CETTE thèse est consacrée à la conception d’un système d’aide au diagnostic dédiéau muscle squelettique humain. Au cours du premier volet de ce manuscrit nousproposons une nouvelle représentation basée sur les modèles parcimonieux dans le cadrede la segmentation d’Images de Résonances Magnétiques (IRM) T1 du muscle squelettiquedu mollet. Notre méthode Sparse Shape Model/ Modèle de Formes Parcimonieux(MFP), apprend un modèle statistique de formes et de textures locales annoté et réussità en tirer une représentation réduite afin de reconstruire le mécanisme musculaire sur unexemple test. Dans la seconde partie du manuscrit, nous présentons une approche baséesur des ondelettes de diffusion pour la segmentation du muscle squelettique. Contrairementaux méthodes de l’état de l’art, notre approche au cours de la phase d’apprentissagepermet à optimiser les coefficients des ondelettes, ainsi que leur nombres et leur positions.Le modèle prend en charge aussi bien les hiérarchies dans l’espace de recherche,que l’encodage des dépendances géométriques complexes et photométriques de la structured’intérêt. Notre modélisation offre ainsi l’avantage de traiter des topologies arbitraires.L’évaluation expérimentale a été effectué sur un ensemble de mollets acquisespar un scanner IRM, ainsi qu’un ensemble d’images tomodensitométriques du ventriculegauche. / THE thesis is dedicated to three dimensional shape analysis and the segmentation ofhuman skeletal muscles in the context of myopathies and their treatment. In particular,we study the local and global structural characteristics of muscles. The methodologicalfocus of the thesis is to devise methods for the segmentation of muscles, theconsistent localization of positions in the anatomy and the navigation within the muscledata across patients. Currently diagnosis and follow-up examinations during therapy ofmyopathies are typically performed by means of biopsy. This has several disadvantages:it is an invasive method, covers only a small muscle region, is mainly restricted to diagnosticpurpose and is not suitable for follow-up evaluation. We develop the followingmethods to make the use of non-invasive imaging modalities such as MRI for a virtualbiopsy possible: first, a novel approach to model shape variations that encodes sparsity,exploits geometric redundancy, and accounts for the different degrees of local variationand image support in data. It makes the modeling and localization of muscles possible,that exhibit sparsely distributed salient imaging features, and heterogeneous shapevariability. Second, we extend the shape representation of 3D structures using diffusionwavelets. The proposed method can represent shape variation and exploits continuousinter-dependencies of arbitrary topology in the shape data. We then explore several approachesfor the shape model search, and appearance representation based on boostingtechniques and canonical correlation analysis. Last we present a robust diffusion wavelettechnique that covers the integration of our two shape models approaches to finally getan enhanced sparse wavelet based method. We validate the approaches on two medicalimaging data sets that represent the properties tackled by the approaches: T1 weightedMRI data of full calf muscles and computed tomography data of the left heart ventricle.

As pragas e doenças apresentam-se como um desafio para a citricultura brasileira em razão do impacto económico que elas causam à produção. Neste trabalho é dado destaque à doença da mancha preta (MPC), causada pelo fungo Guignardia citricarpa. Essa doença provoca lesões no fruto, depreciando-o no mercado de frutas in natura, além de causar amadurecimento e queda precoce. Um dos principais agravantes da doença é a demora no aparecimento dos sintomas, sendo muito importante detectar a presença dos esporos do fungo no pomar, antes que os sintomas apareçam. Dessa maneira, há a possibilidade de se controlar a doença de forma eficaz, aplicando-se quantidades menores de fungicidas e, consequentemente, reduzindo os custos da produção e os efeitos deletérios ao meio-ambiente. Atualmente, a detecção desses esporos é realizada por meio da análise de amostras coletadas nos pomares. Essa análise é efetuada por especialistas que realizam a identificação e a contagem dos ascósporos manualmente. Com o objetivo de automatizar esse processo, um conjunto de técnicas para a análise das imagens e a caracterização dos ascósporos do fungo a partir da forma foi estudado e comparado. Dentre as técnicas, a curvatura e os descritores de Fourier apresentaram resultados bastante satisfatórios e foram utilizados na implementação do protótipo de um sistema de visão computacional - o CITRUSVIS, que analisa e identifica os ascósporos existentes nas imagens dos discos de coleta. / The pest and disease management is one of the significant factors in the citrus culture. This work focuses on the black spot disease ( C B S ) , a fungai disease caused by Guignardia citricarpa that occasions sunken lesions in the rind of fruits causing precocious maturation, accented fali, depreciation for in natura fruit market and increase of the production costs for disease controlling. One of the main problems to control the CBS disease is the delay to appearance of symptom (when the orchard is already infected), and the fungai presence identification is necessary as soon as possible, allowing the appliance of procedures to control it. Nowadays, spores identification, particularly the ascospores (sexual spores), is made by collecting suspended particles in orchards blown on discs, which are analyzed by specialists using the microscope. The use of a computer aided vision system to assist the spores identification is one of the strategies to speed up this process. In this work, methods to analyze and characterize the spores, based on its shape, were studied and compared. Among them, the shape curvature method and the Fourier descriptors, chosen for presenting the best result, were implemented in a system - the CITRUS Vis - to analyze the images and identify the ascospores.

