Analyse par apprentissage automatique des réponses fMRI du cortex auditif à des modulations spectro-temporelles

L'application de classifieurs linéaires à l'analyse des données d'imagerie cérébrale (fMRI) a mené à plusieurs percées intéressantes au cours des dernières années. Ces classifieurs combinent linéairement les réponses des voxels pour détecter et catégoriser différents états du cerveau. Ils sont plus agnostics que les méthodes d'analyses conventionnelles qui traitent systématiquement les patterns faibles et distribués comme du bruit. Dans le présent projet, nous utilisons ces classifieurs pour valider une hypothèse portant sur l'encodage des sons dans le cerveau humain. Plus précisément, nous cherchons à localiser des neurones, dans le cortex auditif primaire, qui détecteraient les modulations spectrales et temporelles présentes dans les sons. Nous utilisons les enregistrements fMRI de sujets soumis à 49 modulations spectro-temporelles différentes. L'analyse fMRI au moyen de classifieurs linéaires n'est pas standard, jusqu'à maintenant, dans ce domaine. De plus, à long terme, nous avons aussi pour objectif le développement de nouveaux algorithmes d'apprentissage automatique spécialisés pour les données fMRI. Pour ces raisons, une bonne partie des expériences vise surtout à étudier le comportement des classifieurs. Nous nous intéressons principalement à 3 classifieurs linéaires standards, soient l'algorithme machine à vecteurs de support (linéaire), l'algorithme régression logistique (régularisée) et le modèle bayésien gaussien naïf (variances partagées).

Rendu de matériaux semi-transparents hétérogènes en temps réel

On retrouve dans la nature un nombre impressionnant de matériaux semi-transparents tels le marbre, le jade ou la peau, ainsi que plusieurs liquides comme le lait ou les jus. Que ce soit pour le domaine cinématographique ou le divertissement interactif, l'intérêt d'obtenir une image de synthèse de ce type de matériau demeure toujours très important. Bien que plusieurs méthodes arrivent à simuler la diffusion de la lumière de manière convaincante a l'intérieur de matériaux semi-transparents, peu d'entre elles y arrivent de manière interactive. Ce mémoire présente une nouvelle méthode de diffusion de la lumière à l'intérieur d'objets semi-transparents hétérogènes en temps réel. Le coeur de la méthode repose sur une discrétisation du modèle géométrique sous forme de voxels, ceux-ci étant utilisés comme simplification du domaine de diffusion. Notre technique repose sur la résolution de l'équation de diffusion à l'aide de méthodes itératives permettant d'obtenir une simulation rapide et efficace. Notre méthode se démarque principalement par son exécution complètement dynamique ne nécessitant aucun pré-calcul et permettant une déformation complète de la géométrie.

Étude des artefacts en tomodensitométrie par simulation Monte Carlo

En radiothérapie, la tomodensitométrie (CT) fournit l’information anatomique du patient utile au calcul de dose durant la planification de traitement. Afin de considérer la composition hétérogène des tissus, des techniques de calcul telles que la méthode Monte Carlo sont nécessaires pour calculer la dose de manière exacte. L’importation des images CT dans un tel calcul exige que chaque voxel exprimé en unité Hounsfield (HU) soit converti en une valeur physique telle que la densité électronique (ED). Cette conversion est habituellement effectuée à l’aide d’une courbe d’étalonnage HU-ED. Une anomalie ou artefact qui apparaît dans une image CT avant l’étalonnage est susceptible d’assigner un mauvais tissu à un voxel. Ces erreurs peuvent causer une perte cruciale de fiabilité du calcul de dose. Ce travail vise à attribuer une valeur exacte aux voxels d’images CT afin d’assurer la fiabilité des calculs de dose durant la planification de traitement en radiothérapie. Pour y parvenir, une étude est réalisée sur les artefacts qui sont reproduits par simulation Monte Carlo. Pour réduire le temps de calcul, les simulations sont parallélisées et transposées sur un superordinateur. Une étude de sensibilité des nombres HU en présence d’artefacts est ensuite réalisée par une analyse statistique des histogrammes. À l’origine de nombreux artefacts, le durcissement de faisceau est étudié davantage. Une revue sur l’état de l’art en matière de correction du durcissement de faisceau est présentée suivi d’une démonstration explicite d’une correction empirique.

