Dissertations / Theses on the topic 'Brain – Models'

In this thesis the mammalian nervous system and mammalian brain have been used as inspiration to develop a computational intelligence model based on the neural structure of fear conditioning and to extend the structure of the previous proposed amygdala-orbitofrontal model. The proposed model can be seen as a framework for developing general computational intelligence based on the emotional system instead of traditional models on the rational system of the human brain. The suggested model can be considered a new data driven model and is referred to as the brain emotional learning-inspired model (BELIM). Structurally, a BELIM consists of four main parts to mimic those parts of the brain’s emotional system that are responsible for activating the fear response. In this thesis the model is initially investigated for prediction and classification. The performance has been evaluated using various benchmark data sets from prediction applications, e.g. sunspot numbers from solar activity prediction, auroral electroject (AE) index from geomagnetic storms prediction and Henon map, Lorenz time series. In most of these cases, the model was tested for both long-term and short-term prediction. The performance of BELIM has also been evaluated for classification, by classifying binary and multiclass benchmark data sets.

Thesis (Ph. D.)--West Virginia University, 2000.
Title from document title page. Document formatted into pages; contains x, 133 p. : ill. Vita. Includes abstract. Includes bibliographical references (p. 122-130).

This dissertation focuses on the development of three classes of brain-inspired machine learning classification models. The models attempt to emulate (a) multi-sensory integration, (b) context-integration, and (c) visual information processing in the brain.The multi-sensory integration models are aimed at enhancing object classification through the integration of semantically congruent unimodal stimuli. Two multimodal classification models are introduced: the feature integrating (FI) model and the decision integrating (DI) model. The FI model, inspired by multisensory integration in the subcortical superior colliculus, combines unimodal features which are subsequently classified by a multimodal classifier. The DI model, inspired by integration in primary cortical areas, classifies unimodal stimuli independently using unimodal classifiers and classifies the combined decisions using a multimodal classifier. The multimodal classifier models are be implemented using multilayer perceptrons and multivariate statistical classifiers. Experiments involving the classification of noisy and attenuated auditory and visual representations of ten digits are designed to demonstrate the properties of the multimodal classifiers and to compare the performances of multimodal and unimodal classifiers. The experimental results show that the multimodal classification systems exhibit an important aspect of the “inverse effectiveness principle” by yielding significantly higher classification accuracies when compared with those of the unimodal classifiers. Furthermore, the flexibility offered by the generalized models enables the simulations and evaluations of various combinations of multimodal stimuli and classifiers under varying uncertainty conditions. The context-integrating model emulates the brain’s ability to use contextual information to uniquely resolve the interpretation of ambiguous stimuli. A deep learning neural network classification model that emulates this ability by integrating weighted bidirectional context into the classification process is introduced. The model, referred to as the CINET, is implemented using a convolution neural network (CNN), which is shown to be ideal for combining target and context stimuli and for extracting coupled target-context features. The CINET parameters can be manipulated to simulate congruent and incongruent context environments and to manipulate target-context stimuli relationships. The formulation of the CINET is quite general; consequently, it is not restricted to stimuli in any particular sensory modality nor to the dimensionality of the stimuli. A broad range of experiments are designed to demonstrate the effectiveness of the CINET in resolving ambiguous visual stimuli and in improving the classification of non-ambiguous visual stimuli in various contextual environments. The fact that the performance improves through the inclusion of context can be exploited to design robust brain-inspired machine learning algorithms. It is interesting to note that the CINET is a classification model that is inspired by a combination of brain’s ability to integrate contextual information and the CNN, which is inspired by the hierarchical processing of visual information in the visual cortex. A convolution neural network (CNN) model, inspired by the hierarchical processing of visual information in the brain, is introduced to fuse information from an ensemble of multi-axial sensors in order to classify strikes such as boxing punches and taekwondo kicks in combat sports. Although CNNs are not an obvious choice for non-array data nor for signals with non-linear variations, it will be shown that CNN models can effectively classify multi-axial multi-sensor signals. Experiments involving the classification of three-axis accelerometer and three-axes gyroscope signals measuring boxing punches and taekwondo kicks showed that the performance of the fusion classifiers were significantly superior to the uni-axial classifiers. Interestingly, the classification accuracies of the CNN fusion classifiers were significantly higher than those of the DTW fusion classifiers. Through training with representative signals and the local feature extraction property, the CNNs tend to be invariant to the latency shifts and non-linear variations. Moreover, by increasing the number of network layers and the training set, the CNN classifiers offer the potential for even better performance as well as the ability to handle a larger number of classes. Finally, due to the generalized formulations, the classifier models can be easily adapted to classify multi-dimensional signals of multiple sensors in various other applications.

Widely distributed brain networks display highly coherent activity at rest. In this work, we combined bottom-up and top-down approaches to investigate the dynamics and underlying mechanisms of this spontaneous activity. We developed a realistic network model to simulate resting-state data, which incorporates biophysical regional dynamics, empirical brain connectivity, time delays, and background noise. At moderately weak coupling strengths, the model produces spontaneous metastable oscillatory states and a novel form of frequency depression, resulting in transient synchronizations between brain regions at reduced collective frequencies. We used fixed and sliding window correlation approaches on the power of band-limited MEG data, and show that brain regions exhibit significant functional connectivity (FC) in the alpha and beta frequency bands on slow ( > 1sec) time scales. We also show that temporal non-stationarity and bistability in FC occur in the same pairs of brain areas, and in the same frequency bands, as stationary measures of FC. We find that the network model reproduces the same frequency-dependency, time-scales, and non-stationary nature of FC as we found in real MEG data. Furthermore, seed-based correlations and independent component analysis also reveal a similar spatial profile of FC in empirical and simulated data, with the existence of widely distributed transient resting state networks in the same frequency bands. Finally, we used the network model simulations to evaluate a range of network estimation methods, and find that often the simplest linear measures perform best and some of the common non-linear measures can often give erroneous estimates. Overall, our results suggest that structured interactions between brain regions in the presence of delays and noise result in spontaneous synchronizations leading to the organized power fluctuations across brain regions, and some of the simplest statistical measures provide excellent estimates of this connectivity. Our work also highlights the potential of computational models in exploring neural mechanisms.

La discipline encore naissante des réseaux complexes est vecteur d’un changement de paradigme dans la neuroscience. Les connectomes estimés à partir de mesures de neuroimagerie comme l’électroencéphalographie, la magnétoencéphalographie ou encore l’imagerie par résonance magnétique fonctionnelle fournissent une représentation abstraite du cerveau sous la forme d’un graphe, ce qui a permis des percées décisives dans la compréhension compacte et objective des propriétés topologiques et physiologiques des cerveaux sains. Cependant, les approches de pointe ignorent souvent l'incertitude et la nature temporelle de données de connectivité fonctionnelles. La plupart des méthodes disponibles dans la littérature ont en effet été développées pour caractériser les réseaux cérébraux fonctionnels comme des graphes statiques composés de nœuds (des régions cérébrales) et des liens (intensité de connectivité fonctionnelle) par métrique de réseau. En conséquence, la théorie des réseaux complexes a été principalement appliquée à des études transversales avec une unique mesure par sujet, produisant au final une caractérisation consistant en une moyenne de phénomènes neuronaux spatiotemporels. Nous avons implémenté des méthodes statistiques pour modéliser et simuler des réseaux cérébraux temporels. Nous avons utilisé des modèles de graphe qui permettent d'étudier simultanément à quel point les différentes propriétés des réseaux influencent la topologie observée dans les réseaux de connectivité cérébrale fonctionnelle. Nous avons identifié avec succès les mécanismes de connectivité locale fondamentaux qui gouvernent les propriétés des réseaux cérébraux. Nous avons proposé l'adaptation temporelle de ces mécanismes fondamentaux pour modéliser et simuler les changements physiologiques dynamiques d'un réseau cérébral. Plus spécifiquement, nous avons exploité des métriques temporelles pour construire des modèles temporels informatifs du rétablissement de patients ayant subit un accident vasculaire cérébral
The emerging area of complex networks has led to a paradigm shift in neuroscience. Connectomes estimated from neuroimaging techniques such as electroencephalography (EEG), magnetoencephalography (MEG) and functional magnetic resonance imaging (fMRI) results in an abstract representation of the brain as a graph, which has allowed a major breakthrough in the understanding of topological and physiological properties of healthy brains in a compact and objective way. However, state of the art approaches often ignore the uncertainty and temporal nature of functional connectivity data. Most of the available methods in the literature have been developed to characterize functional brain networks as static graphs composed of nodes (brain regions) and links (FC intensity) by network metrics. As a consequence, complex networks theory has been mainly applied to cross-sectional studies referring to a single point in time and the resulting characterization ultimately represents an average across spatiotemporal neural phenomena. Here, we implemented statistical methods to model and simulate temporal brain networks. We used graph models that allow to simultaneously study how different network properties influence the emergent topology observed in functional connectivity brain networks. We successfully identified fundamental local connectivity mechanisms that govern properties of brain networks. We proposed a temporal adaptation of such fundamental connectivity mechanisms to model and simulate physiological brain network dynamic changes. Specifically, we exploited the temporal metrics to build informative temporal models of recovery of patients after stroke

