Les parcellations cérébrales sont appliquées en neuroimagerie pour aider les chercheurs à ré- duire la haute dimensionnalité des données d’IRM fonctionnelle. L’objectif principal est une meilleure compréhension de l’organisation fonctionnelle du cerveau tant chez les sujets sains que chez les sujets souffrant de troubles neurologiques, dont la maladie d’Alzheimer. Malgré la vague d’approches de parcellations précédentes, les mesures de performance doivent en- core être améliorées pour générer des parcellations fiables, même avec de longues acquisitions. Autrement dit, une reproductibilité plus élevée qui permet aux chercheurs de reproduire des parcellations et de comparer leurs études. Il est également important de minimiser la perte d’informations entre les données compressées et les données brutes pour représenter avec précision l’organisation d’un cerveau individuel. Dans cette thèse, j’ai développé une nou- velle approche pour parcellaire le cerveau en reconfigurations spatiales distinctes appelées «états dynamiques de parcellations». J’ai utilisé une méthode d’agrégation de cluster simple DYPAC1.0 de parcelles basées sur des semences sur plusieurs fenêtres de temps. J’ai émis l’hypothèse que cette nouvelle façon de formaliser le problème de parcellisation améliorera les mesures de performance par rapport aux parcellations statiques. Le premier chapitre de ce document est une introduction générale au contexte des réseaux à grande échelle du cerveau humain. Je montre également l’importance des parcellations pour une meilleure compréhension du cerveau humain à l’aide de connectomes fonctionnels afin de prédire les schémas de progression de la maladie. Ensuite, j’explique pourquoi le problème de parcelli- sation cérébrale est difficile et les différentes questions de recherche ouvertes associées à ce domaine. Mes contributions à la recherche sont subdivisées en deux articles. Les deuxième et troisième chapitres sont consacrés au premier article principal et à son supplément publié dans Network Neuroscience Journal. Le quatrième chapitre représente le deuxième document en préparation. Le cinquième chapitre conclut mes contributions et ses implications dans le domaine de la neuroimagerie, ainsi que des orientations de recherche ouvertes. En un mot, la principale conclusion de ce travail est l’existence de reconfigurations spatiales distinctes dans tout le cerveau avec des scores de reproductibilité presque parfaits sur les données de test-retest (jusqu’à 0,9 coefficient de corrélation de Pearson). Un algorithme d’agrégation de cluster simple et évolutif appelé DYPAC 1.0 est expliqué pour identifier ces reconfigu- rations ou «états dynamiques de parcellations» pour des sous-réseaux de départ spécifiques (deuxième chapitre). L’analyse de ces états a montré l’existence d’un répertoire plus riche «d’états dynamiques» dans le cas des cortex hétéromodaux (ex: cortex cingulaire posté- rieur et cortex cingulaire antérieur dorsal) par rapport aux cortex unimodaux (ex: cortex visuel). En outre, les résultats de l’analyse de reproductibilité ont montré que DYPAC 1.0 a de meilleurs résultats de reproductibilité (en termes de corrélation de Pearson) par rapport aux parcelles statiques (deuxième chapitre). Plusieurs analyses démontrent que DYPAC 1.0 est robuste au choix de ses paramètres (troisième chapitre). Ces résultats et l’évolutivité de DYPAC 1.0 ont motivé une analyse complète du niveau cérébral. Je présente DYPAC 2.0 comme une approche au niveau cérébral complet pour fragmenter le cerveau en «états dynamiques de parcellations». Des reconfigurations spatiales distinctes et se chevauchant ou «états dynamiques» sont identifiées pour différentes régions du cerveau (quatrième chapitre). Ces états ont des scores de compression prometteurs qui montrent une faible perte d’infor- mations entre les cartes de stabilité d’état réduit et les données d’origine dans les cortex cérébraux, c’est-à-dire jusqu’à seulement 20% de perte de la variance expliquée. Cette thèse présente ainsi de nouvelles contributions dans le domaine de la parcellisation fonctionnelle qui pourraient avoir un impact sur la manière dont les chercheurs modélisent les interactions riches et dynamiques entre les réseaux cérébraux dans la santé et la maladie. / Brain parcellations are applied in neuroimaging to help researchers reduce the high dimen- sionality of the functional MRI data. The main objective is a better understanding of the brain functional organization in both healthy subjects and subjects having neurological dis- orders, including Alzheimer disease. Despite the flurry of previous parcellation approaches, the performance measures still need improvement to generate reliable parcellations even with long acquisitions. That is, a higher reproducibility that allows researchers to replicate par- cellations and compare their studies. It is also important to minimize the information loss between the compressed data and the raw data to accurately represent the organization of an individual brain. In this thesis, I developed a new approach to parcellate the brain into distinct spatial reconfigurations called “dynamic states of parcellations”. I used a simple cluster aggregation method DYPAC1.0 of seed based parcels over multiple time windows. I hypothesized this new way to formalize the parcellation problem will improve performance measures over static parcellations. The first chapter of this document is a general context introduction to the human brain large scale networks. I also show the importance of par- cellations for a better understanding of the human brain using functional