[ES] El aumento de la esperanza de vida en la población con edad por encima de 50 años está generando un mayor número de casos detectados de cáncer de próstata (CaP). Por este motivo, los recursos se destinan al diagnóstico en etapas tempranas y al tratamiento efectivo. A pesar de la multitud de estudios basados en biomarcadores y discriminación histológica, es difícil diferenciar con efectividad los casos de CaP con baja agresividad de aquellos que progresarán y acabarán produciendo mortalidad o una disminución en la esperanza de vida del paciente. Con el objetivo de mejorar el diagnostico, localización y gradación de los tumores malignos, las técnicas de imagen por Resonancia Magnética (MRI) son las más adecuadas para el estudio del cáncer, proporcionando métodos de diagnóstico no-invasivos, sensibles y específicos, basados en secuencias morfológicas (T2w) y funcionales (perfusión de la sangre y difusión del agua). Las diferentes características y parámetros extraídos de estas secuencias, conocidos como biomarcadores de imagen, pueden evaluar las diferencias asociadas al desarrollo de los procesos tumorales, como los modelos farmacocinéticos para estudiar angiogénesis (perfusión) y los modelos mono- y bi-exponenciales para estudiar la caída de la señal en difusión con el objetivo de estudiar la celularización. Normalmente, estos biomarcadores de imagen se analizan de forma "univariante", sin aprovechar la información de las estructuras de correlación interna que existen entre ellos. Una manera de mejorar este análisis es mediante la aplicación de las técnicas estadísticas que ofrece el Análisis Multivariante de Imágenes (MIA), obteniendo estructuras (latentes) simplificadas que ayudan a entender la relación entre los parámetros (variables) y sus propios procesos fisiológicos, además de reducir la incertidumbre en la estimación de los biomarcadores. En esta tesis, se han desarrollado nuevos biomarcadores de imagen para perfusión y difusión con la aplicación de alguna de las herramientas de MIA como la Resolución Multivariante de Curvas con Mínimos Cuadrados Alternos (MCR-ALS), obteniendo parámetros que tienen interpretación clínica directa. A continuación, los métodos basados en mínimos cuadrados parciales (PLS) se aplicaron para estudiar la capacidad de clasificación de estos biomarcadores. En primer lugar, los biomarcadores de perfusión se utilizaron para la detección de tumores (control vs lesión). Posteriormente, la combinación de perfusión + difusión + T2 se empleó para estudiar agresividad tumoral con la aplicación de métodos PLS multibloque, en concreto (secuencial) SMB-PLS. Los resultados mostrados indican que los biomarcadores de perfusión obtenidos mediante MCR son mejores que los parámetros farmacocinéticos en la diferenciación de la lesión. Con lo que respecta al estudio de la agresividad tumoral, la combinación de los biomarcadores de difusión (empleando ambos métodos: modelos paramétricos y MCR) y los valores de T2w normalizados proporcionaron los mejores resultados. En conclusión, MIA se puede aplicar a las secuencias morfológicas y funcionales de resonancia magnética para mejorar el diagnóstico y el estudio de la agresividad de los tumores en próstata. Obteniendo nuevos parámetros cuantitativos y combinándolos con los biomarcadores más ampliamente utilizados en el ambiente clínico. / [CA] El increment de la esperança de vida en la població per damunt dels 50 anys està generant un major nombre de casos detectats de càncer de pròstata (CaP). Per aquest motiu, els recursos es destinen al diagnòstic en etapes primerenques i al tractament efectiu. Tot i la multitud de estudis basats en biomarcadors y discriminació histològica, es difícil diferenciar amb efectivitat els casos de CaP que tenen baixa agressivitat dels que progressaran y acabaran produint mortalitat o una disminució en la esperança de vida del pacient. Amb el objectiu de millorar el diagnòstic, localització y gradació dels tumors malignes, les tècniques de imatge per Ressonància Magnètica (MRI) son els mètodes més adequats per al estudi del càncer, proporcionant metodologies de diagnòstic no-invasius, sensibles y específiques basades en seqüències morfològiques (T2w) y funcionals (perfusió de la sang y difusió del aigua). Les diferents característiques i paràmetres extrets de aquestes seqüències, coneguts com biomarcadors d'imatge, poden