves, for instance alpha waves, whose observation seems to suggest the opposite view the brain is characterized by surprisingly extended coherence? To shed light into this apparently irretrievable contradiction we propose a model based on a generalized form of Langevin equation under the influence of a periodic stimulus. We assume that there exist two different forms of time, a subjective form compatible with Poisson statistical physical and an objective form that is accessible to experimental observation. The transition from the former to the latter form is determined by the brain dynamics interpreted as emerging from the cooperative interaction among many units that, in the absence of cooperation would generate Poisson fluctuations. We call natural time the brain internal time and we make the assumption that in the natural time representation the time evolution of the EEG variable y(t) is determined by a Langevin equation perturbed by a periodic process that in this time representation is hardly distinguishable from an erratic process. We show that the representation of this random process in the experimental time scale is characterized by a surprisingly extended coherence. We show that this model generates a sequence of damped oscillations with a time behavior that is remarkably similar to that derived from the analysis of real EEG's. The main result of this research work is that the existence of crucial events is not incompatible with the alpha wave coherence. In addition to this important result, we find another result that may help our group, or any other research group working on the analysis of brain's dynamics, to prove or to disprove the existence of crucial events. We study the diffusion process generated by fluctuations emerging from the same model after filtering out the alpha coherence, and we study the recursion to the origin. We study the survival probability of this process, namely the probability that up to a given time no re-crossing of the origin occurs. We find that this is an inverse power law with a power that depends on whether or not crucial events exist.

This report is about a project where electroencephalography (EEG) wasused to control a two player game. The signals from the EEG-electrodeswere amplified, filtered and processed. Then the signals from the playerswere compared and an algorithm decided what would happen in the gamedepending on which signal was largest. The controls and the gaming mechanismworked as intended, however it was not possible to gather a signal fromthe brain with the method used in this project. So ultimately the goal wasnot reached.
electroencephalography, EEG

This item is only available in print in the UCF Libraries. If this is your Honors Thesis, you can help us make it available online for use by researchers around the world by following the instructions on the distribution consent form at http://library.ucf.edu/Systems/DigitalInitiatives/DigitalCollections/InternetDistributionConsentAgreementForm.pdf You may also contact the project coordinator, Kerri Bottorff, at kerri.bottorff@ucf.edu for more information.
Bachelors
Arts and Sciences
Psychology

The heritability of frontal EEG asymmetry, a potential endophenotype for depression, was investigated using a large set of adolescent and young adult twins. Additionally, the relationship between polymorphisms within three serotonin genes, two receptor genes and one transporter gene, and frontal EEG asymmetry was also investigated. Using Falconer's estimate, frontal EEG asymmetry was shown to be more heritable at lateral compared to medial cites across nearly all reference montages, and greater in males compared to females. Using structural equation modeling (SEM), and investigating both additive (ACE) and non-additive (ADE) models of genetic heritability, males displayed consistently greater additive genetic contributions to heritability, with greater lateral contributions than medial ones. For female twins pairs, the additive genetic model data provided a mixed picture, with more consistent heritability estimates observed at medial sites, but with larger estimates shown at lateral channels. For non-additive genetic models, male twin pairs demonstrated exclusive non-additive contributions to heritability across channels within AVG and CZ referenced data, with metrics in the CSD and LM montages more mixed between additive and non-additive contributions. However, consistent with Falconer's estimates, lateral channels were nearly always estimated to be more heritable than medial channels regardless of gender. These models demonstrate some combination of additive and non-additive contributions to the heritability of frontal EEG asymmetry, with the CSD and AVG montages showing greater lateral compared to medial heritability and CZ and LM montages showing mixed contributions with additive heritability at lateral channels and non-additive primarily at medial channels. The complex interaction of gender and reference montage on the heritability estimates highlight the subtle yet important roles of age, gender, and recording methodology when investigating proposed endophenotypes. However, no association was found between the proposed polymorphisms in serotonin receptor 1a, 2a or serotonin transporter genes and frontal EEG asymmetry. Although the results support modest heritability of frontal EEG asymmetry, the proposed link to underlying serotonergic genetic markers remains an open question. Overall, these results indicate that frontal asymmetry may be a useful endophenotype for depressive risk with modest heritability, but is one that taps more environmental risk.

