Le remodelage des voies aériennes est un phénomène mal connu, dont le rôle est central dans la sévérité des maladies broncho-pulmonaires obstructives chroniques telles que l’asthme et la broncho-pneumopathie obstructive chronique (BPCO). La recherche est nécessaire pour comprendre les mécanismes et les conséquences fonctionnelles, qui diffèrent dans ces deux maladies. L’imagerie a le potentiel de pouvoir évaluer de façon non-invasive, in vivo, les mécanismes mis en oeuvre dans ces maladies tandis que la recherche de marqueurs pertinents est une nécessité. Dans l’asthme, chez l’animal, nous avons évalué la densité péribronchique dans des modèles murins de sensibilisation allergique. Ce marqueur a été démontré comme étant un reflet du remodelage des voies aériennes et pourrait être utilisé dans des études précliniques. Chez l’homme, nous avons d’abord optimisé les performances d’un logiciel de segmentation 3D des volumes bronchiques, en termes d’automatisation, détection de bronches distales et squelettisation. Ce logiciel a été appliqué lors d’une étude thérapeutique en double aveugle versus placebo pour évaluer l’effet du gallopamil dans l’asthme sévère. Nos résultats montrent que le remodelage musculaire lisse est une composante clé du remodelage dans l’asthme, et que le TDM quantitatif permet d’évaluer les effets de molécules lors d’essais cliniques. Dans la BPCO, nous avons mis en évidence à l’aide de la TDM quantitative que le remodelage bronchique est impliqué dans une complication particulière qui n’est pas mise en oeuvre dans l’asthme, à savoir le développement d’une hypertension artérielle pulmonaire. D’autre part, nous avons montré que la mesure automatique des petits vaisseaux permet une analyse complémentaire et supporte l’hypothèse d’un phénotype distinct de BPCO associé à une maladie vasculaire. Enfin l’IRM pulmonaire reste un challenge pour l’imagerie du fait de la faible teneur en protons et des mouvements cardiaque et respiratoire. Nous avons testé et optimisé une nouvelle séquence permettant d’obtenir des imageries de qualité très proches de celles du TDM. Ces résultats ouvrent des perspectives sur l’imagerie non irradiante des maladies broncho-pulmonaires chroniques par IRM. / Airway remodeling is a critical outcome in broncho-pulmonary obstructive disorders such as asthma, COPD and cystic fibrosis. Research is needed in order to better understand the pathophysiological process underlying these different diseases, as well as their functional significance and consesquences in vivo. Imaging allows non-invasive and quantitative assessment of the remodeling process in vivo. In asthma, we have assessed the value of peribronchial density using microCT in murine models of ovalbumine sensitization. This novel biomarker was shown to relate with remodeling than inflammation of airways and could be used in preclinical studies. In humans, we first optimized an existing chain of post treatment in order to segment in 3 dimensions bronchi volumes in order to improve automation, detection and skeletonization. The software was applied in a prospective randomized double-blinded study in severe asthma, in order to test the effect of a new therapeutic anticalcic agent. Results showed that the software enables a non-invasive assessment of airway smooth muscle remodeling in severe asthma during therapeutic studies. In COPD, we have shown that quantitative CT is able to unravel new complex mechanisms that are involved in COPD but not in asthma, i.e. the development of pulmonary hypertension. CT measurements of small vessels were also shown to add complementary information on COPD phenotypes, supporting the existence of distinct subtypes of COPD related to a vascular rather than a broncho-pulmonary disease. Finally, lung MRI is still a challenging field of investigation, owing to the very low proton density of bronchi, the presence of movement artifact. We have tested and optimized an innovative sequence combined with respiratory synchronization in order to get images in close agreement with CT. Perspectives related to this novel non-invasive, quantitative and radiation-free imaging technique are promising in the evaluation of broncho-pulmonary obstructive diseases.

<div>Fluorescence microscopy has emerged as a powerful tool for studying cell biology because it enables the acquisition of 3D image volumes deeper into tissue and the imaging of complex subcellular structures. Fluorescence microscopy images are frequently distorted by motion resulting from animal respiration and heartbeat which complicates the quantitative analysis of biological structures needed to characterize the structure and constituency of tissue volumes. This thesis describes a two pronged approach to quantitative analysis consisting of non-rigid registration and deep convolutional neural network segmentation. The proposed image registration method is capable of correcting motion artifacts in three dimensional fluorescence microscopy images collected over time. In particular, our method uses 3D B-Spline based nonrigid registration using a coarse-to-fine strategy to register stacks of images collected at different time intervals and 4D rigid registration to register 3D volumes over time. The results show that the proposed method has the ability of correcting global motion artifacts of sample tissues in four dimensional space, thereby revealing the motility of individual cells in the tissue.</div><div><br></div><div>We describe in thesis nuclei segmentation methods using deep convolutional neural networks, data augmentation to generate training images of different shapes and contrasts, a refinement process combining segmentation results of horizontal, frontal, and sagittal planes in a volume, and a watershed technique to enumerate the nuclei. Our results indicate that compared to 3D ground truth data, our method can successfully segment and count 3D nuclei. Furthermore, a microscopy image synthesis method based on spatially constrained cycle-consistent adversarial networks is used to efficiently generate training data. A 3D modified U-Net network is trained with a combination of Dice loss and binary cross entropy metrics to achieve accurate nuclei segmentation. A multi-task U-Net is utilized to resolve overlapping nuclei. This method was found to achieve high accuracy object-based and voxel-based evaluations.</div>