A identificação de unia planta exige, pelos padrões de taxionomia vegetal, a análise de folhas, flores e frutos. O projeto TreeVis surge com uma proposta de auxiliar na identificação de espécies vegetais, por meio do uso de métodos biométricos, a partir da análise de alguns atributos de uma folha. A contribuição inicial deste trabalho de mestrado, para o projeto TreeVis, está obtenção de classificadores por meio do uso de assinaturas de contorno, sob o domínio da frequência, possibilitando a composição de diversos tipos de assinaturas e classificadores para uma mesma espécie. Devido à baixa eficiência obtida por métodos de classificação como distância mínima, optou-se pelo uso de redes neurais. Essa abordagem evidenciou a necessidade de solução de dois problemas: o grande número de possibilidades de composição de sinais o que ocasionaria um grande esforço computacional para a obtenção de todas respectivas redes neurais; e o reduzido número das amostras utilizadas no trabalho - o qual comprometeria as etapas de treinamento e teste de uma rede neural. Para a solução desses problemas, foram desenvolvidos dois métodos: o primeiro método identifica e seleciona as assinaturas que apresentam um maior potencial de sucesso em obter um classificador por meio de redes neurais, solucionando o problema e desperdício de esforço computacional; o segundo método possibilita a geração de amostras artificiais de folhas através da combinação dos espectros de frequência do contorno das amostras reais por meio operadores genéticos de cross-over e mutação. Solucionadas as duas questões, foram obtidas diversas redes neurais, através da indicação das assinaturas de melhor potencial e treinadas com amostras artificiais. Do total de 31 classes, 7 foram descartadas da tentativa de obtenção de classificadores por não apresentarem nenhuma assinatura com potencial de classificação - conforme indicação do método desenvolvido. Das 24 espécies restantes, foram obtidos classificadores para 18 espécies (75%) com taxas médias de 85% de acerto. A execução deste trabalho necessitou do desenvolvimento de um arcabouço para a automatização da geração, treinamento e teste das redes neurais. / The vegetable identifieation is done, in vegetal taxonomy standards, by fiower, fruits and leaves analyses. The TreeVis project proposes to identifv vegetal speeiniens by biometric methods using only same leaf features. The contribution of this work for to TreeVis project is the generation of classifiers by the contour signatures, under frequency domain, niaking be able the coniposition of several types of signatures and classifiers for the same speeimen. Because of poor efficiency results from methods like minimal distance, was chosen to use neural networks. This approach showed the need to solve two probleins: the numerous composition possibilities of signatures - that would be need a big computational effort to obtained ali possible neural networks; and the small number of speeimen samples - that would compromise the training and test. of neural networks. To solve these two probleins was developed two methods: The first identify and select the signatures that have a good pattern recognition potential, before of the network will be done, solving the waste unneeded effort problem. The second method proposed produces artificial leaf sliapes by combination of contour spectrum frequency speeiniens of real leaves, using genetic operators like cross-over and mutation. Solved these probleins several networks was obtained by appointed potential signature methods and trained and tested with artificial leaves. From 31 speeiniens class, 07 were discarded because tliey had not signatures with classification potential - indicated by developed method. From 24 classes remaining were obtained classifiers for 18 classes (75%) with médium rates 85% of set riglit. The execution of this work demanded the construction of a framework to automatize the generation, training and test of the neural networks.