avaluar les diferències associades al desenvolupament dels processos tumorals. Primer, amb els models farmacocinétics per a estudiar angiogènesis (perfusió) y segon, amb els models mono- i bi-exponencials per a estudiar la caiguda de la senyal en difusió amb el objectiu de estudiar la cel·lularització. Normalment, aquests biomarcadors d'imatge s'analitzen de forma "univariant", sense aprofitar la informació de las estructures de correlació interna que existeixen entre ells. Una forma de millorar aquest anàlisis es mitjançant la aplicació de las tècniques estadístiques aportades pel Anàlisis Multivariant de Imatges (MIA), obtenint estructures (latents) simplificades què ajuden a entendre la relació entre els paràmetres (variables) i els seus processos fisiològics, a més de reduir la incertesa en la estimació dels biomarcadors. En aquesta tesis, s'han desenvolupat nous biomarcadors d'imatge per a perfusió i difusió amb la aplicació de alguna de las ferramentes de MIA com la Resolució Multivariant de Corbes i Mínims Quadrats Alterns (MCR-ALS), obtenint paràmetres què tenen interpretació clínica directa. A continuació, els mètodes basats en mínims quadrats parcials (PLS) s'han aplicat per a estudiar la capacitat de classificació d'aquests biomarcadors. En primer lloc, els biomarcadors de perfusió s'han utilitzat per a la detecció de tumors (control contra lesió). Posteriorment, la combinació de perfusió + difusió + T2 s'ha utilitzat per a estudiar agressivitat tumoral amb la aplicació de mètodes PLS multi-bloc, en concret (seqüencial) SMB-PLS. Els resultats mostren què els biomarcadors de perfusió obtinguts mitjançant MCR són millors què els paràmetres farmacocinètics en la diferenciació de la lesió. En lo què es refereix al estudi de la agressivitat tumoral, la combinació dels biomarcadors de difusió (utilitzant els dos mètodes: models paramètrics i MCR) i els valors de T2w normalitzats proporcionaren els millors resultats. En conclusió, MIA es pot aplicar a les seqüències morfològiques i funcionals de ressonància magnètica per a millorar el diagnòstic i el estudi de l'agressivitat dels tumors en pròstata. Obtenint nous paràmetres quantitatius y combinant-los amb els biomarcadors més utilitzats en el ambient clínic. / [EN] The increase in life expectancy and population with age higher than 50 years is producing a major number of detected cases of prostate cancer (PCa). For this reason, the resources are focused in the early diagnosis and effective treatment. In spite of multiple studies with histologic discriminant biomarkers, it is hard to clearly differentiate the low aggressiveness PCa cases from those that will progress and produce mortality or rather a decrease in the life expectancy. With the objective of improving the diagnosis, location and gradation of the malignant tumors, Magnetic Resonance Imaging (MRI) has come up as the most appropriate image acquisition technique for cancer studies, which provides a non-invasive, sensitive and specific diagnosis, based on morphological and functional (blood perfusion and water diffusion) sequences. The different characteristics and parameters extracted from these sequences, known as imaging biomarkers, can evaluate the different processes associated to tumor development, like pharmacokinetic modeling for angiogenesis assessment (perfusion) or mono- and bi-exponential signal decay modeling for cellularization (diffusion). Normally, these imaging biomarkers are analyzed in a "univariate" way, without taking advantage of the internal correlation structures among them. One way to improve this analysis is by applying Multivariate Image Analysis (MIA) statistical techniques, obtaining simplified (latent) structures that help to understand the relation between parameters (variables) and the inner physiological processes, moreover reducing the uncertainty in the estimation of the biomarkers. In this thesis, new imaging biomarkers are developed for perfusion and diffusion by applying MIA tools like Multivariate Curve Resolution Alternating Least Squares (MCR-ALS), obtaining parameters with direct clinical interpretation. Partial Least Squares (PLS) based methods are then used for studying the classification capability of these biomarkers. First, perfusion imaging biomarkers have been tested for tumor detection (control vs lesion). Then, diffusion + perfusion have been combined