We study the following well-established model of reaction-diffusion type for brain tumour growth: This equation describes the change over time of the normalised tumour cell density u as a consequence of two biological phenomena: proliferation and diffusion. We discuss a mathematical method for the inverse problem of locating the brain tumour source (origin) based on the reaction-diffusion model. Our approach consists in recovering the initial spatial distribution of the tumour cells  starting from a later state , which can be given by a medical image. We use the nonlinear Landweber regularization method to solve the inverse problem as a sequence of well-posed forward problems. We give full 3-dimensional simulations of the tumour in time on two types of data, the 3d Shepp-Logan phantom and an MRI T1-weighted brain scan from the Internet Brain Segmentation Repository (IBSR). These simulations are obtained using standard finite difference discretisation of the space and time-derivatives, generating a simplistic approach that performs well. We also give a variational formulation for the model to open the possibility of alternative derivations and modifications of the model. Simulations with synthetic images show the accuracy of our approach for locating brain tumour sources.

Models of whole-brain connectivity are valuable for understanding neurological function. This thesis seeks to develop an optimal framework for extracting models of whole-brain connectivity from clinically acquired diffusion data. We propose new approaches for studying these models. The aim is to develop techniques which can take models of brain connectivity and use them to identify biomarkers or phenotypes of disease. The models of connectivity are extracted using a standard probabilistic tractography algorithm, modified to assess the structural integrity of tracts, through estimates of white matter anisotropy. Connections are traced between 77 regions of interest, automatically extracted by label propagation from multiple brain atlases followed by classifier fusion. The estimates of tissue integrity for each tract are input as indices in 77x77 ”connectivity” matrices, extracted for large populations of clinical data. These are compared in subsequent studies. To date, most whole-brain connectivity studies have characterised population differences using graph theory techniques. However these can be limited in their ability to pinpoint the locations of differences in the underlying neural anatomy. Therefore, this thesis proposes new techniques. These include a spectral clustering approach for comparing population differences in the clustering properties of weighted brain networks. In addition, machine learning approaches are suggested for the first time. These are particularly advantageous as they allow classification of subjects and extraction of features which best represent the differences between groups. One limitation of the proposed approach is that errors propagate from segmentation and registration steps prior to tractography. This can cumulate in the assignment of false positive connections, where the contribution of these factors may vary across populations, causing the appearance of population differences where there are none. The final contribution of this thesis is therefore to develop a common co-ordinate space approach. This combines probabilistic models of voxel-wise diffusion for each subject into a single probabilistic model of diffusion for the population. This allows tractography to be performed only once, ensuring that there is one model of connectivity. Cross-subject differences can then be identified by mapping individual subjects’ anisotropy data to this model. The approach is used to compare populations separated by age and gender.

Functional magnetic resonance imaging (fMRI) is a non-invasive method for studying the human brain that is now widely used to study functional connectivity. Functional connectivity concerns how brain regions interact and how these interactions change over time, between subjects and in different experimental contexts and can provide deep insights into the underlying brain function. Multiregression Dynamic Models (MDMs) are dynamic Bayesian networks that describe contemporaneous, causal relationships between time series. They may therefore be applied to fMRI data to infer functional brain networks. This work focuses on the MDM Directed Graph Model (MDM-DGM) search algorithm for network discovery. The Log Predictive Likelihood (model evidence) factors by subject and by node, allowing a fast, parallelised model search. The estimated networks are directed and may contain the bidirectional edges and cycles that may be thought of as being representative of the true, reciprocal nature of brain connectivity. In Chapter 2, we use two datasets with 15 brain regions to demonstrate that the MDM-DGM can infer networks that are physiologically-interpretable. The estimated MDM-DGM networks are similar to networks estimated with the widely-used partial correlation method but advantageously also provide directional information. As the size of the model space prohibits an exhaustive search over networks with more than 20 nodes, in Chapter 3 we propose and evaluate stepwise model selection algorithms that reduce the number of models scored while optimising the networks. We show that computation time may be dramatically reduced for only a small trade-off in accuracy. In Chapter 4, we use non-local priors to derive new, closed-form expressions for the model evidence with a penalty on weaker, potentially spurious, edges. While the application of non-local priors poses a number of challenges, we argue that it has the potential to provide a flexible Bayesian framework to improve the robustness of the MDM-DGM networks.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2012.
Cataloged from PDF version of thesis.
Includes bibliographical references (p. 131-139).
Connectivity analysis focuses on the interaction between brain regions. Such relationships inform us about patterns of neural communication and may enhance our understanding of neurological disorders. This thesis proposes a generative framework that uses anatomical and functional connectivity information to find impairments within a clinical population. Anatomical connectivity is measured via Diffusion Weighted Imaging (DWI), and functional connectivity is assessed using resting-state functional Magnetic Resonance Imaging (fMRI). We first develop a probabilistic model to merge information from DWI tractography and resting-state fMRI correlations. Our formulation captures the interaction between hidden templates of anatomical and functional connectivity within the brain. We also present an intuitive extension to population studies and demonstrate that our model learns predictive differences between a control and a schizophrenia population. Furthermore, combining the two modalities yields better results than considering each one in isolation. Although our joint model identifies widespread connectivity patterns influenced by a neurological disorder, the results are difficult to interpret and integrate with our regioncentric knowledge of the brain. To alleviate this problem, we present a novel approach to identify regions associated with the disorder based on connectivity information. Specifically, we assume that impairments of the disorder localize to a small subset of brain regions, which we call disease foci, and affect neural communication to/from these regions. This allows us to aggregate pairwise connectivity changes into a region-based representation of the disease. Once again, we use a probabilistic formulation: latent variables specify a template organization of the brain, which we indirectly observe through resting-state fMRI correlations and DWI tractography. Our inference algorithm simultaneously identifies both the afflicted regions and the network of aberrant functional connectivity. Finally, we extend the region-based model to include multiple collections of foci, which we call disease clusters. Preliminary results suggest that as the number of clusters increases, the refined model explains progressively more of the functional differences between the populations.
by Archana Venkataraman.
Ph.D.

A Multiregression Dynamic Model (MDM) is a class of multivariate time series that represents multiple dynamic causal processes in a graphical way. One of the advantages of this class is that, in contrast to many other Dynamic Bayesian Networks, the hypothesised relationships accommodate conditional conjugate inference. We demonstrate for the first time how it is straightforward to search over all possible connectivity networks with dynamically changing intensity of transmission to find the Maximum a Posteriori Probability (MAP) model within this class. This search method is made feasible by using a novel application of the integer programming algorithm. The search over all possible directed (acyclic or cyclic) graphical structures can be made especially fast by utilising the fact that, within this class of models, the joint likelihood factorizes. We proceed to show how diagnostic methods, analogous to those defined for static Bayesian Networks, can be used to suggest embellishment of the model class to extend the process of model selection. A typical goal of experimental neuroscience is to draw conclusions regarding the causal mechanisms that underpin neural communication. Often the main focus of interest in these experiments includes not only a search for the likely model of a specific individual, but an analysis of shared between-subject e↵ects. Currently, such features are analysed using rather coarse aggregation methods over shared time series. However, here we demonstrate that, using the estimation of multiple causal graphical models and Bayesian hyperclustering techniques, it is possible to use the full machinery of Bayesian methods to formally make inferences in a coherent way which contemplates hypotheses about shared dependences between such populations of subjects. Methods developed here are illustrated using simulated and real resting-state and steady-state task functional Magnetic Resonance Imaging (fMRI) data.