The Unruh effect is one of the most startling predictions of quantum field theory. Its interpretation has been controversially discussed, since the first publications of Fulling, Davies and Unruh in the 1970ties. In a recent paper Buchholz and Solveen proposed an application of basic thermodynamic definitions to clarify the meaning of temperature and thermal equilibrium in the Unruh effect. As a result the interpretation of the KMS-parameter as an expression of local temperature has been questioned. The main result of my diploma thesis asserts quasi-equivalence of the disputed KMS states on a subregion of Rindlerspace that infinitely extends in the direction of travel of a uniformly accelerated Rindler-observer. Exploring the consequences of this result, I will present new insights on the asymptotic behaviour of such KMS states and how this fits into the picture drawn by Buchholz and Solveen.

This dissertation presents some of the recent developments in the modelling of shape spaces. Forming the basis for a quantitative analysis of shapes, this is relevant for many applications involving image recognition and shape classification. All shape spaces discussed in this work arise from the general situation of a Lie group acting isometrically on some Riemannian manifold. The first chapter summarizes the most important results about this general set-up, which are well known in other branches of mathematics. A particular focus is laid on Hamiltonian methods that explore the relation of symmetry and conserved momenta. As a classical example these results are applied to Kendall’s shape space. More recent approaches of continuous shape models are then summarized and put in the same concise framework. In more detail the square root velocity shape representation, recently developed by Srivastava et al., is being discussed. In particular, the phenomenon of unclosed orbits under the action of reparametrization is addressed. This issue is partially resolved by an extended equivalence relation along with a well defined, non-degenerate, metric on the resulting quotient space.

Breast cancer is the second leading cause of cancer deaths among women in the United States and microcalcifications clusters are one of the most important indicators of breast disease. Computer methodologies help in the detection and differentiation between benign and malignant lesions and have the potential to improve radiologists' performance and breast cancer diagnosis significantly. A Computer-Aided Diagnosis (CAD-Dx) algorithm has been previously developed to assist radiologists in the diagnosis of mammographic clusters of calcifications with the modules: (a) detection of all calcification-like areas, (b) false-positive reduction and segmentation of the detected calcifications, (c) selection of morphological and distributional features and (d) classification of the clusters. Classification was based on an artificial neural network (ANN) with 14 input features and assigned a likelihood of malignancy to each cluster. The purpose of this work was threefold: (a) optimize the existing algorithm and test on a large database, (b) rank classification features and select the best feature set, and (c) determine the impact of single and two-view feature estimation on classification and feature ranking. Classification performance was evaluated with the NevProp4 artificial neural network trained with the leave-one-out resampling technique. Sequential forward selection was used for feature selection and ranking. Mammograms from 136 patients, containing single or two views of a breast with calcification cluster were digitized at 60 microns and 16 bits per pixel. 260 regions of interest (ROI's) centered on calcification cluster were defined to build the single-view dataset. 100 of the 136 patients had a two-view mammogram which yielded 202 ROI's that formed the two-view dataset. Classification and feature selection were evaluated with both these datasets. To decide on the optimal features for two-view feature estimation several combinations of CC and MLO view features were attempted. On the single-view dataset the classifier achieved an AZ =0.8891 with 88% sensitivity and 77% specificity at an operating point of 0.4; 12 features were selected as the most important. With the two-view dataset, the classifier achieved a higher performance with an AZ =0.9580 and sensitivity and specificity of 98% and 80% respectively at an operating point of 0.4; 10 features were selected as the most important.

No description available.

Realistic assessments of liver injury risk for the entire occupant population require incorporating inter-subject variations into numerical human models. The main objective of this study was to quantify the variations in shape and material properties of the human liver. Statistical shape analysis was applied to analyze the geometrical variation using a surface set of 15 adult human livers recorded in an occupant posture. Principal component analysis was then utilized to obtain the modes of variation, the mean model, and a set of 95% statistical boundary shape models. Specimen-specific finite element (FE) models were employed to quantify material and failure properties of human liver parenchyma. The mean material model parameters were then determined, and a stochastic optimization approach was utilized to determine the standard deviations of the material model parameters. The distributions of the material parameters were used to develop probabilistic FE models of the liver implemented in THUMS human FE model to simulate oblique impact tests under three impact speeds. In addition, the influence of organ preservation on the biomechanical responses of animal livers was investigated using indentation and tensile tests. Results showed that the first five modes of the human liver shape models accounted for more than 70% of the overall anatomical variations. The Ogden material model with two parameters showed a good fit to experimental tensile data before failure. Significant changes of the biomechanical responses of liver parenchyma were found after cooling or freezing storage. The force-deflection responses of THUMS model with probabilistic liver material models were within the test corridors obtained from cadaveric tests. Significant differences were observed in the maximum and minimum principal Green-Lagrangian strain values recorded in the THUMS liver model with the default and updated average material properties. The results from this study could help in the development of more biofidelic human models, which may provide a better understanding of injury mechanisms of the liver during automobile collisions. / Ph. D.