Recent advances in neuroimaging and electromagnetic measurement technology have permitted the exploration of structural and functional brain alterations associated with chronic pain. A number of cortical and subcortical brain regions have been found to be involved in the experience of chronic pain (Baliki et al., 2008; Jensen et al., 2010). Evidence suggests that living with chronic pain shapes the brain from both an architectural and a functional perspective, and that individuals living with chronic pain display altered brainwave activity even at rest. Quantitative EEG (qEEG) is a method of spectral analysis that utilizes a fast Fourier transform algorithm to convert analog EEG signals into digital signals, allowing for precise quantification and analysis of signals both at single electrode locations and across the scalp as a whole. An important advance that has been permitted by qEEG analysis is the development of lifespan normative databases against which individual qEEGs can be compared (Kaiser, 2006; Thatcher et al, 2000). Pilot data utilizing qEEG to examine brainwave patterns of individuals with chronic pain have revealed altered EEG activity at rest compared to age- and gender-matched healthy individuals (Burroughs, 2011). The current investigation extended the findings of the pilot study by utilizing qEEG to examine a larger sample of individuals with chronic pain. Individuals with chronic pain displayed significantly reduced slow wave activity in frontal, central, and temporal regions. Findings will be presented in terms of specific patterns of altered EEG activity seen in individuals with chronic pain.

Thesis (Ph. D.)--University of Oregon, 2007.
Typescript. Includes vita and abstract. Includes bibliographical references (leaves 126-133). Also available for download via the World Wide Web; free to University of Oregon users.

In this work, we present a set of methods aimed at improving the discriminative power of time-varying features of signals that contain noise. These methods use properties of noise signals as well as information theoretic techniques to factor types of noise and support signal inference for electroencephalographic (EEG) based brain-machine interfaces (BMI). EEG data were collected over two studies aimed at addressing Psychophysiological issues involving symmetry and mental rotation processing. The Psychophysiological data gathered in the mental rotation study also tested the feasibility of using dissociations of mental rotation tasks correlated with rotation angle in a BMI. We show the feasibility of mental rotation for BMI by showing comparable bitrates and recognition accuracy to state-of-the-art BMIs. The conclusion is that by using the feature selection methods introduced in this work to dissociate mental rotation tasks, we produce bitrates and recognition rates comparable to current BMIs.

The brain is a highly complex network of nonlinear systems with internal dynamic states that are not easily quantified. As a result, it is essential to understand the properties of the connectivity network linking disparate parts of the brain used in complex cognitive processes, such as working memory. Working memory is the system in control of temporary retention and online organization of thoughts for successful goal directed behavior. Individuals exhibit a typically small capacity limit on the number of items that can be simultaneously retained in working memory. To modify network connections and thereby augment working memory capacity, researchers have targeted brain areas using a variety of noninvasive stimulation interventions. However, few existing methods take advantage of the brain's own structure to actively generate and entrain internal oscillatory modulations in locations deep within the auditory pathways. One technique is known as binaural beats, which arises from the brain's interpretation of two pure tones, with a small frequency mismatch, delivered independently to each ear. The mismatch between these tones is perceived as a so-called beat frequency which can be used to modulate behavioral performance and cortical connectivity. Currently, all binaural stimulation therapeutic systems are open-loop "one-size-fits-all" approaches. However, these methods can prove not as effective because each person's brain responds slightly differently to exogenous stimuli. Therefore, the driving motivation for developing a closed-loop stimulation system is to help populations with large individual variability. One such example is persons with mild cognitive impairment (MCI), which causes cognitive impairments beyond those expected based on age. Therefore, applying a closed-loop binaural beat control system to increase the cognitive load level to people with MCI could potentially maintain their quality of life. In this dissertation, I will present a comparison of algorithms to determine brain connectivity, results of open-loop based binaural stimulation, the development of a closed-loop brain network simulation platform, and finally an experimental study to determine the effectiveness of closed-loop control to modulate brain networks hence influencing cognitive abilities.
Ph. D.

Thesis (Ph.D)--Computing, Georgia Institute of Technology, 2010.
Committee Member: Alexander Gray; Committee Member: Charles Lee Isbell Jr.; Committee Member: Melody Moore Jackson; Committee Member: Paul M. Corballis; Committee Member: Thad Starner. Part of the SMARTech Electronic Thesis and Dissertation Collection.

Decision making can be a complicated process involving perception of the present situation, past experience and knowledge necessary to foresee a better future. This cognitive process is one of the essential human ability that is required from everyday walk of life to making major life choices. Although it may seem ambiguous to translate such a primitive process into quantifiable science, the goal of this thesis is to break it down to signal processing and quantifying the thought process with prominence of EEG signal power variance. This paper will discuss the cognitive science, the signal processing of brain signals and how brain activity can be quantifiable through data analysis. An experiment is analyzed in this thesis to provide evidence that theta frequency band activity is associated with stress and stress is negatively correlated with concentration and problem solving, therefore hindering decision making skill. From the results of the experiment, it is seen that theta is negatively correlated to delta and beta frequency band activity, thus establishing the fact that stress affects internal focus while carrying out a task.