Variabilité individuelle de l'allocation fonctionnelle des régions cérébrales perceptives dans l'autisme

Introduction : Une augmentation de la plasticité cérébrale est susceptible d’être impliquée dans la réallocation des régions corticales et dans les nombreuses altérations microstructurelles observées en autisme. Considérant les nombreux résultats démontrant un surfonctionnement perceptif et un fonctionnement moteur atypique en autisme, l’augmentation de la plasticité cérébrale suggère une plus grande variabilité individuelle de l’allocation fonctionnelle chez cette population, plus spécifiquement dans les régions perceptives et motrices. Méthode : Afin de tester cette hypothèse, 23 participants autistes de haut-niveau et 22 non-autistes appariés pour l’âge, le quotient intellectuel, les résultats au test des Matrices de Raven et la latéralité, ont réalisé une tâche d’imitation visuo-motrice dans un appareil d’imagerie par résonnance magnétique fonctionnelle (IRMf). Pour chaque participant, les coordonnées du pic d’activation le plus élevé ont été extraites des aires motrices primaires (Aire de Brodmann 4 (BA4)) et supplémentaires (BA6), du cortex visuo-moteur pariétal supérieur (BA7) ainsi que des aires visuelles primaires (BA17) et associatives (BA18+19) des deux hémisphères. L’étendue des activations, mesurée en fonction du nombre de voxels activés, et la différence d’intensité des activations, calculée en fonction du changement moyen d’intensité du signal ont également été considérées. Pour chaque région d’intérêt et hémisphère, la distance entre la localisation de l’activation maximale de chaque participant par rapport à celle de la moyenne de son groupe a servi de variable d’intérêt. Les moyennes de ces distances individuelles obtenues pour chaque groupe et chacune des régions d’intérêt ont ensuite été soumises à une ANOVA à mesures répétées afin de déterminer s’il existait des différences de variabilité dans la localisation des activations entre les groupes. Enfin, l’activation fonctionnelle générale à l’intérieur de chaque groupe et entre les groupes a également été étudiée. Résultats : Les résultats démontrent qu’une augmentation de la variabilité individuelle en terme de localisation des activations s’est produite à l’intérieur des deux groupes dans les aires associatives motrices et visuelles comparativement aux aires primaires associées. Néanmoins, malgré le fait que cette augmentation de variabilité dans les aires associatives soit partagée, une comparaison directe de celle-ci entre les groupes a démontré que les autistes présentaient une plus grande variabilité de la localisation des activations fonctionnelles dans le cortex visuo-moteur pariétal supérieur (BA7) et les aires associatives visuelles (BA18+19) de l’hémisphère gauche. Conclusion : Des stratégies différentes et possiblement uniques pour chaque individu semblent être observées en autisme. L’augmentation de la variabilité individuelle de la localisation des activations fonctionnelles retrouvée chez les autistes dans les aires associatives, où l’on observe également davantage de variabilité chez les non-autistes, suggère qu’une augmentation et/ou une altération des mécanismes de plasticité est impliquée dans l’autisme.

