Medical imaging has seen a rapid development in its clinical use in assessment of treatment outcome, disease monitoring and diagnosis over the last few decades. Yet, the vast amount of available image data limits the practical use of this potentially very valuable source of information for radiologists and physicians. Therefore, the design of computer-aided medical image analysis is of great importance to imaging in clinical practice. This thesis deals with the problem of deformable image registration in the context of lung imaging, and addresses three of the major challenges involved in this challenging application, namely: designing an image similarity for multi-modal scans or scans of locally changing contrast, modelling of complex lung motion, which includes sliding motion, and approximately globally optimal mathematical optimisation to deal with large motion of small anatomical features. The two most important contributions made in this thesis are: the formulation of a multi-dimensional structural image representation, which is independent of modality, robust to intensity distortions and very discriminative for different image features, and a discrete optimisation framework, based on an image-adaptive graph structure, which enables a very efficient optimisation of large dense displacement spaces and deals well with sliding motion. The derived methods are applied to two different clinical applications in pulmonary image analysis: motion correction for breathing-cycle computed tomography (CT) volumes, and deformable multi-modal fusion of CT and magnetic resonance imaging chest scans. The experimental validation demonstrates improved registration accuracy, a high quality of the estimated deformations, and much lower computational complexity, all compared to several state-of-the-art deformable registration techniques.

Labeling problem, due to its versatile modeling ability, is widely used in various image analysis tasks. In practice, certain prior information is often available to be embedded in the model to increase accuracy and robustness. However, it is not always straightforward to formulate the problem so that the prior information is correctly incorporated. It is even more challenging that the proposed model admits efficient algorithms to find globally optimal solution. In this dissertation, a series of natural and medical image segmentation tasks are modeled as labeling problems. Each proposed model incorporates different useful prior information. These prior information includes ordering constraints between certain labels, soft user input enforcement, multi-scale context between over-segmented regions and original voxel, multi-modality context prior, location context between multiple modalities, star-shape prior, and gradient vector flow shape prior. With judicious exploitation of each problem's intricate structure, efficient and exact algorithms are designed for all proposed models. The efficient computation allow the proposed models to be applied on large natural and medical image datasets using small memory footprint and reasonable time assumption. The global optimality guarantee makes the methods robust to local noise and easy to debug. The proposed models and algorithms are validated on multiple experiments, using both natural and medical images. Promising and competitive results are shown when compared to state-of-art.

This thesis presents a novel set of image analysis tools developed for the purpose of assisting radiologists with the task of detecting and characterizing breast lesions in image data acquired using magnetic resonance imaging (MRI). MRI is increasingly being used in the clinical setting as an adjunct to x-ray mammography (which is, itself, the basis of breast cancer screening programs worldwide) and ultrasound. Of these imaging modalities, MRI has the highest sensitivity to invasive cancer and to multifocal disease. MRI is the most reliable method for assessing tumour size and extent compared to the gold standard histopathology. It also shows great promise for the improved screening of younger women (with denser, more radio opaque breasts) and, potentially, for women at high risk. Breast MRI presently has two major shortcomings. First, although its sensitivity is high its specificity is relatively poor; i.e. the method detects many false positives. Second, the method involves acquiring several high-resolution image volumes before, during and after the injection of a contrast agent. The large volume of data makes the task of interpretation by the radiologist both complex and time-consuming. These shortcomings have motivated the research and development of the computer-aided detection systems designed to improve the efficiency and accuracy of interpretation by the radiologist. Whilst such systems have helped to improve the sensitivity/specificity of interpretation, it is the premise of this thesis that further gains are possible through automated image analysis. However, the automated analysis of breast MRI presents several technical challenges. This thesis investigates several of these, noise filtering, parametric modelling of contrast enhancement, segmentation of suspicious tissue and quantitative characterisation and classification of suspicious lesions. In relation to noise filtering, a new denoising algorithm for dynamic contrast-enhanced (DCE-MRI) data is presented, called the Dynamic Non-Local Means (DNLM). The DCE-MR image data is inherently contaminated by Rician noise and, additionally, the limited acquisition time per volume and the use of fat-suppression diminishes the signal-to-noise ratio. The DNLM algorithm, specifically designed for the DCE-MRI, is able to attenuate this noise by exploiting the redundancy of the information between the different temporal volumes, while taking into account the contrast enhancement of the tissue. Empirical results show that the algorithm more effectively attenuates noise in the DCE-MRI data than any of the previously proposed algorithms. In relation to parametric modelling of contrast enhancement, a new empiric model of contrast enhancement has been developed that is parsimonious in form. The proposed model serves as the basis for the segmentation and feature extraction algorithms presented in the thesis. In contrast to pharmacokinetic models, the proposed model does not rely on measured parameters or constants relating to the type or density of the tissue. It also does not assume a particular relationship between the observed changes in signal intensity and the concentration of the contrast agent. Empirical results demonstrate that the proposed model fits real data better than either the Tofts or Brix models and equally as well as the more complicated Hayton model. In relation to the automatic segmentation of suspicious lesions, a novel method is presented, based on seeded region growing and merging, using criteria based on both the original image MR values and the fitted parameters of the proposed model of contrast enhancement. Empirical results demonstrate the efficacy of the method, both as a tool to assist the clinician with the task of locating suspicious tissue and for extracting quantitative features. Finally, in relation to the quantitative characterisation and classification of suspicious lesions, a novel classifier (i.e. a set of features together with a classification method) is presented. Features were extracted from noise-filtered and segmented-image volumes and were based both on well-known features and several new ones (principally, on the proposed model of contrast enhancement). Empirical results, based on routine clinical breast MRI data, show that the resulting classifier performs better than other such classifiers reported in the literature. Therefore, this thesis demonstrates that improvements in both sensitivity and specificity are possible through automated image analysis.

