A CT-scan of lungs has become ubiquitous as a thoracic diagnostic tool. Thus, using CT-scan images in developing predictive models for tumor types and survival time of patients afflicted with Non-Small Cell Lung Cancer (NSCLC) would provide a novel approach to non-invasive tumor analysis. It can provide an alternative to histopathological techniques such as needle biopsy. Two major tumor analysis problems were addressed in course of this study, tumor type classification and survival time prediction. CT-scan images of 109 patients with NSCLC were used in this study. The first involved classifying tumor types into two major classes of non-small cell lung tumors, Adenocarcinoma and Squamous-cell Carcinoma, each constituting 30% of all lung tumors. In a first of its kind investigation, a large group of 2D and 3D image features, which were hypothesized to be useful, are evaluated for effectiveness in classifying the tumors. Classifiers including decision trees and support vector machines (SVM) were used along with feature selection techniques (wrappers and relief-F) to build models for tumor classification. Results show that over the large feature space for both 2D and 3D features it is possible to predict tumor classes with over 63% accuracy, showing new features may be of help. The accuracy achieved using 2D and 3D features is similar, with 3D easier to use. The tumor classification study was then extended by introducing the Bronchioalveolar Carcinoma (BAC) tumor type. Following up on the hypothesis that Bronchioalveolar Carcinoma is substantially different from other NSCLC tumor types, a two-class problem was created, where an attempt was made to differentiate BAC from the other two tumor types. To make a three-class problem a two-class problem, misclassification amongst Adenocarcinoma and Squamous-cell Carcinoma were ignored. Using the same prediction models as the previous study and just 3D image features, tumor classes were predicted with around 77% accuracy. The final study involved predicting two year survival time in patients suffering from NSCLC. Using a subset of the image features and a handful of clinical features, predictive models were developed to predict two year survival time in 95 NSCLC patients. A support vector machine classifier, naive Bayes classifier and decision tree classifier were used to develop the predictive models. Using the Area Under the Curve (AUC) as a performance metric, different models were developed and analyzed for their effectiveness in predicting survival time. A novel feature selection method to group features based on a correlation measure has been proposed in this work along with feature space reduction using principal component analysis. The parameters for the support vector machine were tuned using grid search. A model based on a combination of image and clinical features, achieved the best performance with an AUC of 0.69, using dimensionality reduction by means of principal component analysis along with grid search to tune the parameters of the SVM classifier. The study showed the effectiveness of a predominantly image feature space in predicting survival time. A comparison of the performance of the models from different classifiers also indicate SVMs consistently outperformed or matched the other two classifiers for this data.