Multi-view 3D reconstruction using virtual cameras

Ce mémoire s'intéresse à la reconstruction d'un modèle 3D à partir de plusieurs images. Le modèle 3D est élaboré avec une représentation hiérarchique de voxels sous la forme d'un octree. Un cube englobant le modèle 3D est calculé à partir de la position des caméras. Ce cube contient les voxels et il définit la position de caméras virtuelles. Le modèle 3D est initialisé par une enveloppe convexe basée sur la couleur uniforme du fond des images. Cette enveloppe permet de creuser la périphérie du modèle 3D. Ensuite un coût pondéré est calculé pour évaluer la qualité de chaque voxel à faire partie de la surface de l'objet. Ce coût tient compte de la similarité des pixels provenant de chaque image associée à la caméra virtuelle. Finalement et pour chacune des caméras virtuelles, une surface est calculée basée sur le coût en utilisant la méthode de SGM. La méthode SGM tient compte du voisinage lors du calcul de profondeur et ce mémoire présente une variation de la méthode pour tenir compte des voxels précédemment exclus du modèle par l'étape d'initialisation ou de creusage par une autre surface. Par la suite, les surfaces calculées sont utilisées pour creuser et finaliser le modèle 3D. Ce mémoire présente une combinaison innovante d'étapes permettant de créer un modèle 3D basé sur un ensemble d'images existant ou encore sur une suite d'images capturées en série pouvant mener à la création d'un modèle 3D en temps réel.

Segmentation of Brain Tissue from Magnetic Resonance Images

Segmentation of medical imagery is a challenging problem due to the complexity of the images, as well as to the absence of models of the anatomy that fully capture the possible deformations in each structure. Brain tissue is a particularly complex structure, and its segmentation is an important step for studies in temporal change detection of morphology, as well as for 3D visualization in surgical planning. In this paper, we present a method for segmentation of brain tissue from magnetic resonance images that is a combination of three existing techniques from the Computer Vision literature: EM segmentation, binary morphology, and active contour models. Each of these techniques has been customized for the problem of brain tissue segmentation in a way that the resultant method is more robust than its components. Finally, we present the results of a parallel implementation of this method on IBM's supercomputer Power Visualization System for a database of 20 brain scans each with 256x256x124 voxels and validate those against segmentations generated by neuroanatomy experts.

Compact Representations for Fast Nonrigid Registration of Medical Images

We develop efficient techniques for the non-rigid registration of medical images by using representations that adapt to the anatomy found in such images. Images of anatomical structures typically have uniform intensity interiors and smooth boundaries. We create methods to represent such regions compactly using tetrahedra. Unlike voxel-based representations, tetrahedra can accurately describe the expected smooth surfaces of medical objects. Furthermore, the interior of such objects can be represented using a small number of tetrahedra. Rather than describing a medical object using tens of thousands of voxels, our representations generally contain only a few thousand elements. Tetrahedra facilitate the creation of efficient non-rigid registration algorithms based on finite element methods (FEM). We create a fast, FEM-based method to non-rigidly register segmented anatomical structures from two subjects. Using our compact tetrahedral representations, this method generally requires less than one minute of processing time on a desktop PC. We also create a novel method for the non-rigid registration of gray scale images. To facilitate a fast method, we create a tetrahedral representation of a displacement field that automatically adapts to both the anatomy in an image and to the displacement field. The resulting algorithm has a computational cost that is dominated by the number of nodes in the mesh (about 10,000), rather than the number of voxels in an image (nearly 10,000,000). For many non-rigid registration problems, we can find a transformation from one image to another in five minutes. This speed is important as it allows use of the algorithm during surgery. We apply our algorithms to find correlations between the shape of anatomical structures and the presence of schizophrenia. We show that a study based on our representations outperforms studies based on other representations. We also use the results of our non-rigid registration algorithm as the basis of a segmentation algorithm. That algorithm also outperforms other methods in our tests, producing smoother segmentations and more accurately reproducing manual segmentations.

Towards the computer-aided diagnosis of dementia based on the geometric and network connectivity of structural MRI data

We present an intuitive geometric approach for analysing the structure and fragility of T1-weighted structural MRI scans of human brains. Apart from computing characteristics like the surface area and volume of regions of the brain that consist of highly active voxels, we also employ Network Theory in order to test how close these regions are to breaking apart. This analysis is used in an attempt to automatically classify subjects into three categories: Alzheimer’s disease, mild cognitive impairment and healthy controls, for the CADDementia Challenge.