La même entrée sensorielle ne provoque pas toujours la même réaction. Dans les expériences en laboratoire, un stimulus donné peut engendrer une réponse différente à chaque nouvel essai, en particulier à proximité du seuil sensoriel. Ce phénomène est généralement attribué à une source de bruit non spécifique qui affecte la représentation sensorielle du stimulus ou le processus décisionnel. Dans cette thèse, nous examinons l'hypothèse selon laquelle cette variabilité des réponses peut être attribuée en partie à des fluctuations mesurables et spontanées de l'état cérébral. Dans ce but, nous développons et évaluons deux ensembles d'outils. L’un est un ensemble de modèles et de méthodes psychophysiques permettant de suivre les variations de la performance perceptive avec une bonne résolution temporelle et avec précision, sur différentes échelles de temps. Ces méthodes s’appuient sur des procédures adaptatives initialement développées pour mesurer efficacement les seuils de perception statiques et sont étendues ici dans le but de suivre des seuils qui varient au cours du temps. Le deuxième ensemble d'outils que nous développons comprend des méthodes d'analyse de données pour extraire de signaux d’électroencéphalographie (EEG) une quantité prédictive de la performance comportementale à diverses échelles de temps. Nous avons appliqué ces outils à des enregistrements conjoints d’EEG et de données comportementales collectées pendant que des auditeurs normo-entendants réalisaient une tâche de discrimination de fréquence sur des stimuli auditifs proche du seuil de discrimination. Contrairement à ce qui a été rapporté dans la littérature concernant des stimuli visuels, nous n'avons pas trouvé de preuve d’un quelconque effet des oscillations EEG spontanées de basse fréquence sur la performance auditive. En revanche, nous avons trouvé qu'une part importante de la variabilité des jugements peut s’expliquer par des effets de l'historique récent des stimuli et des réponses sur la décision prise à un moment donné
The same sensory input does not always trigger the same reaction. In laboratory experiments, a given stimulus may elicit a different response on each trial, particularly near the sensory threshold. This is usually attributed to an unspecific source of noise that affects the sensory representation of the stimulus or the decision process. In this thesis we explore the hypothesis that response variability can in part be attributed to measurable, spontaneous fluctuations of ongoing brain state. For this purpose, we develop and test two sets of tools. One is a set of models and psychophysical methods to follow variations of perceptual performance with good temporal resolution and accuracy on different time scales. These methods rely on the adaptive procedures that were developed for the efficient measurements of static sensory thresholds and are extended here for the purpose of tracking time-varying thresholds. The second set of tools we develop encompass data analysis methods to extract from electroencephalography (EEG) signals a quantity that is predictive of behavioral performance on various time scales. We applied these tools to joint recordings of EEG and behavioral data acquired while normal listeners performed a frequency-discrimination task on near-threshold auditory stimuli. Unlike what was reported in the literature for visual stimuli, we did not find evidence for any effects of ongoing low-frequency EEG oscillations on auditory performance. However, we found that a substantial part of judgment variability can be accounted for by effects of recent stimulus-response history on an ongoing decision

Thesis: S.M. in Neuroscience, Massachusetts Institute of Technology, Department of Brain and Cognitive Sciences, 2015.
Cataloged from PDF version of thesis.
Includes bibliographical references (page 17).
This work establishes a benchmark by which to measure models of the ventral stream using crowd-sourced human behavioral measurements. We collected human error patterns on an object recognition task across a variety of images. By comparing the error pattern of these models to the error pattern of humans, we can measure how similar to the human behavior the model's behavior is. Each model we tested was composed of two parts: an encoding phase which translates images to features, and a decoding phase which translates features to a classifier decision. We measured the behavioral consistency of three encoder models: a convolutional neural network, and a particular view of neural activity of either are V4 or IT. We measured three decoder models: logistic regression and 2 different types of support vector machines. We found the most consistent error pattern to come from a combination of IT neurons and a logistic regression but found that this model performed far worse than humans. After accounting for performance, the only model that was not invalidated was a combination of IT neurons and an SVM.
by Diego Ardila.
S.M. in Neuroscience