Cette thèse présente les travaux réalisés en vue de l’élaboration d’un modèle cardiaque associant croissance, mouvement et circulation sanguine pour permettre ensuite la construction d’un modèle patient à partir d’un modèle de population. Le premier axe de ce travail est la simulation de la croissance bi-ventriculaire. Un modèle existant de surface unique, calculé à l’aide de méthodes statistiques, a été généralisé à un modèle bi-ventriculaire puis appliqué à la tétralogie de Fallot (ToF). Le deuxième axe concerne la modélisation du mouvement cardiaque au niveau de la population. Un modèle d’ordre réduit basé sur un modèle Polyaffine et LogDemons a été proposé. Il simule la dynamique cardiaque avec peu de paramètres. Les paramètres de transformation sont analysés par des méthodes statistiques. Un modèle de mouvement moyen a été calculé pour représenter le mouvement standard de la population. Le troisième axe s'intéresse à la simulation de l’écoulement sanguin à l’échelle de la population. La complexité des simulations spécifiques à un patient a été réduite grâce à l’utilisation de méthodes d’analyse d’image, de dynamique des fluides numérique et de réduction d’ordre de modèle. La simulation du flux sanguin dans l’artère pulmonaire pour des patients ToF a permis de mieux comprendre l’impact du sang régurgité sur la pression et la vitesse. Étant donné nos contributions sur ces trois axes, nous sommes maintenant en bonne position pour élaborer le modèle couplé des contributions interdépendantes de la croissance, du mouvement et de l'écoulement sanguin. Ce modèle pourrait être utilisé afin d'aider la planification de la thérapie chez les patients atteints de maladies cardiaques. / This thesis presents work towards a coupled model of cardiac growth, motion, and blood flow to enable predictive patient-specific models to be built from a population-based model. The first axis of this work is to simulate bi-ventricular growth through aging. A previously proposed single surface model computed using statistical methods was extended to a bi-ventricular model and applied to Tetralogy of Fallot patients to model the complex evolution of the ventricles due to the pathology. The second axis concerns the development of a model to simulate cardiac motion at a population level. A reduced-order cardiac-specific motion model was proposed to simulate the motion dynamics with a small number of parameters using a Polyaffine and LogDemons based model. From the computed transformations, the parameters were analysed using statistical methods to obtain population-based measures of normality. A mean motion model was derived to represent the normal motion for a given population. The third axis is to develop a model of population-based flow dynamics. The complexity of patient-specific simulations was reduced by combining image analysis, computational fluid dynamics and model order reduction techniques. Blood flow through the pulmonary artery in Tetralogy of Fallot patients was simulated to better understand the impact of regurgitated blood on pressure and velocity. Given our contributions on these three axes, we are now in a good position to couple the models in order to capture the interrelated contributions of growth, motion and flow. Such a model could be used to aid in therapy planning and decision making for patients with heart disease.