connectomes in order to predict patterns of disease progression. Then, I explain why the brain parcellation problem is hard and the different open research questions associated with this field. My research contributions are subdivided into two papers. The second and the third chapters are dedicated to the first main paper and its supplementary published in Network Neuro- science Journal. The fourth chapter represents the second paper under preparation. The fifth chapter concludes my contributions and its implications in the neuroimaging field, along with open research directions. In a nutshell, the main finding of this work is the existence of distinct spatial reconfigurations throughout the brain with near perfect reproducibility scores across test-retest data (up to .9 Pearson correlation coefficient). A simple and scalable clus- ter aggregation algorithm called DYPAC 1.0 is explained to identify these reconfigurations or “dynamic states of parcellations” for specific seed subnetworks (second chapter). The analysis of these states showed the existence of a richer repertoire of “dynamic states” in the case of heteromodal cortices (e.g., posterior cingulate cortex and the dorsal anterior cingulate cortex) compared to unimodal cortices (e.g., visual cortex). Also, the reproducibility analysis results showed that DYPAC 1.0 has better reproducibility results (in terms of Pearson corre- lation) compared to static parcels (second chapter). Several analyses demonstrate DYPAC 1.0 is robust to the choice of its parameters (third chapter). These findings and the scalabil- ity of DYPAC 1.0 motivated a full brain level analysis. I present DYPAC 2.0 as the full brain level approach to parcellate the brain into “dynamic states of parcellations”. Distinct and overlapping spatial reconfigurations or “dynamic states” are identified for different regions throughout the brain (fourth chapter). These states have promising compression scores that show low information loss between the reduced state stability maps and the original data throughout the cerebral cortices, i.e. up to only 20% loss in explained variance. This thesis thus presents new contributions in the functional parcellation field that may impact how researchers model the rich and dynamic interactions between brain networks in health and disease.

No description available.

Les affects, émotions et humeurs sont des processus complexes dont le fonctionnement précis échappe toujours à la neuroscience affective. Un récent mouvement des études IRMf s’est tourné vers la recherche d’effets aux niveaux inter- et intra-individuels en raison du manque d’applicabilité individuelle des résultats provenant de moyennes de groupes basées sur des données transversales. En particulier, la recherche intra-individuelle permet l’étude de liens directs entre l’affectivité et la connectivité chez de mêmes individus à travers le temps. De précédentes études en IRMf rapportent ce type associations chez un unique participant, notre objectif a été d’étudier les effets intra-individuels communs pour un groupe d’individus. Nous avons utilisé le jeu de données Day2day, composé de 40 à 50 sessions pour 6 participants, chaque session incluant des données d’IRMf au repos ainsi que d’auto-évaluations des états affectifs. Nous avons analysé la relation entre l’affectivité et la connectivité fonctionnelle entre des régions cérébrales précédemment liées aux émotions et affects à l’aide de régressions linéaires mixtes multivariées. Nos modèles ont isolé des patrons de connectivité communs et généralisables liés aux variations intra-individuelles de l’affectivité observées au cours de plusieurs semaines et mois. Ces modèles impliquaient particulièrement l’amygdale et l’insula. Nos résultats ouvrent la possibilité de reproduire de tels modèles sur des jeux de données plus larges ainsi qu’à évaluer l’hétérogénéité entre sujets au-delà des effets moyens. La caractérisation de tels processus neurobiologiques pourrait être d’une grande utilité en clinique comme biomarqueur transdiagnostique de l’état affectif ou potentielle cible thérapeutique. / Affects, emotions and moods are complex processes, the precise functioning of which still eludes affective neuroscience. A recent movement in fMRI has turned to research of effects at the inter- and intra-individual level in response to the lack of individual-level applicability of results from cross-sectional group mean studies. In particular, intra-individual research enables the study of direct links between affective states and underlying connectivity in individuals across time. Previous fMRI studies have described these associations in a single participant, our objective was to find shared intraindividual effects across multiple subjects. We have used the Day2day dataset, comprising 40 to 50 sessions for six participants, each session including data from resting-state fMRI scans and self-report measures of state affectivity. We have investigated the relationship between affectivity and connectivity in brain regions linked to emotions and affects using multivariate mixed linear analysis. Our models have isolated common and generalizable patterns of connectivity linked to variations in affectivity observed over multiple weeks and months. These models involved mainly the amygdala and insula. Our results incentivize the re-creation of such modelsin larger datasets, and to assess heterogeneity beyond group mean effects. The characterization of such neurobiological processes could be of great use in a clinical setting as a transdiagnostic biomarker or as a potential therapeutic target.