Conselho Nacional de Desenvolvimento CientÃfico e TecnolÃgico
CoordenaÃÃo de AperfeiÃoamento de Pessoal de NÃvel Superior
The disabling effects of central nervous system (CNS) tumors are out of proportion to their low incidence. Theyâre second only to stroke as neurologic mortality causes. Brain metastases are the commonest intracranial tumors in adults, almost 10 times more frequent than primary brain tumors. Their diagnosis and treatment have met significant advances, although much more research about drug resistance and new treatment modalities are needed. New and even better brain tumor animal models will help to evaluate novel drug regimens and adjuvant therapies for CNS neoplasms. In the present work, the author presents a simple and easily reproducible brain tumor model utilizing the tumor cell line W256 transplanted to Wistar rats. They tested a drug widely used for palliative treatment of tumoral brain edema (dexamethasone), for survival impact. They also have tested the effects of a drug newly proposed as multidrug resistance reversal agent (cyclosporin â CS). Wistar albino rats had stereotaxic intracranial tumor inoculation after the surgical installation of a permanent canulla on the area of interest (right subfrontal caudate). The brain tumor model, as a model of metastatic brain disease, was successful, with reproducible tumor growth (95%), low incidence of extracranial tumor implantation (21% local, no distant metastasis) and few evidence of surgical site infection (21%). The median survival of the animals was 12.5 days (control), 13 days (CS vehicle treated), 11 days (CS treated), 9.5 and 9 days (dexamethasone 0.3 and 3.0 mg/kg/day). These differences were not significant, although the survival rates on the 12th day post-inoculation have showed a significant survival decrease for the case of dexamethasone 3,0 mg/kg/day (p < 0.05), but not for CS treatment (Fischerâs Exact Test). The estimated tumor volume was 17.08 Â 6.7 mm3 (control) and 12.61 Â 3.6 mm3 (CS treatment, not significant, Studentâs t-test). The tumor volume in the 9th day post-inoculation was estimated in 67,25 Â 19,8 mm3. The doubling time was 24.25 h. This model behaved as an undifferentiated tumor, with local invasiveness features compared with that of primary brain tumors. It fits well, in this way, for the study of tumor cell migration on CNS parenchyma. Phenomena like neuronal degeneration, neuron cell edema and death, and gliosis, as well as perivascular cell infiltrates, were seen frequently. One could find, also, neoangiogenesis, satellite tumor growth, and tumor cell migration in normal brain parenchyma. Besides heavy parenchymatous infiltration, it was also disclosed markedly tumor cell migration along white matter tracts, such as callosal fibers and infiltration in the Virchow-Robins perivascular space. The model presents as a dual brain tumor and leptomeningeal carcinomatosis model. It could be used for the study and treatment test in the scenario of these two pathologies. The intracerebral tumor growth induced peripheral blood neutrophil count elevation (ANOVA, p < 0.01), higher chance for neutrophilia (Fischerâs Exact Test, p < 0.01), higher chance for lymphopenia (Fischerâs Exact Test, p < 0.01) and brain weight increase (Studentâs t-test, p < 0.001) comparing to control. There was no significant change in any of the other hematologic, biochemical and biological parameters tested. CS treatment did not alter any of the tests, as compared to non-treated brain tumor animals. The only exception was the mean animal weight on the first week post-inoculation (ANOVA, p < 0.05). CS, in this way, was responsible for an early cachexia in the brain tumor inoculated animals. CS treatment of brain tumor animals did show non-significant effects indicating a volume (26%) and weight tumor decrease, and tumor infiltrating neutrophil increase (odds ratio - OR = 5.6). This indicates the necessity to further study morphologically and functionally the local inflammation in brain tumor inoculated animals, as well the effects of CS administration. In conclusion, the W256 intracerebral tumor model is simple, easily performed, reproducible and of great potential utility. In this model, tumor inoculation can lead to hematologic and biologic modifications in the experimental animals. CS could apparently lead to early tumor caquexia in this tumor model. However, CS treatment did not modify the survival chance of the brain tumor animals, in sharp contrast to dexamethasone 3.0mg/kg/day, a much-used drug in the treatment of brain tumors, which decreased the animal survival.
Os importantes efeitos incapacitantes dos tumores do sistema nervoso central (SNC) sÃo desproporcionais a sua baixa incidÃncia. Mesmo assim, entre as doenÃas neurolÃgicas, ficam atrÃs apenas dos acidentes vasculares do SNC como causa de morte. MetÃstases cerebrais constituem os tumores intracranianos mais comuns do adulto, ocorrendo atà 10 vezes mais freqÃentemente que tumores primÃrios. AvanÃos significativos ocorreram em seu diagnÃstico e tratamento, embora mais pesquisa sobre os fenÃmenos que diminuem o efeito de drogas em metÃstases cerebrais e tratamentos eficazes para estas patologias sejam cada vez mais necessÃrios. O desenvolvimento de melhores modelos animais de tumores do SNC serà necessÃrio para a avaliaÃÃo in vivo de novas formas de quimioterapia (QT) e terapia adjuvante para tumores cerebrais. No presente trabalho, o autor objetivou desenvolver um modelo de tumor cerebral simples e de fÃcil reproduÃÃo utilizando a linhagem W256, alÃm de testar o efeito na sobrevida animal de uma droga largamente usada para o tratamento de efeitos secundÃrios a edema cerebral (dexametasona). O autor tambÃm testou uma droga envolvida numa nova proposta de reversÃo de multirresistÃncia a drogas anti-neoplÃsicas em tumores cerebrais (ciclosporina â CS). Ratos albinos (Wistar) tiveram o tumor inoculado atravÃs de estereotaxia, apÃs a instalaÃÃo cirÃrgica de uma cÃnula no ponto escolhido (caudato subfrontal direito). O modelo de tumor implantado no cÃrebro de ratos, simulando uma metÃstase cerebral, mostrou-se bem sucedido e reprodutÃvel (95% de crescimento tumoral), com baixa incidÃncia de disseminaÃÃo tumoral extracraniana local (21%), baixa evidÃncia de infecÃÃo local (21%), ausÃncia de metÃstases à distÃncia e ausÃncia de sinais de infecÃÃo sistÃmica. Os animais sobreviveram uma mediana de 12,5 dias (grupo controle), 13 dias (tratados com veÃculo da CS), 11 dias (tratados com CS), 9,5 e 9 dias (dexametasona 0,3 e 3,0 mg/kg/dia, respectivamente). As diferenÃas entre estas medianas nÃo foram significantes (teste de Kruskal-Wallis), embora as diferenÃas entre as taxas de sobrevida no 12o dia apÃs a inoculaÃÃo tenham mostrado reduÃÃo significante no grupo que recebeu dexametasona 3,0 mg/kg/dia (p < 0,05), mas nÃo no grupo tratado com CS (teste de Fischer). O volume tumoral estimado (VTE) no sÃtimo dia pÃs-inoculaÃÃo (7DPI) foi de 17,08  6,7 mm3 no controle e 12,61 3,6 mm3 apÃs tratamento com CS, sem diferenÃa significante (teste t-Student). O VTE no 9DPI de animais do grupo Tumor foi de 67,25  19,8 mm3. O tempo de duplicaÃÃo foi de 24,25 h. O modelo comportou-se como um tumor de caracterÃsticas indiferenciadas, apresentando invasividade local comparada à de tumores primÃrios do SNC, prestando-se ao estudo da migraÃÃo de cÃlulas tumorais no SNC. Observaram-se fenÃmenos como degeneraÃÃo neuronal hidrÃpica, edema celular neuronal, sinais de morte celular neuronal e gliose, alÃm da presenÃa de infiltrados celulares tumorais e inflamatÃrios perivasculares. Observaram-se, tambÃm, neoformaÃÃo vascular, formaÃÃo de nÃdulos tumorais satÃlites ao tumor principal e migraÃÃo celular tumoral no parÃnquima cerebral normal. Observou-se, alÃm da infiltraÃÃo parenquimatosa, marcante migraÃÃo celular tumoral ao longo de tratos de substÃncia branca (corpo caloso) e ao longo dos espaÃos perivasculares de Virchow-Robins. O modelo apresenta-se como um misto de tumor cerebral intraparenquimatoso e carcinomatose leptomenÃngea, podendo ser utilizado para estudar o comportamento e testar formas de tratamento para ambas as patologias. O crescimento tumoral intracerebral induziu aumento do nÃmero de neutrÃfilos no sangue perifÃrico (ANOVA, p < 0,01), maior chance de apresentar neutrofilia (teste de Fischer, p < 0,01), maior chance de apresentar linfopenia (teste de Fischer, p < 0,01) e aumento do peso dos cÃrebros dos animais experimentais (teste t-Student, p < 0,001) em relaÃÃo ao controle. Nenhum dos outros valores hematolÃgicos, bioquÃmicos e biolÃgicos foi alterado de maneira significante. O tratamento de animais inoculados com tumor com a CS, nÃo alterou nenhuma das medidas hematolÃgicas, bioquÃmicas ou biolÃgicas em relaÃÃo aos animais inoculados com tumor e nÃo tratados, exceto o peso dos animais na primeira semana apÃs inoculaÃÃo tumoral (ANOVA, p < 0,05). A CS, dessa forma, induziu significantemente uma caquexia precoce nos animais inoculados com tumor cerebral. O tratamento com CS de animais inoculados com tumor mostrou tendÃncias nÃo significantes a diminuir volume (26%) e massa (7%) tumorais e aumentar nÃmero de neutrÃfilos infiltrantes de tumor (razÃo de chance - RC = 5,6) e necrose tumoral, indicando a necessidade de posteriores estudos para caracterizar morfolÃgica e funcionalmente a resposta inflamatÃria local em animais inoculados com tumor e a influÃncia da CS neste processo, alÃm do efeito da CS na angiogÃnese tumoral. Concluindo, o modelo de W256 intracerebral mostrou-se simples, de fÃcil execuÃÃo, reprodutÃvel e Ãtil. Neste modelo, a inoculaÃÃo tumoral induz modificaÃÃes hematolÃgicas e biolÃgicas nos animais. A CS pareceu exarcebar a caquexia tumoral neste modelo. A CS, todavia, nÃo alterou a chance de sobrevida de animais inoculados com tumor cerebral, ao contrÃrio da dexametasona 3,0 mg/kg/dia, que reduziu esta chance. A CS, assim, parece ser mais segura neste modelo tumoral que uma droga largamente utilizada para tratamento de pacientes com metÃstase cerebral.

Two brain segmentation approaches based on Hidden Markov Models are proposed. The first approach aims to segment normal brain 3D multi-channel MR images into three tissues WM, GM, and CSF. Linear Discriminant Analysis, LDA, is applied to separate voxels belonging to different tissues as well as to reduce their features vector size. The second approach aims to detect MS lesions in Brain 3D multi-channel MR images and to label WM, GM, and CSF tissues. Preprocessing is applied in both approaches to reduce the noise level and to address sudden intensity and global intensity correction. The proposed techniques are tested using 3D images from Montereal BrainWeb data set. In the first approach, the results were numerically assessed and compared to results reported using techniques based on single channel data and applied to the same data sets. The results obtained using the multi channel HMM-based algorithm were better than the results reported for single channel data in terms of an objective measure of overlap, Dice coefficient, compared to other methods. In the second approach, the segmentation accuracy is measured using Dice coefficient and total lesions load percentage

This work is about understanding the dynamics of neuronal systems, in particular with respect to brain connectivity. It addresses complex neuronal systems by looking at neuronal interactions and their causal relations. These systems are characterized using a generic approach to dynamical system analysis of brain signals - dynamic causal modelling (DCM). DCM is a technique for inferring directed connectivity among brain regions, which distinguishes between a neuronal and an observation level. DCM is a natural extension of the convolution models used in the standard analysis of neuroimaging data. This thesis develops biologically constrained and plausible models, informed by anatomic and physiological principles. Within this framework, it uses mathematical formalisms of neural mass, mean-field and ensemble dynamic causal models as generative models for observed neuronal activity. These models allow for the evaluation of intrinsic neuronal connections and high-order statistics of neuronal states, using Bayesian estimation and inference. Critically it employs Bayesian model selection (BMS) to discover the best among several equally plausible models. In the first part of this thesis, a two-state DCM for functional magnetic resonance imaging (fMRI) is described, where each region can model selective changes in both extrinsic and intrinsic connectivity. The second part is concerned with how the sigmoid activation function of neural-mass models (NMM) can be understood in terms of the variance or dispersion of neuronal states. The third part presents a mean-field model (MFM) for neuronal dynamics as observed with magneto- and electroencephalographic data (M/EEG). In the final part, the MFM is used as a generative model in a DCM for M/EEG and compared to the NMM using Bayesian model selection.