Water-flooding in water-wet fractured reservoirs can recover significant amounts of oil through capillary driven imbibition. Unfortunately, many of the fractured reservoirs are mixed-wet/oil-wet and water-flooding leads to poor recovery as the capillary forces hinder imbibition. Surfactant injection and immiscible gas injection are two possible processes to improve recovery from fractured oil-wet reservoirs. In both these EOR methods, the gravity is the main driving force for oil recovery. Surfactant has been recommended and shown a great potential to improve oil recovery from oil-wet cores in the laboratory. To scale the results to field applications, the physics controlling the imbibition of surfactant solution and the scaling rules needs to be understood. The standard experiments for testing imbibition of surfactant solution involves an imbibition cell, where the core is placed in the surfactant solution and the recovery is measured versus time. Although these experiments prove the effectiveness of surfactants, little insight into the physics of the problem is achieved. This dissertation provides new core scale and pore scale information on imbibition of anionic surfactant solution into oil-wet porous media. In core scale, surfactant flooding into oil-wet fractured cores is performed and the imbibition of the surfactant solution into the core is monitored using X-ray computerized tomography(CT). The surfactant solution used is a mixture of several different surfactants and a co-solvent tailored to produce ultra-low interfacial tension (IFT) for the specific oil used in the study. From the CT images during surfactant flooding, the average penetration depth and the water saturation versus height and time is calculated. Cores of various sizes are used to better understand the effect of block dimension on imbibition behavior. The experimental results show that the brine injection into fractured oil-wet core only recovers oil present in the fracture; When the surfactant solution is injected, the CT images show the imbibition of surfactant solution into the matrix and increase in oil recovery. The surfactant solution imbibes as a front. The imbibition takes place both from the bottom and the sides of the core. The highest imbibition is observed close to the bottom of the core. The imbibition from the side decreases with height and lowest imbibition is observed close to the top of the core. Experiments with cores of different sizes show that increase in either the length or the diameter of the core causes decrease in the fractional recovery rate (%OOIP). Numerical simulation is also used to determine the physics that controls the imbibition profiles. %The numerical simulations show that the relative permeability curves strongly affect the imbibition profiles and should be well understood to accurately model the process. Both experimental and numerical simulation results imply that the gravity is the main driving force for the imbibition process. The traditional scaling group for gravity dominated imbibition only includes the length of the core to upscale the recovery for cores of different sizes. However based on the measurements and simulation results from this study, a new scaling group is proposed that includes both the diameter and the length of the core. It is shown that the new scaling group scales the recovery curves from this study better than the traditional scaling group. In field scale, the new scaling group predicts that the recovery from fractured oil-wet reservoirs by surfactant injection scales by both the vertical and horizontal fracture spacing. In addition to core scale experiments, capillary tube experiments are also performed. In these experiments, the displacement of oil by anionic surfactant solutions in oil-wet horizontal capillary tubes is studied. The position of the oil-aqueous phase interface is recorded with time. Several experimental parameters including the capillary tube radius and surfactant solution viscosity are varied to study their effect on the interface speed. Two different models are used to predict the oil-aqueous phase interface position with time. In the first model, it is assumed that the IFT is constant and ultra-low throughout the experiments. The second model involves change of wettability and IFT by adsorption of surfactant molecules to the oil-water interface and the solid surface. Comparing the predictions to the experimental results, it is observed that the second model provides a better match, especially for smaller capillary tubes. The model is then used to predict the imbibition rate for very small capillary tubes, which have equivalent permeability close to oil reservoirs. The results show that the oil displacement rate is limited by the rate of diffusion of surfactant molecules to the interface. In addition to surfactant flooding, immiscible gas injection can also improve recovery from fractured oil-wet reservoirs. In this process, the injected gas drains the oil in the matrix by gravity forces. Gravity drainage of oil with gas is an efficient recovery method in strongly water-wet reservoirs and yields very low residual oil saturations. However, many of the oil-producing fractured reservoirs are not strongly water-wet. Thus, predicting the profiles and ultimate recovery for mixed and oil-wet media is essential to design and optimization of improved recovery methods based on three-phase gravity drainage. In this dissertation, we provide the results from two- and three-phase gravity drainage experiments in sand-packed columns with varying wettability. The results show that the residual oil saturation from three-phase gravity drainage increases with increase in the fraction of oil-wet sand. A simple method is proposed for predicting the three-phase equilibrium saturation profiles as a function of wettability. In each case, the three-phase results were compared to the predictions from two-phase results of the same wettability. It is found that the gas/oil and oil/water transition levels can be predicted from pressure continuity arguments and the two-phase data. The predictions of three-phase saturations work well for the water-wet media, but become progressively worse with increasing oil-wet fraction. / text

The mechanical behavior of closed cell metallic foams strongly depends on their geometry at the scale of cells and cell walls. Two approaches are proposed in this work to address this computationally:(i) a controlled geometrical description of foam morphology features by exploiting an advanced tessellation-based procedure, allowing to generate realistic microstructural geometry,(ii) a procedure allowing to extract geometrical features of a foam morphology based on image-based modelling using CT scans. The first approach proposes a methodology that allows the automated generation of RVEs with a detailed control of the microstructure, including of cell geometries. It is primarily based on an inclusions packing algorithm assisted by distance fields control. Such distance fields can subsequently be used to morph inclusions, producing generalized tessellations with the possibility of incorporating curved and irregular boundaries. 3D morphologies of closed cell foams are produced by extracting the geometry from a proper combination of distance field functions. The procedure allows controlling the cell size distribution, spatial cell wall thickness distribution (correlated or not with the cell size distribution), wall curvatures and/or defects. An automated 3D meshing tool for implicit geometries was exploited to produce high quality tetrahedral meshes from the generated implicit foam geometries. Representative volume element based simulations were performed using this approach to assess the different morphological features relative importance on the mechanical behaviour of ALPORAS. An original extension of this tool was incorporating the transformation of 3D geometry into a shell-based finite element model. This resulted in a significant gain in computation time and allowed for simulating compression test up to densification (being out of reach with 3D solid finite element models) showing a good qualitative match with experimental results from the literature.The second approach proposes a robust methodology for the automated generation of shell-based finite element models directly from X-ray Computed Tomography (CT) scans.An in situ X-ray CT compression test of the sample was performed to serve as basis of comparison to the computations. As first steps, raw CT images are segmented using various image processing techniques and an implicit 3D geometry is reconstructed for each cell by using a Euclidean distance field computation technique. An automated geometrical procedure is used next to extract a (surface) shell geometry from this implicit 3D geometry, followed by subsequent meshing step. A direct comparison of the performed simulations with raw experimental data is performed. The detailed deformation and failure mechanisms of closed-cell foams under quasi static uniaxial compressive loading are investigated numerically and compared directly with the result of the in situ experimental measurement. / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished

Cochlear implants are electronic devices intended for restoring hearing to the profoundly deaf. Unfortunately the degree of restored hearing varies greatly between subjects. To investigate some of the mechanisms that determine this variability, mathematical models of the auditory system are used. The level of detail that these models incorporate varies greatly. The present study describes the development of a method to create high detail, subject specific cochlea models. μ-CT scans and photomicrographs were used to obtain the morphology and histology of a specific guinea pig cochlea. A 3D model was constructed from this data and the finite element method was used to determine the potential distribution inside the cochlea. The potential distribution was calculated for different stimulus protocols applied to different modelled electrodes. A neuron model was then used to obtain neural excitation profiles. The modelled excitation profiles were compared to data from literature and it was found that this model is valid and can be used as a tool in electric hearing research. The model output was also compared to brainstem response data from the specific subject to assess the degree to which this model can predict brain stem data from a specific subject. Possible improvements to the model were also discussed. / Dissertation (MEng)--University of Pretoria, 2009. / Electrical, Electronic and Computer Engineering / unrestricted

Historique et Objectif : La littérature est conflictuelle sur le rôle des calcifications aortiques dans la rupture d’anévrisme de l’aorte abdominale (AAA). La prédiction de rupture d’AAA basée sur le sexe et le diamètre est peu précise. Le but de ce projet était donc de déterminer si les calcifications permettent de mieux prédire la rupture d’AAA que le sexe et le diamètre à eux seuls. Méthodologie : Lors de cette étude rétrospective, 80 patients traités pour rupture d’AAA entre Janvier 2001 et Août 2018 ont été appariés à 80 patients non-rompus sur la base du diamètre maximal d’AAA, de l’âge, du sexe et de la présence de contraste lors du scan. La charge et la répartition des calcifications de la paroi aortique ainsi que certaines variables morphologiques d’anévrisme ont été comparées entre les deux groupes par analyse univariée et apprentissage machine. Résultats : L’âge moyen des patients était de 74.0 ± 8.4 ans et 89% étaient des hommes. Les diamètres d’AAA étaient équivalents entre groupes (80.9 ± 17.5 vs 79.0 ± 17.3 mm, p= 0.505). Selon l’analyse univariée, les anévrismes rompus comportaient significativement moins d’agrégats de calcifications (18.0 ± 17.9 vs 25.6 ± 18.9, p=0.010) et étaient moins enclins à avoir un collet (45.0% vs 76.3%, p<0.0001). Les 5 variables les plus importantes délivrées par l’apprentissage machine étaient: collet, antiplaquettaires, nombre de calcifications, distance d’Euler entre calcifications et finalement l’écart-type de la distance d’Euler entre calcifications. Le modèle à 5 variables a produit une aire sous la courbe (AUC) de 0.81 ± 0.02 (sensibilité 83% et spécificité 71%), supérieure à une AUC de 0.67(IC 95%, 0.58-0.77%) (sensibilité 60% et spécificité 77%) obtenues dans une étude antérieure avec une population similaire à celle-ci et ne tenant compte que du sexe et du diamètre. Conclusion : La charge en calcifications des anévrismes rompus était moins bien répartie que celle des non-rompus. Le modèle d’apprentissage machine a mieux prédit la rupture que le modèle basé uniquement sur le diamètre et le sexe. / Background and Purpose: Literature is conflictual regarding the role of aortic calcification in AAA rupture. AAA rupture prediction based on sex and diameter could be improved. The goal of this project was to assess whether aortic calcification could better predict AAA rupture. Methods: In this retrospective study, 80 patients treated for a ruptured AAA between January 2001 and August 2018 were matched with 80 non-ruptured patients based on maximal AAA diameter, age, sex and contrast enhancement status of the CT scan. Calcification load and dispersion, morphologic and clinical variables were compared between both groups using a univariable analysis and machine learning. Results: Mean age of patients was 74.0 ± 8.4 years and 89% were men. AAA diameters were equivalent in both groups (80.9 ± 17.5 vs 79.0 ± 17.3 mm, p= 0.505). Ruptured aneurysms contained a smaller number of calcification chunks than the non-ruptured (18.0 ± 17.9 vs 25.6 ± 18.9, p=0.010) and were less likely to have a proximal neck than the non-ruptured (45.0% vs 76.3%, p<0.0001). In the machine learning analysis, 5 variables were associated to AAA rupture: proximal neck, antiplatelets, calcification number, Euler distance between calcifications and standard deviation of the Euler distance between calcifications. The model including these 5 variables yielded an area under the curve (AUC) of 0.81 ± 0.02 (83% sensitivity and 71% specificity) which was better than a previous study with a similar population reporting a 0.67 AUC (95% CI, 0.58-0.77%) (60% sensitivity and 77% specificity) for sex and diameter only. Conclusion: Ruptured aneurysms were more likely to have their calcification load concentrated in a small number of clusters closer to each other. Our 5-variable model predicted rupture better than the model based on age and sex.