This doctoral thesis considers the automatic verification of parameterized systems, i.e. systems with an arbitrary number of communicating components, such as mutual exclusion protocols, cache coherence protocols or heap manipulating programs. The components may be organized in various topologies such as words, multisets, rings, or trees. The task is to show correctness regardless of the size of the system and we consider two methods to prove safety:(i) a backward reachability analysis, using the well-quasi ordered framework and monotonic abstraction, and (ii) a forward analysis which only needs to inspect a small number of components in order to show correctness of the whole system. The latter relies on an abstraction function that views the system from the perspective of a fixed number of components. The abstraction is used during the verification procedure in order to dynamically detect cut-off points beyond which the search of the state-space need not continue. Our experimentation on a variety of benchmarks demonstrate that the method is highly efficient and that it works well even for classes of systems with undecidable property. It has been, for example, successfully applied to verify a fine-grained model of Szymanski's mutual exclusion protocol. Finally, we applied the methods to solve the complex problem of verifying highly concurrent data-structures, in a challenging setting: We do not a priori bound the number of threads, the size of the data-structure, the domain of the data to store nor do we require the presence of a garbage collector. We successfully verified the concurrent Treiber's stack and Michael & Scott's queue, in the aforementioned setting. To the best of our knowledge, these verification problems have been considered challenging in the parameterized verification community and could not be carried out automatically by other existing methods.

Cette thèse présente une méthodologie appliquée à la caractérisation de la dynamique de structures déformables sur des séquences temporelles (2D+t). Des indicateurs sont proposés pour estimer la mobilité de formes non-rigides, à partir de leurs contours. Deux approches complémentaires sont développées: En premier lieu, les descripteurs de forme sont utilisés pour quantifier les déformations globales des formes, et pour estimer des repères géométriques spécifiques. La deuxième approche repose sur l'appariement difféomorphique pour déterminer une paramétrisation unifiée des formes, afin de décrire les déformations. Une évaluation permet d'apprécier la qualité des indicateurs en termes de coût algorithmique, de robustesse face aux données altérées, et de capacité à différencier deux séquences.Cette approche de caractérisation est appliquée à des séquences IRM dynamiques de la cavité pelvienne, où les principaux organes pelviens (vessie, utérus-vagin, rectum) ont une grande variabilité morphologique, ils se déplacent et se déforment. Cette caractérisation est validée dans le cadre de deux applications. L'analyse statistique effectuée sur un ensemble de séquences permet de mettre en évidence des comportements caractéristiques des organes, d'identifier des références anatomiquement significatives, et d'aider à l'interprétation des diagnostics des organes. Aussi, dans le contexte de la réalisation d'une modélisation de la dynamique pelvienne patiente-spécifique, la caractérisation vise à évaluer quantitativement la précision de la modélisation, en utilisant l'IRM dynamique comme vérité-terrain. Ainsi, elle apporte des indications sur la correction des paramètres du modèle. / This thesis presents a methodology for the characterization of the dynamics of deformable structures on time-series data (2D+t). Some indicators are proposed in order to estimate non-rigid shape variations from their contours. Two complementary approaches are developed : First, shape descriptors are used to quantify the global deformations of the shapes, and to estimate specific geometric references. The second approach relies on the diffeomorphic mapping to determinate a unified parametrization of the shapes. Then, features are used to describe the deformations locally. Furthermore, the methodology has an evaluation step which consists in the assessment of the quality of the indicators in the algorithmic complexity, in the stability against data with a small variability, and in the ability to differentiate two sequences.The characterization is applied to dynamic MRI sequences of the pelvic cavity, where the main pelvic organs (bladder, uterus-vagina, rectum) have a high morphological variability, they undergo displacements and deformations. The characterization is validated within the context of two applications. Firslty, a statistical analysis is carried out on a set of sequences. It allows to highlight some properties of the organ behaviors, and to identify meaningful anatomical landmarks. The analysis helps also for the automatic interpretation of the organ diagnoses. Secondly, within the context of the development of a patient-specific pelvic dynamics modeling system, the characterization aims at assessing quantitatively the modeling precision. It uses the dynamic MRI as a ground truth. Thereby, it brings some clues about the correction of the model parameters.