The neurological underpinnings of cluttering: some initial findings

Background Cluttering is a fluency disorder characterised by overly rapid or jerky speech patterns that compromise intelligibility. The neural correlates of cluttering are unknown but theoretical accounts implicate the basal ganglia and medial prefrontal cortex. Dysfunction in these brain areas would be consistent with difficulties in selection and control of speech motor programs that are characteristic of speech disfluencies in cluttering. There is a surprising lack of investigation into this disorder using modern imaging techniques. Here, we used functional MRI to investigate the neural correlates of cluttering. Method We scanned 17 adults who clutter and 17 normally fluent control speakers matched for age and sex. Brain activity was recorded using sparse-sampling functional MRI while participants viewed scenes and either (i) produced overt speech describing the scene or (ii) read out loud a sentence provided that described the scene. Speech was recorded and analysed off line. Differences in brain activity for each condition compared to a silent resting baseline and between conditions were analysed for each group separately (cluster-forming threshold Z > 3.1, extent p < 0.05, corrected) and then these differences were further compared between the two groups (voxel threshold p < 0.01, extent > 30 voxels, uncorrected). Results In both conditions, the patterns of activation in adults who clutter and control speakers were strikingly similar, particularly at the cortical level. Direct group comparisons revealed greater activity in adults who clutter compared to control speakers in the lateral premotor cortex bilaterally and, as predicted, on the medial surface (pre-supplementary motor area). Subcortically, adults who clutter showed greater activity than control speakers in the basal ganglia. Specifically, the caudate nucleus and putamen were overactive in adults who clutter for the comparison of picture description with sentence reading. In addition, adults who clutter had reduced activity relative to control speakers in the lateral anterior cerebellum bilaterally. Eleven of the 17 adults who clutter also stuttered. This comorbid diagnosis of stuttering was found to contribute to the abnormal overactivity seen in the group of adults who clutter in the right ventral premotor cortex and right anterior cingulate cortex. In the remaining areas of abnormal activity seen in adults who clutter compared to controls, the subgroup who clutter and stutter did not differ from the subgroup who clutter but do not stutter. Conclusions Our findings were in good agreement with theoretical predictions regarding the neural correlates of cluttering. We found evidence for abnormal function in the basal ganglia and their cortical output target, the medial prefrontal cortex. The findings are discussed in relation to models of cluttering that point to problems with motor control of speech.

Dynamic Changes in the Mental Rotation Network Revealed by Pattern Recognition Analysis of fMRI Data

We investigated the temporal dynamics and changes in connectivity in the mental rotation network through the application of spatio-temporal support vector machines (SVMs). The spatio-temporal SVM [Mourao-Miranda, J., Friston, K. J., et al. (2007). Dynamic discrimination analysis: A spatial-temporal SVM. Neuroimage, 36, 88-99] is a pattern recognition approach that is suitable for investigating dynamic changes in the brain network during a complex mental task. It does not require a model describing each component of the task and the precise shape of the BOLD impulse response. By defining a time window including a cognitive event, one can use spatio-temporal fMRI observations from two cognitive states to train the SVM. During the training, the SVM finds the discriminating pattern between the two states and produces a discriminating weight vector encompassing both voxels and time (i.e., spatio-temporal maps). We showed that by applying spatio-temporal SVM to an event-related mental rotation experiment, it is possible to discriminate between different degrees of angular disparity (0 degrees vs. 20 degrees, 0 degrees vs. 60 degrees, and 0 degrees vs. 100 degrees), and the discrimination accuracy is correlated with the difference in angular disparity between the conditions. For the comparison with highest accuracy (08 vs. 1008), we evaluated how the most discriminating areas (visual regions, parietal regions, supplementary, and premotor areas) change their behavior over time. The frontal premotor regions became highly discriminating earlier than the superior parietal cortex. There seems to be a parcellation of the parietal regions with an earlier discrimination of the inferior parietal lobe in the mental rotation in relation to the superior parietal. The SVM also identified a network of regions that had a decrease in BOLD responses during the 100 degrees condition in relation to the 0 degrees condition (posterior cingulate, frontal, and superior temporal gyrus). This network was also highly discriminating between the two conditions. In addition, we investigated changes in functional connectivity between the most discriminating areas identified by the spatio-temporal SVM. We observed an increase in functional connectivity between almost all areas activated during the 100 degrees condition (bilateral inferior and superior parietal lobe, bilateral premotor area, and SMA) but not between the areas that showed a decrease in BOLD response during the 100 degrees condition.