No description available.

L’objet de cette thèse est de développer des méthodologies permettant d’identifier la distribution spatiale des valeurs de perméabilité dans des échantillons de roches. Nous avons tout d’abord développé en laboratoire des expériences d’injection de fluide miscible très visqueux dans des échantillons initialement saturés par une saumure peu visqueuse. Pendant l’injection, l’évolution au cours du temps de la pression différentielle entre les deux faces de l’échantillon a été enregistrée par des capteurs de pression. En outre, des mesures scanner ont fourni une carte 3D de la porosité ainsi que des cartes 3D décrivant la distribution spatiale des concentrations dans l’échantillon à différents temps. Nous avons mis en place une méthode d’interprétation donnant directement le profil 1D de la perméabilité le long de la direction d’écoulement à partir de la pression différentielle mesurée au cours du temps. Cette méthode a été validée numériquement et expérimentalement. Puis, afin d’affiner la description de l’agencement des valeurs de perméabilité dans l’échantillon, c’est à dire d’obtenir un modèle 3D de perméabilité représentatif de l’échantillon, nous avons développé une méthodologie itérative de calage des pressions et des concentrations. Cette méthode passe par deux étapes : une optimisation simple pour capturer l’hétérogénéité dans la direction de l’écoulement et une optimisation complexe pour capturer l’hétérogénéité transverse. Cette méthode a été validée à partir de tests numériques. La méthode a été appliquée à deux des expériences d’injection de fluide visqueux. Nous avons pu alors déterminer des modèles de perméabilité capables de reproduire assez bien les données de pression et de concentration acquises pendant l’injection / The objective of this study is to develop new methodologies to identify the spatial distribution of permeability values inside the heterogeneous core samples. We developed laboratory viscous miscible displacements by injecting high viscosity glycerin into the core samples initially saturated by low viscosity brine. The pressure drop across the samples was measured as a function of time until breakthrough. Meanwhile, CT scan measurements provided a 3D porosity map plus several 3D maps of concentration distribution inside the core samples at different times. A simple permeability mapping technique was developed deducing a one-dimensional permeability profile along the flow direction from the measured pressure drop data. The method was validated with both numerical and laboratory experiments. To go beyond one-dimensional characterization of permeability into cores, we developed an iterative process for matching pressure and concentration data. This method consisted of two steps: a simple optimization for capturing the permeability heterogeneity along the flow direction axis and a complex optimization for capturing transversal permeability heterogeneities. The methodology was validated by numerical data. It was also applied to the data collected from two laboratory viscous miscible displacements. We showed that the final 3D permeability models reproduce well the measured pressure drop and concentration data

