Traditional Magnetic Resonance Imaging (MRI) reconstruction methods, which may be highly time-consuming and sensitive to noise, heavily depend on solving nonlinear optimization problems. By contrast, deep learning (DL)-based reconstruction methods do not need any explicit analytical data model and are robust to noise due to its large data-based training, which both make DL a versatile tool for fast and high-fidelity MR image reconstruction. While DL can be performed completely independently of traditional methods, it can, in fact, benefit from incorporating these established methods to achieve better results. To test this hypothesis, we proposed a hybrid DL-based MR image reconstruction method, which combines two state-of-the-art deep learning networks, U-Net and Generative Adversarial Network with Cycle loss (CycleGAN), with a traditional data reconstruction method: Projection Onto Convex Sets (POCS). Experiments were then performed to evaluate the method by comparing it to several existing state-of-the-art methods. Our results demonstrate that the proposed method outperformed the current state-of-the-art in terms of higher peak signal-to-noise ratio (PSNR) and higher Structural Similarity Index (SSIM). / Electrical and Computer Engineering

No description available.

No description available.

Les micro-saignements cérébraux (MSC) sont des dépôts d’hémosidérine particulièrement visibles sur des séquences IRM sensibles à la susceptibilité magnétique, comme par exemple la séquence en écho de gradient pondérée (GRE) en T2*. Néanmoins, leur détection in-vivo à partir de l’image d'amplitude GRE T2* obtenue en routine clinique est peu exacte et très sensible aux paramètres d’acquisition. Une détection automatique des MSC permettrait d’augmenter la portée et la pertinence de cette séquence, mais il est avant tout nécessaire de mieux caractériser les MSC pour augmenter la spécificité de détection. Cette thèse présente une nouvelle méthode pour extraire l’information d’intérêt (la carte champ interne) à partir d’images de phase obtenues avec les mêmes acquisitions cliniques GRE T2*. En effet, l’image de phase T2* contient directement une information sur la susceptibilité et pourrait donc améliorer la détection des MSC. Cette méthode a été évaluée avec succès sur des données multicentriques de qualité compatible avec la routine clinique pour des sujets avec un nombre très variable de MSC et a permis de différencier les MSC des micro-calcifications. Une étude de validation a également été menée pour évaluer l’utilité clinique de la carte de champ interne pour la détection par un expert, par rapport à d’autres types d’images et reconstructions utilisées en clinique. Elle a montré une amélioration de la spécificité. La thèse comprend également une preuve de concept d’une méthode d’identification automatique utilisant l’information provenant de plusieurs types d’image et de reconstruction afin d’augmenter la spécificité de l’identification. L’évaluation est menée sur les sujets précédemment décrits. Cette preuve de concept est basée sur un algorithme d’apprentissage supervisé ; cela consiste à combiner les informations issues des différents types d’image à partir desquels des descripteurs d’intensités et de forme ont été extraits pour créer un modèle de prédiction permettant discriminer les MSC. / CMBs, small hypo-intense foci with a maximum diameter of 10 millimeters, were first thought clinically silent. They are now considered as an imaging marker of cerebral small vessel diseases and their clinical involvement is increasingly recognized; they may be associated with an increased risk of hemorrhagic stroke, ischemia and dementia such as Alzheimer's disease. However, their relation with pathology and its causality remains largely to be understood, partly because of their tricky characterization in-vivo. Large scale studies or meta-analyses are made difficult because their identification varies with MRI sequence parameters and suffers from reproducibility issues and is time-consuming. Automatic identification methods have been proposed to address these issues but they all require manual post processing selection steps, because of a very high number of false positives. This suggests that a better characterization of CMBs may be the key to improve their detection, as it would allow better identifying them from misleading structures and lesions.This PhD focused on achieving a better characterization of CMBs to better detect them with an automatic method. It covers multiple aspects to improve CMBs identification. First, MR phase image was taken into account in addition to the standard MR magnitude image, because of its sensitivity to CMBs. A new MR phase image processing technique was thus developed to obtain the magnetic field of interest free of contamination from background sources in datasets equivalent to clinical routine. A comparison study was carried-out to evaluate the outcome of this tool for CMBs detection in a standardized dataset in a clinical environment. A proof-of-concept is given to illustrate the advantages of new features for automatically identifying CMBs.

Brain tumors present the second highest cause of death among pediatric cancers. About 60% are located in the posterior fossa region of the brain; among the most frequent types the ones considered for this project were astrocytomas, medulloblastomas, and ependymomas. Diagnosis can be done either through invasive histopathology exams or by non-invasive magnetic resonance (MR) scans. The tumors listed can be difficult to diagnose, even for trained radiologists, so machine learning methods, in particular deep learning, can be useful in helping to assess a diagnosis. Deep learning has been investigated only in a few other studies.The dataset used included 115 different subjects, some with multiple scan sessions, for which there were 142 T2-w, 119 T1Gd-w, and 89 volumes that presented both MR modalities. 2D slices have been manually extracted from the registered and skull-stripped volumes in the transversal, sagittal, and frontal anatomical plane and have been preprocessed by normalizing them and selecting the slices containing the tumor. The scans employed are T2-w, T1Gd-w, and a combination of the two referred to as multimodal images. The images were divided session-wise into training, validation, and testing, using stratified cross-validation and have also been augmented. The convolutional neural networks (CNN) investigated were ResNet50, VGG16, and MobileNetV2. The model performances were evaluated for two-class and three-class classification tasks by computing the confusion matrix, accuracy, receiver operating characteristic curve (ROC), the area under the curve (AUROC), and F1-score. Moreover,  explanations for the behavior of networks were investigated using GradCAMs and occlusion maps. Preliminary investigations showed that the best plane and modality were the transversal one and T2-w images. Overall the best model was VGG16, for the two-class tasks the best classification was between astrocytomas and medulloblastomas which reached an F1-score of 0.86 for both classes on multimodal images, followed by astrocytomas and ependymomas with an F1-score of 0.76 for astrocytomas and 0.74 for ependymomas on T2-w, and last F1-score of 0.30 for ependymomas and 0.65 for medulloblastomas on multimodal images. The three-class classification reached F1-score values of 0.59 for astrocytomas, 0.46 for ependymomas, and 0.64 for medulloblastomas on T2-w images. GradCAMs and occlusion maps showed that VGG16 was able to focus mostly on the tumor region but that there also seemed to be other information in the background of the images that contributed to the final classification.To conclude, the classification of infratentorial pediatric brain tumors can be achieved with acceptable results by means of deep learning and using a single MR modality, though one might have to account for the dataset size, number of classes and class imbalance. GradCAMs and occlusion maps offer important insights into the decision process of the networks