Visualização de estruturas internas em volumes de dados multimodais

Com o aperfeiçoamento de técnicas de aquisição de imagens médicas, como, por exemplo, a tomografia computadorizada e ressonância magnética, a capacidade e a fidelidade do diagnóstico por imagens foram ampliadas. Atualmente, existe a tendência de utilizarem-se imagens através de diversas modalidades para um único diagnóstico, principalmente no caso de doenças graves. Entretanto, o registro e a fusão dessas imagens, chamadas mutimodais, em uma única representação 3D do paciente é uma arefa extremamente dif[icil, que consome tempo e que está sujeita a erros. Sendo assim, a integração de imagens de diferentes modalidades tem sido objeto de pesquisa sob a denominação de Visualização de Volumes de Dados Multimodais. Sistemas desenvolvidos com este objetivo são usados, principalmente, para combinar informações metabólicas e funcionais com dados de anatomia, aumentando a precisão do diagnóstico, uma vez que possibilitam extrrair uma superfície ou região da imagem que apresenta a anatomia, e, então, observar a atividade funcional na outra modalidade. Durante a análise de tais imagens, os médicos estão interessados e quantificar diferentes estruturas. Seusobjetivos envolvem, por exemplo, a visualização de artérias e órgãos do corpo humano para análise de patologias, tais como tumores, má-formações artério-venosas, ou lesões em relação às estuturas que as circundam. Assim, um dos principais obetivos de um algoritmo de visualização volumétrica é permitir a identificação e exploração de estruturas internas no volume. Como o volume é normalmente um "bloco de dados", não se pode visualizar o seu interior, a menos que se assuma que é possível ver através de voxels transparentes, ou que é possivel remover voxels que estão na frente na qual o usuário está interessado, o que foi feito através de técnicas de segmentação ou de corte. Este trabalho presenta uma abordagem para a visualização de estruturas internas em volumes de dados multimodais. A abordagem está fundamentada na utilização de ferramentas de corte, tanto geométricas quanto baseadas em conteúdo, evitando, assim, o uso de técnicas de segmentação; e na integração dos dados multimodais na etapa de acumulação de pipeline de visualização volumétrica. Considerando que as aplicações que suportam este tipo de visualização envolvem a integração de várias ferramentas, tais como registro, corte e visualização, também é apresentado o projeto de um framework que permite esta integração e um alto grau de interação com usuário. Para teste e validação das técnicas de visualização de estruturas internas propostas e do algoritmo desenvolvido, que consiste numa extensão do algoritmo de ray casting tradicional, foram implementadas algumas classes desse framework. Uma revisão baseada na análise e na classificação das ferramentas de corte e funções de transferências, que correspondem a técnicas que permitem visualizar estruturas internas, também é apresentada.

Avaliação in vivo do metabolismo do córtex pré-frontal dorsolateral em pacientes bipolares em episódio maníaco : um estudo com espectroscopia por ressonância magnética