Duchenne muscular dystrophy (DMD) is an X-linked recessive muscle wasting disease caused by mutations in the DMD gene, which encodes the large cytoskeletal protein dystrophin. Alongside severe muscle pathology, one-third of DMD patients exhibit cognitive problems ranging from reduced verbal intelligence to severe autism. There is conclusive evidence that the muscle pathology exhibited by DMD patients is progressive, yet it remains unknown whether the cognitive impairments in DMD are also progressive. Previous studies have highlighted a cognitive impairment in the mdx mouse model of DMD, but no studies have investigated if this cognitive impairment worsens with age. We assessed the consequences of dystrophin deficiency on brain morphology and cognitive function in two dystrophin-deficient mouse models (mdx and Cmah-/-mdx mice). The overall project aim was to identify outcome measures to monitor central nervous system (CNS) pathology non-invasively in DMD mice. Magnetic resonance imaging (MRI) identified a total brain volume increase in DMD mice, alongside morphological changes in brain ventricles. Behavioural testing revealed a deficit in hippocampal spatial learning and memory, particularly long-term memory, in mdx mice, which appears to progressively worsen with age. Immunoblotting identified a progressive reduction of aquaporin-4 (AQP4) expression, the major water channel of the CNS, in DMD mice. Moreover, contrast enhancing MRI and Evans blue extravasation demonstrated a progressive impairment in blood-brain barrier (BBB) integrity in mdx mice. Proteomic profiling of the mdx cerebellum identified changes in expression of mitochondrial subunit complexes, suggestive of changes in mitochondrial function. Additionally, elevated levels of inflammatory markers were identified and confirmed in the mdx cerebellum. Our studies suggest that dystrophin deficiency causes a progressive cognitive impairment in mdx mice. We also present evidence showing that changes in osmotic equilibrium may be involved in the pathogenesis of DMD, with reductions in AQP4 expression and BBB disruptions. We speculate that some of the changes in the mdx cerebellar proteome, in comparison to wild type mice, iii serve as compensatory mechanisms whilst others may contribute directly to cognitive dysfunction in DMD. These results support a role for dystrophin in normal brain morphology and cognitive function.

Aim: The general aim of this research is development and evaluation of novel methods for predicting active transport over the human blood-brain-barrier (BBB), while this project specifically aims to i) review the literature and select suitable methods and substrates, ii) develop models for determining in vitro kinetic properties of selected compounds, analyze the in vitro data using the developed models and to use Maximum Transport Activity (MTA) approach (Karlgren et al., 2012), iii) perform Physiology Based Pharmacokinetic (PBPK) modelling and compare to in vivo literature data. Background: Drug permeation to the brain through blood circulation is primarily limited by blood-brain barrier (BBB), due to existence of tight junctions in endothelial cells of blood vessels as well as active efflux and influx transporters in the barrier. Toxicity and CNS related side effects can be caused by peripheral targeted drugs crossing BBB. Hence, prediction of BBB permeability and estimation of drug concentration in the brain tissue are challenging in drug discovery. To resolve this, estimating the human BBB permeability using improved in vitro and in silico predictive models can be a facilitator. Methods: In vitro data provided by the Drug Delivery research group was used to develop in vitro predictive models for BBB penetration of Verapamil, Risperidone, and Prazosin using R-studio 1.2.5. The MTA approach was adjusted for extrapolation of BBB in vitro transporter activity to in vivo condition. For PBPK modelling, we took advantage of PK-Sim® to simulate drug disposition and time profile of Risperidone in human and animal species. Results: It was shown that MDR1 is the major transporter for efflux transport of Prazosin and Risperidone in brain while both BCRP and MDR1 have similar impact on transport of Verapamil. Furthermore, it was presented in PBPK models that the predicted brain concentration of Risperidone increases in rat and nonhuman primate (NHP) when MDR1 And BCRP are knocked out while the brain concentration of Risperidone in dog is not affected by expression level of the efflux transporters. Conclusion: Both MDR1 and BCRP are contributing in efflux transport of Verapamil, Risperidone, and Prazosin across the BBB. Additionally, expression of the efflux transporters shown to have an impact on brain exposure of Risperidone in animal PBPK models.

This PhD thesis is focused in the application of brain network models obtained from different modalities of magnetic resonance imaging (MRI) to characterize anomalies in neurodevelopment of a prenatal origin, using intrauterine growth restriction (IUGR) as a clinical model. Importantly, IUGR due to placental insufficiency affects 5-10% of all pregnancies and is a leading cause of fetal morbidity and mortality. The thesis is presented as a compendium of four studies published in international journals. Each of the studies is focused in the characterization of IUGR using brain networks obtained from a specific MRI modality (structural, diffusion and functional MRI) in a specific pediatric stage in the life of subjects with IUGR (neonatal, early infancy and pre-adolescence age). The first study focuses in the characterization of brain reorganization produced by IUGR at one year of age using brain networks based on a tractography obtain from diffusion MRI, using diffusion tensor imaging (DTI) approach. In this study it is demonstrated that brain network features of IUGR infants have alterations associated with an altered neurodevelopment later in life. The second study assess the viability to use a novel methodology to obtain structural brain networks based on simple anatomical MRI based on the similitude of gray matter (GM) patterns among different areas of the brain. We demonstrated alterations in infants with IUGR using this technique, and that the alterations found are also associated with neurodevelopmental problems found later in life. In the third study we used a rabbit model of IUGR to explore if the alterations in the structural brain network persist at long-term, during preadolescence. We demonstrated that indeed, there are alterations in the structural brain network organization that persist at long-term and that this alterations are associated with neurobehavioral outcomes. Finally, using normalization approaches, we observed a peculiar compensatory effect in the subjects with IUGR. In the forth study, we assessed functional brain networks of neonates with IUGR, demonstrating that this condition produces a reorganization of functional brain connectivity since such an early age, characterized by a pattern of increased co-activation and synchronization of brain regions together with a suboptimal organization when assessing normalized networks. In addition, functional brain network features were also associated to neurobehavioral alterations. Overall, our main conclusion is that IUGR condition produces structural and functional brain reorganization since early life that persists postnatally up to pre-adolescence. We hypothesize that the observed functional and structural reorganization could be a potential substrate of high risk of altered neurodevelopment in infants with IUGR, and postulate this condition as a possible brain network disorder. Importantly, the association of network features with neurobehavior and neurodevelopment since an early age opens the opportunity to further develop early image biomarkers of altered neurodevelopment, a clinical chance to improve the management of a condition that affects up to 10% of deliveries.
Aquesta tesis doctoral està centrada en l'aplicació de models de xarxa del cervell obtinguts a partir de diferents modalitats de ressonància magnètica (RM) per caracteritzar anomalies en el desenvolupament d'origen prenatal utilitzant la restricció de creixement intrauterí (RCIU) com a model clínic. La tesi està presentada com a compendi de quatre estudis publicats en revistes internacionals de primer quartil. Cada un dels estudis està centrat en la caracterització de la RCIU mitjançant xarxes cerebrals obtingudes a partir d'una modalitat de RM determinada en una etapa pediàtrica diferent, en la vida de subjectes amb RCIU. Així doncs, el primer estudi es centra en la caracterització de la reorganització cerebral produïda per RCIU a l'any de vida mitjançant xarxes cerebrals estructurals basades en RM per difusió. En aquest estudi es demostra que les característiques de xarxa en els subjectes amb RCIU presenten una sèrie d'alteracions relacionades amb un neuro-desenvolupament futur anormal. El segon projecte analitza la utilització de xarxes estructurals cerebrals basades en RM anatòmica convencional per caracteritzar alteracions en nens d'un any amb RCIU. Es demostra que efectivament amb aquesta tècnica també es troben alteracions en els infants amb IUGR, i que aquestes alteracions estan també relacionades amb problemes en el neuro-desenvolupament posterior. En el tercer projecte s'utilitza un model animal de conill amb RCIU per explorar les alteracions en la xarxa cerebral estructural que persisteix a llarg termini. Es demostra que efectivament existeixen alteracions en la organització estructural del cervell persistents a llarg termini i s'observa un efecte compensatori en els subjectes amb RCIU. En el quart projecte s'analitzen les xarxes cerebrals funcional en neonats amb RCIU, demostrant que aquesta condició prenatal genera una reorganització en la connectivitat cerebral que té un substrat funcional, que es pot observar des d'etapes molt precoces de la vida i que està relacionada amb resultats de neuro-comportament.