abstract: In recent years, Convolutional Neural Networks (CNNs) have been widely used in not only the computer vision community but also within the medical imaging community. Specifically, the use of pre-trained CNNs on large-scale datasets (e.g., ImageNet) via transfer learning for a variety of medical imaging applications, has become the de facto standard within both communities. However, to fit the current paradigm, 3D imaging tasks have to be reformulated and solved in 2D, losing rich 3D contextual information. Moreover, pre-trained models on natural images never see any biomedical images and do not have knowledge about anatomical structures present in medical images. To overcome the above limitations, this thesis proposes an image out-painting self-supervised proxy task to develop pre-trained models directly from medical images without utilizing systematic annotations. The idea is to randomly mask an image and train the model to predict the missing region. It is demonstrated that by predicting missing anatomical structures when seeing only parts of the image, the model will learn generic representation yielding better performance on various medical imaging applications via transfer learning. The extensive experiments demonstrate that the proposed proxy task outperforms training from scratch in six out of seven medical imaging applications covering 2D and 3D classification and segmentation. Moreover, image out-painting proxy task offers competitive performance to state-of-the-art models pre-trained on ImageNet and other self-supervised baselines such as in-painting. Owing to its outstanding performance, out-painting is utilized as one of the self-supervised proxy tasks to provide generic 3D pre-trained models for medical image analysis. / Dissertation/Thesis / Masters Thesis Computer Science 2020

Táto práca sa zaoberá určením neistoty v predikciách modelov hlbokého učenia. Aj keď sa týmto modelom darí dosahovať vynikajúce výsledky v mnohých oblastiach počítačového videnia, ich výstupy sú väčšinou deterministické a neposkytujú mnoho informácií o tom, ako si je model istý svojou predpoveďou. To je obzvlášť dôležité pri analýze lekárskych obrazových dát, kde môžu mať omyly vysokú cenu a schopnosť detekovať neisté predikcie by umožnila dohliadajúcemu lekárovi spracovať relevantné prípady manuálne. V tejto práci aplikujem niekoľko rôznych metrík vyvinutých v nedávnom výskume pre určenie neistoty na modely hlbokého učenia natrénované pre lokalizáciu cefalometrických landmarkov. Následne ich vyhodnotím a porovnávam v sade experimentov, ktorých úlohou je určiť, nakoľko jednotlivé metriky poskytujú užitočnú informáciu o tom, ako si je model istý svojou predpoveďou.

Cancer is one of the leading causes of death in the world, but advances in research and development of treatment methods is constantly ongoing to reduce the number of deaths and the amount of suffering. One of many approaches is radiation therapy, which uses high doses of radiation to kill tumors. Radiation therapy requires advanced software in image analysis to create careful treatment plans, evaluate treatment responses and to perform dose accumulation, among other things. One important tool for this is deformable image registration (DIR) which is used to find a correspondence between the images. The aim with this master thesis is to improve the DIR method ANACONDA by automatically provide additional information to the algorithm before the registration is performed.This work focuses on the registration of internal tubular structures in lung and liver images (bronchial and vascular tree, respectively). Two challenges in registering lung images are the sliding motion of lung surfaces and large motion of small internal structures. Several DIR methods have been proposed for solving the challenging internal structures, however most of them do not take into account the alignment of surrounding tissues. DIR methods applied to the liver are published less frequently, but accurate registration of the liver is of high interest since, for example, knowledge of the anatomy of the vascular tree is essential when removing tumors through liver surgeries. In this work, corresponding (anatomical) points are automatically found in two images and added to the DIR algorithm. The points are found by extracting and comparing the tubular structures between the images, and with use of different distance requirements, nearby points are paired.The new method manages to achieve good registration of both internal structures and surrounding tissue. Mean target registration errors for the internal structures of lungs was 1.11 ± 0.75 and for liver 1.67 ± 1.15 mm.

Modern medical imaging techniques have revolutionized health care over the last decades, providing clinicians with high-resolution 3D images of the inside of the patient's body without the need for invasive procedures. Detailed images of the vascular anatomy can be captured by angiography, providing a valuable source of information when deciding whether a vascular intervention is needed, for planning treatment, and for analyzing the success of therapy. However, increasing level of detail in the images, together with a wide availability of imaging devices, lead to an urgent need for automated techniques for image segmentation and analysis in order to assist the clinicians in performing a fast and accurate examination. To reduce the need for user interaction and increase the speed of vascular segmentation,  we propose a fast and fully automatic vascular skeleton extraction algorithm. This algorithm first analyzes the volume's intensity histogram in order to automatically adapt the internal parameters to each patient and then it produces an approximate skeleton of the patient's vasculature. The skeleton can serve as a seed region for subsequent surface extraction algorithms. Further improvements of the skeleton extraction algorithm include the expansion to detect the skeleton of diseased arteries and the design of a convolutional neural network classifier that reduces false positive detections of vascular cross-sections. In addition to the complete skeleton extraction algorithm, the thesis presents a segmentation algorithm based on modified onion-kernel region growing. It initiates the growing from the previously extracted skeleton and provides a rapid binary segmentation of tubular structures. To provide the possibility of extracting precise measurements from this segmentation we introduce a method for obtaining a segmentation with subpixel precision out of the binary segmentation and the original image. This method is especially suited for thin and elongated structures, such as vessels, since it does not shrink the long protrusions. The method supports both 2D and 3D image data. The methods were validated on real computed tomography datasets and are primarily intended for applications in vascular segmentation, however, they are robust enough to work with other anatomical tree structures after adequate parameter adjustment, which was demonstrated on an airway-tree segmentation.