to study tumor aggressiveness by applying PLS-multiblock methods (SMB-PLS). The results showed that MCR-based perfusion biomarkers performed better than state-of-the-art pharmacokinetic parameters for lesion differentiation. Regarding the assessment of tumor aggressiveness, the combination of diffusion-based imaging biomarkers (using both the parametric models and MCR) and normalized T2-weighted measurements provided the best discriminating outcome, while perfusion was not needed as it did not supply additional information. In conclusion, MIA can be applied to morphologic and functional MRI to improve the diagnosis and aggressiveness assessment of prostate tumors by obtaining new quantitative parameters and combining them with state-of-the-art imaging biomarkers. / Aguado Sarrió, E. (2019). Application of multivariate image analysis to prostate cancer for improving the comprehension of the related physiological phenomena and the development and validation of new imaging biomarkers [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/134023

Through the use of multiple methodologies and analytical approaches, this dissertation combines (1) sales control; (2) call center service; and (3) motivational theory to extend sales control literature beyond its current state, to consider the conflicting motivational perspectives an inside sales agent has to experience. To achieve this unification, this dissertation consists of three essays intended to: (1) identify the influence of autonomous and controlled motivation on operational sales outcome controls and performance; (2) explore the influence these motivators have on sales controls and sales performance; and, (3) understand the impact of autonomous and controlled motivation on sales agent tenure.

The purpose of this research was to identify key determinants of service provider effectiveness and how it impacts outsourced security success. As environments have become more robust and dynamic, many organizations have made the decision to leverage external security expertise and have outsourced many of their information technology security functions to Managed Security Service Providers (MSSPs). Information Systems Outsourcing, at its core, is when a customer chooses to outsource certain information technology functions or services to a service provider and engages in a legally binding agreement. While legal contracts govern many aspects of an outsourcing arrangement, it cannot serve as the sole source of determining the outcome of a project. Organizations are viewing outsourcing success as an attainment of net benefits achieved through the use of a service provider. The effectiveness of the service provider has an impact on a company’s ability to meet business objectives and adhere to service level agreements. Many empirical studies have focused on outsourcing success, but few have focused on service provider effectiveness, which can serve as a catalyst to outsourcing success. For this research, Agency Theory (AT) was proposed as a foundation for developing the research model which included key areas of focus in information asymmetry, the outsourcing contract, moral hazard, trust, service provider effectiveness, and security outsourcing success. Agency Theory helped uncover several hypotheses deemed germane to service provider effectiveness and provided insight into helping understand the principal-agent paradigm that exists with security outsourcing. Confirmatory Factor Analysis (CFA) and Partial Least Squares-Structured Equation Modeling (PLS-SEM) were used with SmartPLS to analyze the data and provided clarity and validation for the research model and helped uncover key determinants of service provider effectiveness. The statistical results showed support for information asymmetry, contract, and trust, all of which were mediated through service provider effectiveness. The results also showed that service provider effectiveness is directly correlated to increasing security outsourcing success. This concluded that the research model showed significant results to support 4 of the 5 hypotheses proposed and helped uncover key findings on how security outsourcing success can be impacted. This research served as an original contribution to information security while viewing outsourcing success from the perspective of the client, security services, and customer expectations.