While task-based functional magnetic resonance imaging (fMRI) has helped us understand the functional role of many regions in the human brain, many diseases and complex behaviors defy explanation. Alternatively, if no task is performed, the fMRI signal between distant, anatomically connected, brain regions is similar over time. These correlations in “resting state” fMRI have been strongly linked to behavior and disease. Previous work primarily calculated correlation in entire fMRI runs of six minutes or more, making understanding the neural underpinnings of these fluctuations difficult. Recently, coordinated dynamic activity on shorter time scales has been observed in resting state fMRI: correlation calculated in comparatively short sliding windows and quasi-periodic (periodic but not constantly active) spatiotemporal patterns. However, little relevance to behavior or underlying neural activity has been demonstrated. This dissertation addresses this problem, first by using 12.3 second windows to demonstrate a behavior-fMRI relationship previously only observed in entire fMRI runs. Second, simultaneous recording of fMRI and electrical signals from the brains of anesthetized rats is used to demonstrate that both types of dynamic activity have strong correlates in electrophysiology. Very slow neural signals correspond to the quasi-periodic patterns, supporting the idea that low-frequency activity organizes large scale information transfer in the brain. This work both validates the use of dynamic analysis of resting state fMRI, and provides a starting point for the investigation of the systemic basis of many neuropsychiatric diseases.

The unprecedented turbulence and uncertainty experienced in global economic and financial markets because of the 'credit crunch' has had a damaging impact on consumer confidence. Trust and credibility have been eroded as many customers feel let down by the banks suggesting the need for banks to rebuild constructive dialogue and long-term, meaningful relationships with their customers again. Though financial service, in this case, is considered a utilitarian service, based on the fact that money is needed to support people‘s daily activities, the present state of financial service has suggested the need for banks to appeal to consumers‘ emotions with the aim of improving their reputation. Also, the competition within the industry also could suggest the need to adopt an emotionally appealing advertisement strategy as emotions are known to play an influential role in building robust brand preference. This study builds on the communication theory, meaning transfer theory and consumer involvement theory, to understand the messages the banks are sending out and to elicit consumers‘ emotional reaction. One thousand, two hundred and seventy-four UK bank advertisements in nine national newspapers were content-analysed to identify the emotional appeals presented by the banks. The perception of these appeals and their associated meanings were sought through semi-structured interviews with 33 participants in London and Luton. The results of the analysis indicated that UK Banks are utilising emotional appeal in their advertisements to reach out to the consumers to convince them to upgrade their account, to open an additional account or switch their account. The most predominantly used appeals were relief and relaxation followed by excitement and happiness or satisfaction with the bank, and finally, security and adventure. However, variations were found in different financial products that employed emotional appeals. It was found that high-involvement products such as mortgages and loans used fewer emotional appeals. Both bank groups - high street banks, including the big four (Barclays, HSBC, Lloyds and RBS) and non-high street banks, such as the new entrants, supermarket brands, and online banks were using emotional appeals. However, it is acknowledged that the communication strategies between these banks could be different as the non-high street banks are more likely to repeat and publish the same messages across many newspapers, instead of publishing different emotionally appealing advertisements. Though consumers acknowledged these emotional appeals in the advertisements, they were more concerned about their relationship with the banks as they don‘t rely on advertisements to make a financial decision. Rather, recommendations from families, friends and associates and also branch location are more important when deciding on which bank to choose. The lack of congruency between financial services and emotional appeals in advertisements is also observed as customers are more likely to be persuaded by rational appeals however this study has not completely ruled out emotional appeals in bank advertisements as the use of both types of appeals is recommended. The study provides important theoretical and managerial contributions to understanding how the consumers understand meaning-embedded advertisements produced by the banks. Managers will be able to consider the implications of advertisements in enhancing their brand equity and building relationships with customers in anticipation that, by word of the mouth and established relationship, their bank‘s reputation will be enhanced. Limitations of the study and opportunities for future research are identified.