Diffusion MRI microstructure imaging provides a unique noninvasive probe into tissue microstructure. The technique relies on mathematical models, relating microscopic tissue features to the MR signal. The assumption of Gaussian diffusion oversimplifies the behaviour of water in complex media. Multi-compartment models fit the signal better and enable the estimation of more specific indices, such as axon diameter and density. A previous model comparison framework used data from fixed rat brains to show that three compartment models, designed for intra/extra-axonal diffusion, best explain multi-b-value datasets. The purpose of this PhD work is to translate this analysis to in vivo human brain white matter. It updates the framework methodology by enriching the acquisition protocol, extending the model base and improving the model fitting. In the first part of this thesis, the original fixed rat study is taken in-vivo by using a live human subject on a clinical scanner. A preliminary analysis cannot differentiate the models well. The acquisition protocol is then extended to include a richer angular resolution of diffusion- sampling gradient directions. Compared with ex-vivo data, simpler three-compartment models emerge. Changes in diffusion behaviour and acquisition protocol are likely to have influenced the results. The second part considers models that explicitly seek to explain fibre dispersion, another potentially specific biomarker of neurological diseases. This study finds that models that capture fibre dispersion are preferred, showing the importance of modelling dispersion even in apparently coherent fibres. In the third part, we improve the methodology. First, during the data pre-processing we narrow the region of interest. Second, the model fitting takes into account the varying echo time and compartmental tissue relaxation; we also test the benefit to model performance of different compartmental diffusivities. Next, we evaluate the inter- and intra-subject reproducibility of ranking. In the fourth part, high-gradient Connectom-Skyra data are used to assess the generalisability of earlier results derived from a standard Achieva scanner. Results showed a reproducibility of major trends in the model ranking. In particular, dispersion models explain low gradient strength data best, but cannot capture Connectom signal that remains at very high b-values. The fifth part uses cross-validation and bootstrapping as complementary means to model ranking. Both methods support the previous ranking; however, the leave-one-shell-out cross- validation supports less difference between the models than bootstrapping.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Brain and Cognitive Sciences, 2019
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 91-100).
Neural networks and probabilistic models have different and in many ways complementary strengths and weaknesses: neural networks are flexible and support efficient inference, but rely on large quantities of labeled training data. Probabilistic models can learn from fewer examples, but in many cases remain limited by time-consuming inference algorithms. Thus, both classes of models have drawbacks that both limit their engineering applications and prevent them from being fully satisfying as process models of human learning. This thesis aims to address this state of affairs from both directions, exploring case studies where we make neural networks that learn from less data, and in which we design more efficient inference procedures for generative models. First, we explore recurrent neural networks that learn list-processing procedures (sort, reverse, etc.), and show how ideas from type theory and programming language theory can be used to design a data augmentation scheme that enables effective learning from small datasets. Next, we show how error-driven proposal mechanisms can speed up stochastic search for generative model inversion, first developing a symbolic model for inferring Boolean functions and Horn clause theories, and then a general-purpose neural network model for doing inference in continuous domains such as inverse graphics.
by Owen Lewis.
Ph. D.
Ph.D. Massachusetts Institute of Technology, Department of Brain and Cognitive Sciences

This thesis develops and applies statistical methodologies to model brain atrophy in humans among multiple brain regions and how this may change over time. Throughout this work, Bayesian multilevel models are progressively developed for single and multiple regions at a given time point as well as modelling how connectivity between multiple regions evolves over time in conjunction with region level estimates. The application of these models provide insight into the detection of longitudinal brain atrophy patterns consistent with healthy ageing or progression towards Alzheimer's disease, and should be of interest to biostatisticians and researchers who deal with neurological spatial data.

En el área de la neurociencia de la conciencia, actualmente existe una tendencia de contrastar y comparar modelos de conciencia. Aunque la palabra final es empírica, esfuerzos teóricos son esenciales para poner en contexto tanto suposiciones conceptuales como resultados experimentales. Desde allí, podemos diseñar mejores pruebas y responder a la pregunta sobre que modelo es optimo. En esta dirección, esta tesis explora modelos y enfoques computacionales integrativos. El documento clasifica modelos científicos de conciencia de acuerdo a sus "perfiles explicativos". Resultados empíricos son describidos a la luz de teoría de redes. Luego, herramientas computacionales inspiradas por la integración conceptual de dos de los mas influyentes modelos son implementados para cuantificar las diferencias entre la condiciones de despierto y anestesiado. Finalmente, la tesis introduce nuevos conceptos para evadir el actual reducionismo de algunos modelos, orientando el texto hacia polémicas discusiones. Esta tesis es un trabajo teórico y conceptual inspirado por resultados empíricos que intenta revelar el poder de modelos computacionales y matemáticos en la búsqueda de desarrollar hipótesis testeables y entender mejor la neurociencia de la conciencianciencia.
In the field of neuroscience of consciousness, there is a current trend to contrast and compare existing models of consciousness. Even though the final word is empirical, theoretical efforts are essential to place both, conceptual assumptions and experimental results in context. From that, we can design better assessments and answer the question about what model is optimal. In this direction, this thesis explores models and computational integrative approaches. The document classifies scientific models of consciousness according to their "explanatory profile". Empirical data is described in light of network theory. Then, computational tools inspired by the conceptual integration of two influential models are implemented to quantify differences between awake and anaesthetic conditions. Finally, the thesis introduces new concepts to avoid the current reductionism of some models, pushing the text to controversial discussions. This thesis is a theoretical and conceptual work inspired by empirical results, attempting to reveal the power of computational and mathematical models in order to develop testable hypotheses and understand better the neuroscience of consciousness.

Automated medical image analysis has made significant progress over the past decades. With recent advances in acquiring high quality in vivo images of the developing human brain, analysing this data for the purpose of understanding brain development is rapidly becoming feasible. Premature birth increases the risk of developing neurocognitive and neurobehavioural disorders. Studying the morphology and function of the brain during maturation, provides us not only with a better understanding of normal development, but may help identify causes for these. A difficulty is to differentiate between neurodevelopmental consequences and normal variation. Reference models are therefore needed. This thesis presents computational methods used to obtain such models. As a prerequisite, an efficient topology-preserving registration is required. Existing methods have been evaluated mostly on adult brain images, with considerably different shape and appearance. We evaluate approaches for the fast diffeomorphic registration on a publicly available neonatal brain image dataset, and present an improved inverse consistent variant of the stationary velocity free-form deformation algorithm. We employ this algorithm for the construction of a spatio-temporal atlas of the neonatal brain, and compare two different approaches. The first approach is based on the registration of all pairs of images. Residual misalignment thereby still impacts the sharpness of the atlas. More detail is preserved with an iterative refinement of the transformations relating each image to the atlas space. We developed a second approach, which jointly estimates mean shape and longitudinal change iteratively. The final atlas demonstrates increased sharpness and temporal consistency. Finally, we present deformable models for the reconstruction of the neonatal cortex, which correct for common errors observed in state-of-the-art neonatal brain segmentations. Our models were found by experts to be superior to the original segmentation in terms of accurately delineating the cortical anatomy, and form a vital component of image processing pipelines of the Developing Human Connectome Project.

Our motivation is to develop an automated technique for the segmentation of sub-cortical human brain structures from MR images. To this purpose, models of shape-and-appearance are constructed and fit to new image data. The statistical models are trained from 317 manually labelled T1-weighted MR images. Shape is modelled using a surface-based point distribution model (PDM) such that the shape space is constrained to the linear combination of the mean shape and eigenvectors of the vertex coordinates. In addition, to model intensity at the structural boundary, intensities are sampled along the surface normal from the underlying image. We propose a novel Bayesian appearance model whereby the relationship between shape and intensity are modelled via the conditional distribution of intensity given shape. Our fully probabilistic approach eliminates the need for arbitrary weightings between shape and intensity as well as for tuning parameters that specify the relative contribution between the use of shape constraints and intensity information. Leave-one-out cross-validation is used to validate the model and fitting for 17 structures. The PDM for shape requires surface parameterizations of the volumetric, manual labels such that vertices retain a one-to-one correspondence across the training subjects. Surface parameterizations with correspondence are generated through the use of deformable models under constraints that embed the correspondence criterion within the deformation process. A novel force that favours equal-area triangles throughout the mesh is introduced. The force adds stability to the mesh such that minimal smoothing or within-surface motion is required. The use of the PDM for segmentation across a series of subjects results in a set surfaces that retain point correspondence. The correspondence facilitates landmark-based shape analysis. Amongst other metrics, vertex-wise multivariate statistics and discriminant analysis are used to investigate local and global size and shape differences between groups. The model is fit, and shape analysis is applied to two clinical datasets.