Ce travail de recherche propose de nouveaux outils pour améliorer la prise en compte des caractéristiques spécifiques des hedge funds, dans l'évaluation de leur risque et de leur performance. Tout d'abord, nous mettons en évidence l'intérêt des développements basés sur la théorie des valeurs extrêmes pour analyser et quantifier le risque extrême des hedge funds. Une procédure de backtesting démontre que la valeur en risque, estimée à partir de la distribution de Pareto généralisée s'ajustant aux pertes extrêmes (VaREVT), est plus fiable que les mesures de risque usuelles. Puis, nous suggérons un nouvel indicateur de performance, lequel permet de prendre en compte la non normalité des distributions de rentabilités des hedge funds ainsi que, le niveau de rentabilité minimum acceptable de l'investisseur. Enfin, quatre modèles ont été construits en vue de déterminer les principaux facteurs explicatifs de l'évolution de la rentabilité journalière des stratégies alternatives. Ce dernier point donne l'occasion de mettre en évidence les avantages de la méthode de régression PLS pour identifier les facteurs pertinents. Cette recherche offre, non seulement, des résultats intéressants pour mieux comprendre le monde des hedge funds mais également, de nouvelles perspectives pour l'évaluation du risque et de la performance des autres actifs financiers ayant une distribution de rentabilités leptokurtique et asymétrique.

Les modèles d'équations structurelles à variables latentes constituent des modèles statistiques complexes permettant de mettre en relation des concepts non observables. Les deux méthodes d'estimation de ces modèles, que sont, d'une part, l'analyse de la structure de covariance (méthode LISREL) et, d'autre part, l'approche PLS, offrent des solutions à la fois concurrentes et complémentaires. Dans ce travail, un certain nombre de questionnements liés à ce type de modèles et de méthodes sont abordés aussi bien d'un point de vue théorique qu'empirique. Nous étudions la construction du modèle initial par le biais d'algorithmes itératifs, au niveau des relations du modèle de mesure et du modèle structurel. Nous présentons des tests basés sur des permutations afin de comparer des échantillons non appariés sur un modèle donné. Des transformations dites optimales des variables sont utilisées afin de rechercher des relations non linéaires. Finalement, des méthodes permettant le traitement de données manquantes spécifiques induites par des filtres sont décrites. Pour chaque cas, la théorie existante est présentée et des approches novatrices sont introduites. Nous appliquons l'ensemble de ces méthodes dans le cadre de l'analyse de la satisfaction et de la fidélité des clients sur des données provenant d'Electricité de France.

Lean management is becoming the standard for systematic productivity improvement, but the majority of implementations fail to sustain. Hence, the critical success factors for lean were the focus of this work. Literature review showed that the causality for lean success was not empirically developed beyond case study contextualisation. A multifaceted work was developed with contextualisation studies, survey of lean knowledge (758 responses), and a comprehensive case-study questionnaire (1253 responses from 44 countries). The statistical methods included exploratory factor analysis and path analysis by structural equation modelling (SEM). The first questionnaire revealed two different understandings of lean, and the second explored the underlying causality for lean success, including contingency for business size and product variety. Many contributions to the body of knowledge issued from this work. First of all, there was a methodological contribution, pioneering explorative structural modelling of full scope lean implementation. Second, SEMs of the lean knowledge-based view showed the profound positive effects of management knowledge on the primary factors for lean success. These factors were shown to be leadership and employee development. Third, the most beneficial lean methods were highlighted for specific scenarios. Fourth, the negligible and negative effects of a consultant-based approach to lean were uncovered. The results showed that the majority of consultants did not aid the long-term performance and sustainability of lean but significantly hindered it, except where masterful consultants acted as coaches. Fifth, a shortage of lean knowledge was observed in New Zealand; their participants averaged only half of what the USA�s did. Sixth, as culture has been emphasised in current literature, the present danger of overly focusing on it was discussed. Seventh was a conceptual contribution integrating lean and risk management, and a practical application with a risk analysis. This developed a risk matrix for the assessment and prioritisation of implementation components. Eighth, some adjustments to government lean strategies were proposed. And finally, the work integrated the findings in a tangible stage process model for implementation in SMEs. The dissemination of this knowledge has the potential to enhance productivity and commercial success of industries in New Zealand and abroad through successful lean implementations. Lean is not a weak methodology but it has been misunderstood and misapplied.