This master thesis report describes signal processing parameters of electroencephalography (EEG) signals with a significant difference between the signals from the animal model of clinical depression and the non-depressed animal model. The signal from the depressed model had a weaker power in gamma (30 - 80 Hz) than the non-depressed model during awake and it had a stronger power in delta (1.5 - 4 Hz) during sleep. The report describes the process of using visualisation to understand the shape of the signal which helps with interpreting results and helps with the development of parameters. A generic tool for time-frequency analysis was improved to cope with the size of the weeklong EEG dataset. A method for evaluating the quality of how well the EEG parameters are able to separate the strains with as short recordings as possible was developed. This project shows that it is possible to separate an animal model of depression from an animal model of non-depression based on its EEG and that EEG-classifiers may work as indicative classifiers for depression. Not a lot of data is needed. Further studies are needed to verify that the results are not overly sensitive to recording setup and to study to what extent the results are translational. It might be some of the EEG parameters with significant differences described here are limited to describe the difference between the two strains FSL and SD. But the classifiers have reasonable biological explanations that makes them good candidates for being translational EEG-based classifiers for clinical depression.

Un problema principal de la ciència de xarxes és com reconstruir (inferir) la topologia d’una xarxa real a partir de senyals mesurades de les seves unitats internes. Entendre la arquitectura d’una xarxa complexa és clau, no només per comprendre el seu funcionament, sinó també per predir i controlar el seu comportament. Els mètodes actualment disponibles es centren principalment en la detecció d’enllaços de xarxes no direccio- nals i sovint requereixen suposicions fortes sobre el sistema. Tanmateix, molts d’aquests mètodes no es poden aplicar a xarxes amb connexions direccionals. Per abordar aquest problema, en aquesta tesis ens centrarem en la inferència de xarxes direccionals. Concretament, desenvolupem un mètode de reconstrucció de xarxes basat en models que combina estadístiques de correlacions de derivades amb recuit simulat. A més, desenvolupem un mètode de reconstrucció basat en dades cimentat en una mesura d’interpedendència no lineal. Aquest mètode permet inferir la topologia de xarxes direccionals d’oscil.ladors caòtics de Lorenz per un subordre de la força d’acoblament i la densitat de la xarxa. Finalment, apliquem el mètode basat en dades a gravacions electroencefalogràfiques d’un pacient amb epilèpsia. Les xarxes cerebrals funcionals obtingu- des a partir d’aquest mètode són coherents amb la informació mèdica disponible.
Un problema principal de la ciencia de redes es cómo reconstruir (inferir) la topología de una red real usando la señales medidas de sus unidades internas. Entender la arquitectura de redes complejas es clave, no solo para entender su funcionamiento pero también para predecir y controlar su comportamiento. Los métodos existentes se focalizan en la detección de redes no direccionales y normalmente requieren fuertes suposicio- nes sobre el sistema. Sin embargo, muchos de estos métodos no pueden ser aplicados en redes con conexiones direccionales. Para abordar este problema, en esta tesis estudiamos la reconstrucción de redes direccio- nales. En concreto, desarrollamos un método de reconstrucción basado en modelos que combina estadísticas de correlaciones de derivadas con recocido simulado. Además, desarrollamos un método basado en datos cimentado en una medida d’interdependencia no lineal. Este método permite inferir la topología de redes direccionales de osciladores caóticos de Lorenz para un subrango de la fuerza de acoplamiento y densidad de la red. Finalmente, aplicamos el método basado en datos a grabaciones electroencefalográficas de un paciente con epilepsia. Las redes cerebra- les funcionales obtenidas usando este método son consistentes con la información médica disponible.
A foremost problem in network science is how to reconstruct (infer) the topology of a real network from signals measured from its internal units. Grasping the architecture of complex networks is key, not only to understand their functioning, but also to predict and control their behaviour. Currently available methods largely focus on the detection of links of undirected networks and often require strong assumptions about the system. However, many of these methods cannot be applied to networks with directional connections. To address this problem, in this doctoral work we focus at the inference of directed networks. Specifically, we develop a model-based network reconstruction method that combines statistics of derivative-variable correlations with simulated annealing. We furthermore develop a data-driven reconstruction method based on a nonlinear interdependence measure. This method allows one to infer the topology of directed networks of chaotic Lorenz oscillators for a subrange of the coupling strength and link density. Finally, we apply the data-driven method to multichannel electroencephalographic recordings from an epilepsy patient. The functional brain networks obtained from this approach are consistent with the available medical information.

Frontal electroencephalographic (EEG) alpha (α) asymmetry may index the activation of lateralized affect and motivation systems in humans. Resting EEG activation was measured and its relationship to Iowa gambling task (IGT) performance was evaluated. No effects were found for α power asymmetry. However, beta (β) power asymmetry, an alternative measure of resting EEG activation, was associated with the number of risky decisions made in the early portion of the task. Additionally, IGT deck selection patterns suggest there are at least three distinct performance styles in healthy individuals. Interestingly, β power asymmetry contradicts performance predictions based on accepted frontal asymmetry affect and motivation models.