Wo und wie werden komplexe Regelsätze im Gehirn repräsentiert? Drei empirische Studien dieser Doktorarbeit untersuchen dies experimentell. Eine weitere methodische Studie liefert Beiträge zur Weiterentwicklung der genutzten empirischen Methode. Die empirischen Studien nutzen multivariate Musteranalyse (MVPA) funktioneller Magnetresonanzdaten (fMRT) gesunder Probanden. Die Fragestellungen der methodischen Studie wurden durch die empirischen Arbeiten inspiriert. Wirkung und Anwendungsbreite der entwickelten Methode gehen jedoch über die Anwendung in den empirischen Studien dieser Arbeit hinaus. Die empirischen Studien bearbeiten Fragen wie: Wo werden Hinweisreize und Regeln repräsentiert, und sind deren Repräsentationen voneinander unabhängig? Wo werden Regeln repräsentiert, die aus mehreren Einzelregeln bestehen, und sind Repräsentationen der zusammengesetzten Regeln Kombinationen der Repräsentationen der Einzelregeln? Wo sind Regeln verschiedener Hierarchieebenen repräsentiert, und gibt es einen hierarchieabhängigen Gradienten im ventrolateralen präfrontalen Kortex (VLPFK)? Wo wird die Reihenfolge der Regelausführung repräsentiert? Alle empirischen Studien verwenden informationsbasiertes funktionales Mapping ("Searchlight"-Ansatz), zur hirnweiten und räumlich Lokalisierung von Repräsentationen verschiedener Elemente komplexer Regelsätze. Kernergebnisse der Arbeit beinhalten: Kompositionalität neuronaler Regelrepräsentationen im VLPFK; keine Evidenz für Regelreihenfolgenrepräsentation im VLPFK, welches gegen VLPFK als generelle Task-Set-Kontrollregion spricht; kein Hinweis auf einen hierarchieabhängigen Gradienten im VLPFK. Die komplementierende methodische Studie präsentiert "The Same Analysis Approach (SAA)", ein Ansatz zur Erkennung und Behebung experimentspezifischer Fehler, besonders solcher, die aus Design–Analyse–Interaktionen entstehen. SAA ist für relevant MVPA, aber auch für anderen Bereichen innerhalb und außerhalb der Neurowissenschaften. / Where and how does the brain represent complex rule sets? This thesis presents a series of three empirical studies that decompose representations of complex rule sets to directly address this question. An additional methodological study investigates the employed analysis method and the experimental design. The empirical studies employ multivariate pattern analysis (MVPA) of functional magnetic resonance imaging (fMRI) data from healthy human participants. The methodological study has been inspired by the empirical work. Its impact and application range, however, extend well beyond the empirical studies of this thesis. Questions of the empirical studies (Studies 1-3) include: Where are cues and rules represented, and are these represented independently? Where are compound rules (rules consisting of multiple rules) represented, and are these composed from their single rule representations? Where are rules from different hierarchical levels represented, and is there a hierarchy-dependent functional gradient along ventro-lateral prefrontal cortex (VLPFC)? Where is the order of rule-execution represented, and is it represented as a separate higher-level rule? All empirical studies employ information-based functional mapping ("searchlight" approach) to localise representations of rule set features brain-wide and spatially unbiased. Key findings include: compositional coding of compound rules in VLPFC; no order information in VLPFC, suggesting VLPFC is not a general controller for task set; evidence against the hypothesis of a hierarchy-dependent functional gradient along VLPFC. The methodological study (Study 4) introduces "The Same Analysis Approach (SAA)". SAA allows to detect, avoid, and eliminate confounds and other errors in experimental design and analysis, especially mistakes caused by malicious experiment-specific design-analysis interactions. SAA is relevant for MVPA, but can also be applied in other fields, both within and outside of neuroscience.