Submitted by Rosivalda Pereira (mrs.pereira@ufma.br) on 2017-09-04T18:08:18Z No. of bitstreams: 1 LuisClaudioSilva.pdf: 4393489 bytes, checksum: 5c3cb715bd7992ba1cc5380c1a940f32 (MD5) / Made available in DSpace on 2017-09-04T18:08:18Z (GMT). No. of bitstreams: 1 LuisClaudioSilva.pdf: 4393489 bytes, checksum: 5c3cb715bd7992ba1cc5380c1a940f32 (MD5) Previous issue date: 2017-07-27 / Breast cancer is the second type of cancer that most a ects women in the world, losing only for non melanoma skin cancer. Breast density can hinder the location of masses, especially in early stages. In this work, the use of independent component analysis for detecting and segmentation lesions in dense breasts is proposed. Several works suggests the use of computer aided diagnosis, increasing sensitivity to over 90% in detecting cancer in non dense breasts, however there are few published studies about detecting in dense breasts. To analyse its e ciency in relation to other segmentation techniques, we compare the performance with principal component analysis. To measure the quality of the segmentation obtained by the two methods, a area overlay measure will be used. To verify if there was any di erence between the results of the proposed methods in the detection of lesions in nondense breasts and in dense breasts, a statistic test for two proportions was used. Experimental results on the Mini-MIAS and DDSM database showed an accuracy of 92.71% in detecting masses in nondense and 79.17% in dense breasts. All experiments showed that the ICA lters have a better performance for detect lesions in dense breast, compared with PCA. Contrary to previous works, our experiments showed that there is actually a signi cant di erence between the detection of masses in dense and nondense breasts. This study can help specialist to detect lesions in dense breast. / O câncer de mama é o segundo tipo de câncer que mais afeta mulheres no mundo, perdendo apenas para o câncer de pele não melanoma. A densidade da mama pode di cultar a localização de massas, especialmente em estágios iniciais. Neste trabalho, propõe-se o uso de análise de componentes independentes para detectar e segmentar lesões em mamas densas. Vários trabalhos sugerem o uso do diagnóstico auxiliado por computador, aumentando a sensibilidade para acima de 90% na detecção de câncer em mamas não densas, no entanto, existem poucos estudos publicados sobre a detecção em mamas densas. Para analisar a e ciência do método proposto em relação a outras técnicas de segmentação, comparamos o desempenho com a análise de componentes principais. Para medir a qualidade da segmentação obtida pelos dois métodos, será utilizada uma medida de sobreposição de área. Para veri car se houve diferença entre os resultados dos métodos propostos na detecção de lesões em mamas não densas e nas mamas densas, foi utilizado um teste estatístico para duas proporções. Os resultados experimentais usando os bancos de dados Mini-MIAS e DDSM mostraram uma acurácia de 92,71% na detecção de massas em mamas não densas e 79,17% em mamas densas. Todas as experiências mostraram que os ltros de ICA usados têm um melhor desempenho para detectar lesões em mamas densas, em comparação com PCA. Contrariamente aos trabalhos anteriores, nossos experimentos mostraram que existe realmente uma diferença signi cativa entre a detecção de massas em mamas densas e não densas. Este estudo pode ajudar o especialista a detectar lesões em mamas densas

Medical image analysis aims at developing techniques to extract information from medical images. Among its many sub-fields, image registration and segmentation are two important topics. In this report, we present four pieces of work, addressing different problems as well as coupling them into a unified framework of shape based image segmentation. Specifically: 1. We link the image registration with the point set registration, and propose a globally optimal diffeomorphic registration technique for point set registration. 2. We propose an image segmentation technique which incorporates the robust statistics of the image and the multiple contour evolution. Therefore, the method is able to simultaneously extract multiple targets from the image. 3. By combining the image registration, statistical learning, and image segmentation, we perform a shape based method which not only utilizes the image information but also the shape knowledge. 4. A multi-scale shape representation based on the wavelet transformation is proposed. In particular, the shape is represented by wavelet coefficients in a hierarchical way in order to decompose the shape variance in multiple scales. Furthermore, the statistical shape learning and shape based segmentation is performed under such multi-scale shape representation framework.