La complexité de la connaissance humaine est révèlée dans l'organisation spatiale et temporelle de la dynamique du cerveau. Nous pouvons connaître cette organisation grâce à l'analyse des signaux dépendant du niveau d'oxygène sanguin (BOLD), lesquels sont obtenus par l'imagerie par résonance magnétique fonctionnelle (IRMf). Nous observons des dépendances statistiques entre les régions du cerveau dans les données BOLD. Ce phénomène s' appelle connectivité fonctionnelle (CF). Des modèles computationnels sont développés pour reproduire la connectivité fonctionnelle (CF). Comme les études expérimentales précédantes, ces modèles assument que la CF est stationnaire, c'est-à-dire la moyenne et la covariance des séries temporelles BOLD utilisées par la CF sont constantes au fil du temps. Cependant, des nouvelles études expérimentales concernées par la dynamique de la CF à différentes échelles montrent que la CF change dans le temps. Cette caractéristique n'a pas été reproduite dans ces modèles computationnels précédants. Ici on a augmenté la non-linéarité de la dynamique locale dans un modèle computationnel à grande échelle. Ce modèle peut reproduire la grande variabilité de la CF observée dans les études expérimentales
The complexity of human cognition is revealed in the spatio-temporal organization of brain dynamics. We can gain insight into this organization through the analysis of blood oxygenation-level dependent (BOLD) signals, which are obtained from functional magnetic resonance imaging (fMRI). In BOLD data we can observe statistical dependencies between brain regions. This phenomenon is known as functional connectivity (FC). Computational models are being developed to reproduce the FC of the brain. As in previous empirical studies, these models assume that FC is stationary, i.e. the mean and the covariance of the BOLD time series used for the FC are constant over time. Nevertheless, recent empirical studies focusing on the dynamics of FC at different time scales show that FC is variable in time. This feature is not reproduced in the simulated data generated by some previous computational models. Here we have enhanced the non-linearity of local dynamics in a large-scale computational model. By enhancing this non-linearity, our model is able to reproduce the variability of the FC found in empirical data

La biophotométrie sur fibre est une technique puissante utilisée en neuroscience pour surveiller les fluctuations dynamiques des niveaux de calcium en corrélation avec des événements neuronaux, tels que la génération de potentiel d’action, l’exocytose de neurotransmetteurs, des modifications de la plasticité synaptique et la transcription de gènes dans les structures cérébrales profondes d’animaux de laboratoire vivants. Cette approche permet également d’étudier la corrélation entre les processus neuronaux et le comportement de modèles animaux vivants afin de percer les mystéres du cerveaux et de nombreuses maladies comme la maladie d’Alzheimer. Les appareils de biophotométrie sur fibre de table classiques utilisent une fibre optique attachée pour émettre de la lumière et récupérer les signaux de fluorescence, ce qui présente un risque de rupture, de contrainte et de blessure potentielle. Ces systèmes sont également encombrants et nécessitent des tensions de fonctionnement élevées. Par conséquent, leur utilité dans les études sur des animaux vivants est limitée. Le but de ce projet est de mettre en place une interface neuronale optique sans fil pour effectuer la détection de fluorescence avec des modèles animaux vivants sans restreindre leurs mouvements ni induire de stress dû au câble. Nous avons conçu un système de biophotométrie par fibre optique sans fil légère et compacte pour une utilisation chronique basée sur un capteur de fluorescence CMOS (Complementary Metal-Oxide- Semiconductor) intégré offrant une sensibilité élevée, une plage dynamique élevée et une consommation d’énergie très faible. Le système de biophotométrie à fibre présenté incorpore tous les aspects d’un système de biophotométrie à fibres englobé dans un sans fil. Les principales contributions de ce travail ont été rapportées dans neuf conférences et trois articles de journaux publiés ou soumis, ainsi que dans une divulgation d’invention. Les mesures de biophotométrie en fluorescence nécessitent un appareil de laboratoire à large plage dynamique (DR) et à haute sensibilité. Cependant, il est souvent très difficile de mesurer avec précision les petites variations de fluorescence en présence de bruit et d’autofluorescence de tissu de fond élevée. Une contribution importante de ce travail concerne le développement de biocapteurs optoélectroniques CMOS intégrés sur mesure et de circuits de traitement permettant de détecter les signaux de fluorescence très faibles et de les convertir en codes numériques de haute précision, afin de construire des dispositifs de détection du cerveau montables sur la tête de souris de laboratoire, très compacts et légers. Nous avons conçu une première puce de biocapteur CMOS haute précision offrant une plage de tension de fonctionnement basse, une basse consommation, une haute sensibilité et une gamme dynamique élevée basée sur une architecture basse tension intégrant un circuit frontal à détection différentielle avec heure [sigma delta] modulation avec un amplificateur de transconductance capacitif différentiel (ATCCD).
Ce nouveau système offre une mise en oeuvre simplifiée ainsi qu’une architecture à faible consommation utilisant une stratégie de partage du matériel. La détection différentielle et les photodiodes factices avec le ATCCD permettent d’atteindre une sensibilité élevée en supprimant les dark current de la photodiode, en utilisant un petit condensateur d’intégration dans le ATCCD. Les résultats de mesure sont présentés pour le capteur de biophotométrie proposé, fabriqué avec une technologie CMOS de 0.18 mm, consommant 41 mWd’une tension d’alimentation de 1.8 V, tout en atteignant une gamme dynamique maximale de 86 dB, une bande passante de 50 Hz, une sensibilité de 24 mV/nW et un courant minimum détectable de 2.6-pArms à un taux d’échantillonnage de 20 kS/s. Un autre défi critique pour un système de photométrie à fibre pour petits animaux concerne la gestion de la consommation de courant importante nécessaire à la source de lumière d’excitation pour fournir une puissance de sortie de lumière suffisante au tissu afin de déclencher la fluorescence. Par conséquent, des impulsions lumineuses d’excitation courtes doivent être utilisées par rapport à la période d’échantillonnage du signal de fluorescence (>10 ms), afin de réduire la consommation de courant moyenne et d’allonger la durée de vie de la batterie. Pour répondre à cette exigence critique, nous avons amélioré notre conception avec un deuxième prototype de biocapteur utilisant de nouvelles techniques de circuit pour offrir une sensibilité élevée et une plage dynamique élevée avec un temps de conversion réduit permettant l’utilisation d’impulsions lumineuses à cycle de fonctionnement réduit et de consommation faible. Le biocapteur est basé sur un convertisseur analogique-numérique (CAN) à comptage étendu, et un convertisseur analogique-numérique de premier ordre SD, dont le fonctionnement est synchronisé avec les impulsions lumineuses d’excitation. Le biocapteur présente une gamme dynamique de 104 dB à un temps de conversion de 3 % de la période d’échantillonnage du signal de fluorescence et réduit la consommation électrique de la DEL de 97 %. Un dernier aspect critique concerne la flexibilité du biocapteur pour effectuer des tests fiables in vivo. Réaliser un test pratique in vivo nécessite d’ajuster la sensibilité du biocapteur et la puissance de sortie de la DEL du biocapteur afin de s’adapter à différents niveaux de fluorescence et différents environnements physiologiques à l’intérieur des tissus de l’animal vivant. Ainsi, nous avons conçu un troisième biocapteur incorporant une sensibilité et un temps de conversion programmables afin d’optimiser la consommation d’énergie de DEL et de permettre un très faible facteur de fonctionnement excitation/détection. Cette toute nouvelle architecture de capteurs utilise un CAN à temps discret [sigma delta] avec une technique de double échantillonnage numérique corrélée permettant la détection de photocourants inférieurs à 1 pArms. Cette conception a été utilisée comme module de base pour développer un prototype de headstage sans fil. Nous avons mis en place et testé in vitro avec succès ce système de biophotométrie à fibre, qui comprend la puce de biocapteur proposée, avec une tranche de cerveau de souris exprimant GCaMP6, un indicateur de calcium génétiquement codé.
Fiber biophotometry is a powerful technique in neuroscience to monitor the dynamic fluctuations in calcium levels correlated with neural events, such as action potential generation, exocytosis of neurotransmitters, changes in synaptic plasticity, and gene transcription in deep brain structures in live laboratory animals. This approach allows studying the correlation between neuronal processes and the behavior of live animal models in order to learn more about the brain function and its associated diseases. Conventional bench-top fiber biophotometry apparatus use a tethered optical fiber to deliver light and to retrieve fluorescence signals, which involves risk of breakage, stress, and potential injury. These systems are also bulky and require high operating voltages. Therefore, their usefulness to conduct studies with live animals is limited. The goal of this project is to implement a wireless optical neural interface to perform fluorescence sensing with live animal models without restraining their movement or inducing stress due to cable tethering. We designed a lightweight and compact size wireless fiber biophotometry headstage for chronic utilization based on a custom integrated Complementary Metal-Oxide-Semiconductor (CMOS) fluorescence sensor providing high-sensitivity, high-dynamic range, and very low-power consumption. The presented head-mountable fiber biophotometry system incorporates all aspects of a conventional tethered fiber-based biophotometry system encompassed into a wireless headstage. The main contributions of this work were reported in nine conferences and three journal papers published or submitted, and in one invention disclosure. Fluorescence biophotometry measurements require wide dynamic range (DR) and high-sensitivity laboratory apparatus. But, it is often very challenging to accurately resolve the small fluorescence variations in presence of noise and high background tissue autofluorescence. An important contribution of this work concerns the development of custom integrated CMOS optoelectronic biosensors and processing circuits to detect very weak fluorescence signals, and to convert them into high-precision digital codes, for building very compact and lightweight head-mountable brain sensing devices for laboratory mice. We first designed a high-precision CMOS biosensor chip providing low operating voltage, low-power, high-sensitivity, and high-dynamic range based on a low-voltage architecture that embeds a differential sensing front-end circuitry with a continuous-time [sigma delta] modulation with a differential capacitive transconductance amplifier (DCTIA). This novel system offers a simplified implementation as well as a low-power architecture leveraging a hardware sharing strategy. Differential sensing and dummy photodiodes with the DCTIA enables to achieve high-sensitivity by suppressing the photodiode dark currents and using a small integration capacitor in the DCTIA. Measurement results are presented for the proposed biophotometry sensor fabricated in a 0.18-mm CMOS technology, consuming 41 mW from a 1.8-V supply voltage, while achieving a peak dynamic range of 86 dB over a 50-Hz input bandwidth, a sensitivity of 24 mV/nW and a minimum detectable current of 2.46-pArms at a 20-kS/s sampling rate.
Another critical challenge for a head-mountable fiber photometry system is when handling the large current consumption needed for the excitation light source to provide sufficient light output power to the tissue in order to trigger fluorescence. Hence, short excitation light pulses must be used, relative to the sampling period of the fluorescence signal (>10 ms), in order to decrease the average current consumption, and extend the battery lifetime. To address this critical requirement, we improved our design with a second biosensor prototype using novel circuit techniques to provide high-sensitivity and a high-dynamic range with a short conversion time to allow the utilization of low-duty cycle light pulses and low-power consumption. The biosensor is based on an extended counting ADC, first-order [sigma delta] and single slope ADC, whose operation is synchronized with the excitation light pulses. The biosensor presents a high-dynamic range of 104 dB at a conversion time of 3 % of the fluorescence signal sampling period and decreases the power consumption of the excitation light source by 97%. A last critical aspect concerns the flexibility of the biosensor to perform reliable tests in-vivo. Performing a practical test in-vivo requires to adjust the biosensor sensitivity and the excitation light source output power of the biosensor to adapt to different fluorescence levels and different physiological environments inside the live animal tissues. Thus, we designed a third biosensor incorporating a programmable sensitivity and a programmable conversion time to optimize the excitation light power consumption, and to enable very low excitation/sensing duty cycle. This completely new sensor architecture utilizes a discrete time SD ADC with digital correlated double sampling technique enabling detection of low photocurrents as low as 1 pArms. This design was used as a core module to develop a wireless head-mountable optical headstage prototype. We have implemented and sucessfully tested this fiber photometry headstage, which includes the proposed biosensor chip, in-vitro with a mouse brain slice expressing GCaMP6, a genetically encoded calcium indicator.