Natural archives such as lake sediments allow us to assess contemporary ecosystem responses to climate and environmental changes in a long-term context beyond the few decades to at most few centuries covered by monitoring or historical data. To achieve a comprehensive view of the changes preserved in sediment records, multi-proxy studies – ideally in high resolution – are necessary. However, this combination of including a range of analyses and high resolution constrains the amount of material available for analyses and increases the analytical costs. Infrared spectroscopic methods are a cost-efficient alternative to conventional methods because they offer a) a simple sample pre-treatment, b) a rapid measurement time, c) the non- or minimal consumption of sample material, and d) the potential to extract quantitative and qualitative information about organic and inorganic sediment components from a single measurement. The main objective of this doctoral thesis was twofold. The first part was to further explore the potential of Fourier transform infrared (FTIR) and visible-near infrared (VNIR) spectroscopy in paleolimnological studies as a) an alternative tool to conventional methods for quantifying biogenic silica (bSi) – a common proxy of paleoproductivity in lakes – in sediments and b) as a tool to infer past lake-water total organic carbon (TOC) levels from sediments. In a methodological study, I developed an independent application of FTIR spectroscopy and PLS modeling for determining bSi in sediments by using synthetic sediment mixtures with known bSi content. In contrast to previous models, this model is independent from conventional wet-chemical techniques, which had thus far been used as the calibration reference, and their inherent measurement uncertainties. The second part of the research was to apply these techniques as part of three multi-proxy studies aiming to a) improve our understanding of long-term element cycling in boreal and arctic landscapes in response to climatic and environmental changes, and b) to assess ongoing changes, particularly in lake-water TOC, on a centennial to millennial time scale. In the first applied study, high-resolution FTIR measurements of the 318-m long sediment record of Lake El’gygytgyn provided a detailed insight into long-term climate variability in the Siberian Arctic over the past 3.6 million years. Highest bSi accumulation occurred during the warm middle Pliocene (3.6-3.3 Ma), followed by a gradual but variable decline, which reflects the first onset of glacial periods and then the finally full establishment of glacial–interglacial cycles during the Quaternary. The second applied study investigated the sediment record of Torneträsk in subarctic northern Sweden also in relation to climate change, but only over the recent post-glacial period (~10 ka). By comparing responses to past climatic and environmental forcings that were recorded in this large-lake system with those recorded in small lakes from its catchment, I determined the significance and magnitude of larger-scale changes across the study region. Three different types of response were identified over the Holocene: i) a gradual response to the early landscape development following deglaciation (~10000-5300 cal yr BP); ii) an abrupt but delayed response following climate cooling during the late Holocene, which occurred c. 1300 cal yr BP – about 1000-2000 years later than in smaller lakes from the area; and iii) an immediate response to the ongoing climate change during the past century. The rapid, recent response in a previously rather insensitive lake-ecosystem emphasizes the unprecedented scale of ongoing climate change in northern Fennoscandia. In the third applied study, VNIR-inferred lake-water TOC concentrations from lakes across central Sweden showed that the ongoing, observed increase in surface water TOC in this region was in fact preceded by a long-term decline beginning already AD 1450-1600. These dynamics coincided with early human land use activities in the form of widespread summer forest grazing and farming that ceased over the past century. The results of this study show the strong impact of past human activities on past as well as ongoing TOC levels in surface waters, which has thus far been underestimated. The research in this thesis demonstrates that infrared spectroscopic methods can be an essential component in high-resolution, multi-proxy studies of past environmental and climate changes.