Introdução: Diversas anormalidades neuroquímicas têm sido relatadas no Transtorno Bipolar (TB), mas os verdadeiros mecanismos envolvidos na fisiopatologia do TB permanecem a ser elucidados. A técnica de espectroscopia por ressonância magnética (1H-MRS) permite a mensuração de certos neurometabólitos no cérebro humano in vivo. Nós utilizamos a 1H-MRS para investigar o N-acetil-L-aspartato (NAA), compostos de colina (Cho), creatina/fosfocreatina (Cr) e o myoinositol (Ino) no córtex pré-frontal dorsolateral (CPFDL) em indivíduos bipolares durante episódio maníaco/misto. Métodos: Dez pacientes bipolares (9 maníacos, 1 misto), diagnosticados através de uma entrevista clínica semi-estruturada (SCID), e 10 voluntários normais pareados por sexo e idade foram estudados. Os neurometabólitos foram mensurados através de voxels de 8cm3 localizados no CPFDL direito e esquerdo para aquisição da 1H-MRS de 1.5T. Imagens de ressonância magnética anatômica ponderadas em T1 e T2 foram obtidas para excluir quaisquer anormalidades neuroanatômicas. Resultados: Não foram encontradas diferenças significativas para NAA, Cho, Cr, Ino, NAA/Cr, Cho/Cr, ou Ino/Cr entre pacientes e controles. Pacientes maníacos/mistos apresentaram níveis significativamente aumentados de myoinositol no CPFDL esquerdo em relação ao CPFDL direito (p = 0,044). Conclusões: Elevação do myoinositol no CPFDL esquerdo em pacientes bipolares durante mania aguda pode representar uma disfunção na via de sinalização do fosfatidilinositol. Estudos longitudinais com maior amostra avaliando o pré e pós-tratamento são necessários para melhor esclarecer este tema.

Utilizando mapas de conectividade fuzzy no desenvolvimento de algoritmos reparadores de imagens binárias 3D

A 3D binary image is considered well-composed if, and only if, the union of the faces shared by the foreground and background voxels of the image is a surface in R3. Wellcomposed images have some desirable topological properties, which allow us to simplify and optimize algorithms that are widely used in computer graphics, computer vision and image processing. These advantages have fostered the development of algorithms to repair bi-dimensional (2D) and three-dimensional (3D) images that are not well-composed. These algorithms are known as repairing algorithms. In this dissertation, we propose two repairing algorithms, one randomized and one deterministic. Both algorithms are capable of making topological repairs in 3D binary images, producing well-composed images similar to the original images. The key idea behind both algorithms is to iteratively change the assigned color of some points in the input image from 0 (background)to 1 (foreground) until the image becomes well-composed. The points whose colors are changed by the algorithms are chosen according to their values in the fuzzy connectivity map resulting from the image segmentation process. The use of the fuzzy connectivity map ensures that a subset of points chosen by the algorithm at any given iteration is the one with the least affinity with the background among all possible choices

Synchrotron microComputed tomography of the mature bovine dentinoenamel junction

The mature dentinoenamel junction (DEJ) is viewed by some investigators and the current authors, not as a fossilized, sharp transition between enamel and dentin, but as a relatively broad structural transition zone including the mantle dentin and the inner aprismatic enamel. In this study, the DEJ structure in bovine incisors was studied with synchrotron microComputed Tomography (microCT) using small cubes cut parallel to the tooth surface. The reconstructions revealed a zone of highly variable punctate contrast between bulk dentin and enamel; the mean linear attenuation coefficients and their standard deviations demonstrated that this zone averaged less mineral than dentin or enamel but had more highly variable structure than either. The region with the punctuate contrast is, therefore, the mantle dentin. The thickness of the mantle dentin seen in a typical data set was about 30 mu m, and the mantle dentin-enamel interface deviated +/- 15 mu m from the average plane over a distance of 520 mu m. In the highest resolution data (similar to 1.5 mu m isotropic voxels, volume elements), tubules in the dentin could be discerned in the vicinity of the DEJ. Contrast sensitivity was high enough to detect differences in mineral content between near-surface and near-DEJ volumes of the enamel. Reconstructions before and after two cubes were compressed to failure revealed cracks formed only in the enamel and did not propagate across the mantle dentin, regardless of whether loading was parallel to or perpendicular to the DEJ. (C) 2007 Elsevier B.V. All rights reserved.

Phosphorus magnetic resonance spectroscopy in malformations of cortical development

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)