Thesis: M. Eng., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, February, 2021
Cataloged from the official PDF of thesis.
Includes bibliographical references (pages 55-58).
Recent research has seen the rise of powerful neural-network language models that are sufficiently computationally precise and neurally plausible as to serve as a jumping-off base for our understanding of language processing in the brain. Because these models have been developed for optimizing a similar objective (word prediction), their brain predictions are often correlated, even though the models differ along several architectural and conceptual features, yielding a major challenge for testing which model features are most relevant for predicting language processing in the brain. Here, we address this challenge by synthesizing new sentence stimuli that maximally expose the disagreement between the predictions of a set of language models ('controversial stimuli'), which would not naturally occur in large language corpora . To do so, we develop a platform for systematizing this sentence synthesis process, providing a way to test different model-based hypotheses easily and efficiently. An initial exploration with this platform has begun to give us some intuition for how choosing from different pools of candidate words affect the kinds of sentences produced, and what kinds of changes tend to produce controversial sentences. For example, we show that the disagreement score, or the maximum amount of disagreement between models for a sentence, converges. This approach will eventually allow us to determine which models perform in the most human-like way and are most successful in predicting language processing in the brain, thus hopefully leading to insights on the mechanisms of human language understanding.
by Lara I. Rakocevic.
M. Eng.
M.Eng. Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science

Preterm birth is the leading cause of neonatal mortality, with survivors experiencing motor, cognitive and other deficits at increased rates. In preterm infancy, the developing brain undergoes folding, myelination, and rapid cellular growth. Diffusion-Weighted Magnetic Resonance Imaging (DW MRI) is an imaging modality that allows noninvasive inference of cellular microstructure in living tissue, and its parameters reflect changes in brain tissue composition. In this thesis, we employ compartment models of DW MRI to investigate the microstructure in preterm-born subjects at different ages. Within infants, we have used the NODDI model to investigate longitudinal changes in neurite density and orientation dispersion within the white matter, cerebral cortex and thalamus, explaining known trends in diffusion tensor parameters with greater specificity. We then used a quantitative T2 sequence to develop and investigate a novel, multi-modal parameter known as the ‘g-ratio’. We have also investigated changing microstructural geometry within the cortex. Immediately after preterm birth, the highly-ordered underlying cellular structure makes diffusion in the cortex almost entirely radial. This undergoes a transition to a disordered and isotropic state over the first weeks of life, which we have used the DIAMOND model to quantify. This radiality decreases at a rate that depends on the cortical lobe. In a cohort of young adults born extremely preterm, we have quantified differences in brain microstructure compared to term-born controls. In preterm subjects, the brain structures are smaller than for controls, leading to increased partial volume in some regions of interest. We introduce a method to infer diffusion parameters in partial volume, even for regions which are smaller than the diffusion resolution. Overall, this thesis utilises and evaluates a variety of compartment models of DW MRI. By developing and applying principled and robust methodology, we present new insights into microstructure within the preterm-born brain.

Dans cette thèse, nous avons proposé deux modèles multi-compartiments en IRM de relaxométrie T2 (MCT2) fournissant des informations sur la microstructure des tissus cérébraux. Trois compartiments de relaxométrie T2 ont été considérés dans chaque voxel représentant des tissus avec des temps de relaxation T2 courts, T2 moyens et T2 élevés. La complexité associée à l'estimation des paramètres de tels modèles paramétriques a ensuite été explorée. Le premier modèle MCT2 que nous avons proposé a estimé la représentation fractionnelle de compartiments T2 prédéfinis. Dans le second modèle, les représentations fractionnaires et le paramètre de relaxation moyenne ont été estimés pour le compartiment T2 moyen. Dans les deux modèles, le choix de l'approche était justifié par une analyse de la fonction de coût et un cadre d’estimation a été proposé. Le modèle MCT2 a été utilisé pour deux applications. Dans la première application, l’évolution des biomarqueurs de MCT2 a été étudiée dans les lésions de sclérose en plaques (SEP) présentant une prise de contraste gadolinium (Gd) ou non chez 10 patients présentant un syndrome cliniquement isolé. La seconde application a démontré le potentiel de combinaison des biomarqueurs MCT2 avec les informations de microstructure dérivées de l'IRM de diffusion pour identifier les régions présentant une prise de contraste Gd dans les lésions de SEP. Les résultats montrent que les biomarqueurs MCT2 proposés peuvent constituer des outils efficaces pour étudier l’état et l’évolution de la microstructure tissulaire dans le cerveau. La combinaison des biomarqueurs MCT2 avec les informations de microstructure dMRI nous a permis de progresser vers la résolution d’un problème critique et délicat consistant à limiter l'utilisation de gadolinium dans la détection de régions de lésion améliorantes dans les lésions de SEP
In this thesis, we propose two multi-compartment T2 relaxometry (MCT2) models which provide information on brain tissue microstructure. Three T2 relaxometry compartments were considered in each voxel representing tissues with short T2, medium T2 and high T2 relaxation times. The complexity associated with the estimation of the parameters for such parametric models has then been explored. The first MCT2 model we propose computes the fractional representation of pre-defined T2 pools. In the next MCT2 model the fractional representations as well as T2 pool parameter were estimated for the medium T2 compartment. For both models the choice of approach was justified using a cost function analysis and a dedicated estimation framework was proposed.Our MCT2 model was used for two applications. In the first application the evolution of MCT2 biomarkers was studied in gadolinium (Gd) enhancing and nonenhancing regions of multiple sclerosis (MS) lesions in 10 patients with clinically isolated syndrome. The potential of combining the MCT2 biomarkers with diffusion MRI (dMRI) derived microstructure information to identify Gd enhancing regions in MS lesions was then demonstrated in the second application. The results show that the proposed MCT2 biomarkers can be effective tools to study the condition and evolution of tissue microstructures in the brain. Combining the MCT2 biomarkers with dMRI microstructure information enabled us to address a critical and challenging problem of limiting the use of gadolinium usage in detecting enhancing lesion regions in MS patients