Activated carbon is an adsorbent that is commonly used for removing organic contaminants from air due to its abundant pores and large internal surface area. This thesis is concerned with the static adsorption capacity and adsorption kinetics for single and binary organic compounds on different types of activated carbon. These are important parameters for the design of filters and for the estimation of filter service life. Existing predictive models for adsorption capacity and kinetics are based on fundamental “hard” knowledge of adsorption mechanisms. These models have several drawbacks, especially in complex situations, and extensive experimental data are often needed as inputs. In this work we present a systematic approach that can contribute to the further development of predictive models, especially for complex situations. The approach is based on Multivariate Data Analysis (MVDA), which is ideally suited for the development of soft models without incorporating any assumptions about the mathematical form or fundamental physical principles involved. Adsorption capacity and adsorption kinetics depend on the properties of the carbon and the adsorbate as well as experimental conditions. Therefore, to make general statements regarding adsorption capacity and kinetics it is important for the resulting models to be representative of the conditions they will simulate. Accordingly, the first step in the investigations underlying this thesis was to select a minimum number of representative and chemically diverse organic compounds. The next steps were to study the dependence of the derived affinity coefficient, β, in the Dubinin-Radushkevich equation on properties of organic compounds and to establish a new, improved model. This new model demonstrates the importance of adding descriptors for the specific interaction with the carbon surface to the size and shape descriptors. The adsorption capacities of the same eight organic compounds at low relative pressures were correlated with compound properties. It was found that different compound properties are important in the various stages of adsorption, reflecting the fact that different mechanisms are involved. Ideal adsorbed solution theory (IAST) in combination with the Freundlich equation was developed to predict the adsorption capacities of binary organic compound mixtures. A new model was proposed for predicting the rate coefficient of the Wheeler-Jonas equation which is valid for breakthrough ratios up to 20%. Finally, it was shown that the Wheeler-Jonas equation can be adapted to describe the breakthrough curves of binary mixtures. New models were proposed for predicting its parameters, the adsorption rate coefficients, and the adsorption capacities for both components of the binary mixture. Thus, multivariate data analysis can not only be used to assist in the understanding of adsorption mechanisms, but also contribute to the development of predictive models of adsorption capacity and breakthrough time for single and binary organic compounds.

Supply chains are increasingly incorporating sustainable elements as mean to meet intensified TBL awareness in the market and to promote competitiveness. Suppliers play a key role in enhancing Triple Bottom Line (TBL) outcomes in supply chain since they are responsible for raw materials that will be incorporated in the final product. Supplier Development (SD) can improve suppliers\' TBL capabilities, improving TBL outcomes throughout the supply chain. This research aims to explore Triple Bottom Line Supplier Development (TBL SD) for an improved TBL performance at supply chain level. To address the research general purpose a systematic literature review was conducted, followed by a basic literature review, in order to identify TBL SD practices, enablers and barriers. The systematic literature analysis was performed with support of QDA Miner software. A survey was performed involving Brazilian buying organizations to investigate the impact of TBL Supplier Development on the organizations\' performance at the operational, environmental and social dimensions. Data were analyzed using Partial Least Square Structural Equation Modeling (PLS SEM) through Smart PLS 3 software. The results obtained point that practices adopted by organizations are in accordance to what literature suggests, highlighting the role of training and resources sharing as important practices adopted. The most impactful enablers to enable TBL SD are supplier evaluation, resources availability, and TBL culture at organizational level. Surprisingly, barriers appeared to be only slightly negatively correlated to TBL SD Adoption. It can be concluded that TBL SD adoption leads to improvements in all TBL legs (Operational, Environmental, and Social) although the impact on environmental performance is stronger. In turn, it was found that enhanced environmental performance can lead to improvements in both social and operational performance. Another interesting finding is that organizations in the Brazilian context are adopting TBL SD for internal reasons, and not due to pressures coming from the local market and community. / As cadeias de suprimentos estão incorporando cada vez mais elementos sustentáveis como meio de atender à crescente conscientização do mercado e promover competitividade. Fornecedores desempenham um papel fundamental no aprimoramento dos resultados sustentáveis na cadeia de fornecimento, uma vez que são responsáveis pelas matérias-primas que serão incorporadas no produto final. Nesse contexto, o desenvolvimento de fornecedores pode melhorar as capabilidades sustentáveis dos fornecedores, entregando melhores resultados a toda a cadeia de suprimentos. Esta pesquisa tem como objetivo explorar o Desenvolvimento de Fornecedores Sustentáveis para um melhor desempenho sustentável no nível da cadeia de suprimentos. Para atingir o objetivo geral da pesquisa, foi realizada uma revisão sistemática de literatura, seguida de uma revisão básica da literatura, a fim de identificar as práticas de desenvolvimento sustentáveis, fatores influenciadores e barreiras. A análise da revisão sistemática da literatura foi realizada com o apoio do software QDA Miner. Uma survey envolvendo organizações brasileiras foi realizada para investigar o impacto da adoção do Desenvolvimento de Fornecedores Sustentáveis no desempenho das organizações pesquisadas, tanto nos aspectos operacionais, ambientais e sociais. Os dados foram analisados usando a Modelagem de Equações Estruturais por Mínimos Quadrados Parciais através do software Smart PLS 3. Os resultados obtidos apontam que as práticas adotadas pelas organizações estão de acordo com o que sugere a literatura, destacando o papel do treinamento e compartilhamento de recursos como práticas importantes adotadas. Os fatores influenciadores mais impactantes para a adoção do Desenvolvimento de Fornecedores Sustentáveis são avaliação de fornecedores, a disponibilidade de recursos e a cultura sustentável no nível organizacional. Surpreendentemente, as barreiras parecem estar ligeiramente correlacionadas de forma negativa com adoção do Desenvolvimento de Fornecedores Sustentáveis. Pode-se concluir que a adoção do Desenvolvimento de Fornecedores Sustentáveis leva a melhorias em todas as dimensões da Sustentabilidade (Operacional, Ambiental e Social), embora o impacto no desempenho ambiental seja mais forte. Por sua vez, o desempenho ambiental pode levar a melhorias no desempenho social e operacional. Outra constatação interessante é que as organizações no contexto brasileiro estão adotando o Desenvolvimento de Fornecedores Sustentáveis por motivos internos, e não devido a pressões vindas do mercado e da comunidade local.

Extracting structure from initially incomprehensible streams of events is fundamental to a range of human abilities: from navigating in a new environment to learning a language. These skills rely on our ability to extract spatial and temporal regularities, often with minimal explicit feedback, that is known as statistical learning. Despite the importance of statistical learning for making perceptual decisions, we know surprisingly little about the brain circuits and how they change when learning temporal regularities. In my thesis, I combine behavioural measurements, Diffusion Tensor Imaging (DTI) and resting-state fMRI (rs-fMRI) to investigate the structural and functional circuits that are involved in statistical learning of temporal structures. In particular, I compare structural connectivity as measured by DTI and functional connectivity as measured by rs-fMRI before vs. after training to investigate learning-dependent changes in human brain pathways. Further, I combine the two imaging modalities using graph theory and regression analyses to identify key predictors of individual learning performance. Using a prediction task in the context of sequence learning without explicit feedback, I demonstrate that individuals adapt to the environment’s statistics as they change over time from simple repetition to probabilistic combinations. Importantly, I show that learning of temporal structures relates to decision strategy that varies among individuals between two prototypical distributions: matching the exact sequence statistics or selecting the most probable outcome in a given context (i.e. maximising). Further, combining DTI and rs-fMRI, I show that learning-dependent plasticity in dissociable cortico-striatal circuits relates to decision strategy. In particular, matching relates to connectivity between visual cortex, hippocampus and caudate, while maximisation relates to connectivity between frontal and motor cortices and striatum. These findings have potential translational applications, as alternate brain routes may be re-trained to support learning ability when specific pathways (e.g. memory-related circuits) are compromised by age or disease.