Depuis le début des années soixante-dix, le nombre de scanners par hôpitaux n’a fait qu’augmenter et leur utilisation est de plus en plus fréquente. Même si cette technique permet de donner des informations cliniques précieuses, elle a un revers qui est l’exposition du patient à des rayonnements ionisants. La sensibilité des enfants aux rayonnements est plus grande que celle des adultes, les enfants ont une espérance de vie importante et ont donc plus de risques de développer des cancers dans le futur. Il y a donc nécessité de tenter de réduire la dose au patient. Cette thèse vise donc à développer des stratégies d’optimisation sur les protocoles cliniques en utilisant des méthodes de simulation et de modélisation permettant de comprendre l’influence des paramètres des protocoles sur les indicateurs de qualité d’image et sur la dose délivrée au patient. Ce travail se divise en quatre parties: La première partie porte sur la modélisation de l’influence des paramètres des protocoles scanographiques sur deux indicateurs de qualité d’image et un indicateur de dose en utilisant la méthodologie des plans d’expériences. La seconde partie traite du développement d’un Protocole Générique Optimisé (PGO) pour la région de l’abdomen. A partir des données des modèles développés, un PGO recalculé pour cinq morphologies de patients pédiatriques et pour quatre modèles de scanners a été réalisé. L’ensemble des résultats, ont permis le développement d’un outil d’aide à l’optimisation permettant à l’utilisateur de générer un protocole optimisé en fonction du modèle de scanner et de la morphologie du patient. / For the last 10-years, computed tomography (CT) procedures and their increased use have been a major source for concern in the scientific community. This concern has been the starting point for several studies aiming to optimize the dose while maintaining a diagnostic image quality. In addition, it is important to pay special attention to dose levels for children (age range considered to be from a newborn baby to a 16-y-old patient). Indeed, children are more sensitive to ionizing radiations, and they have a longer life expectancy. Optimizing the CT protocols is a very difficult process due to the complexity of the acquisition parameters, starting with the individual patient characteristics, taking into account the available CT device and the required diagnostic image quality. This PhD project is contributing to the advancement of knowledge by: (1) Developing a new approach that can minimize the number of testing CT scans examinations while developing a predictive mathematical model allowing radiologists to prospectively anticipate how changes in protocols will affect the image quality and the delivered dose for four models of CT scan. (2) Setting-up a Generic Optimized Protocol (based on the size of the phantom CATPAHN 600) for four models of CT scan. (3) Developing a methodology to adapt the GOP to five sizes of pediatric patient using Size Specific Dose Estimate calculation (SSDE). (4) Evaluating subjective and objective image quality between size-based optimised CT protocol and age-based CT protocols. (5) Developing a CT protocol optimization tool and a tutorial helping the radiologists in the process of optimization.

Les imageurs embarqués sur les appareils de traitement par radiothérapie sont des dispositifs très efficaces pour améliorer la précision géométrique des irradiations. Cependant, ils ne sont pertinents que s'ils font l'objet de contrôles de performances précis et réguliers. Notre propos a donc été de développer des solutions logicielles permettant d'automatiser l'analyse des images de contrôle de qualité des principaux modes utilisés en imagerie de repositionnement : le mode 2D-MV qui mesure l'image radiante bidimensionnelle haute énergie transmise par le patient pendant le traitement, les modes 2D-kV (images bidimensionnelles basse énergie) et mode 3D (images tridimensionnelles MV ou kV) qui permettent de réajuster la position du patient avant de traiter. Nous avons également automatisé l'analyse des images du test de Winston & Lutz utilisé pour évaluer la précision mécanique des appareils de traitement sur lesquels s'exercent des contraintes supplémentaires dues au poids supplémentaire de l'imageur embarqué. Enfin nous avons développé et mis en oeuvre une méthodologie originale et performante utilisant des objetstest numériques pour évaluer les performances des solutions logicielles mises au point. Au bilan, l'automatisation des contrôles de qualité des imageurs embarqués avec les solutions logicielles développées dans le cadre de ces travaux de thèse permet de réduire d'un facteur de 100 le temps consacré par l'équipe de physique médicale à l'analyse des résultats des contrôles tout en améliorant leur précision grâce à l'utilisation d'analyses objectives et reproductibles et leur traçabilité grâce à l'édition automatique de rapports de contrôle et d'études statistiques.

The understanding of the mesorectal spread in rectal cancer has lead to wide acceptance of total mesorectal excision (TME) as the surgical technique of choice for carcinoma in the lower and mid rectum. While oncological results and survival have improved with TME-surgery, morbidity and mortality remain important issues. The most feared complication is symptomatic anastomotic leakage. The aim of this thesis was to focus on the role of the defunctioning stoma, risk factors, and postoperative monitoring in regard to anastomotic leakage in sphincter saving resection of the rectum. Intraoperative adverse events were analysed in a retrospective population based case-control study in which all patients who underwent elective anterior resection in Sweden between 1987 and 1995, and who died within 30 days or during the initial hospital stay (n=140), were compared with patients chosen at random (n=423) who underwent the same operation during the same period, but survived the operation. Intraoperative adverse events were more frequent in those who died, and reconstruction of an anastomosis judged unsatisfactory by the surgeon improved the outcome. In a population based retrospective case-control study, risk factors for symptomatic anastomotic leakage were investigated in randomly chosen sample of patients who underwent anterior resection in Sweden between 1987 and 1995 (n=432). Twelve per cent of the patients developed symptomatic leakage, and 25% of the patients with leakage ended up with a permanent stoma. In multivariate regression analysis, low anastomosis, preoperative radiotherapy, male gender and intraoperative adverse events were independent riskfactors for anastomotic leakage. In a randomised multicentre trial patients operated with sphincter saving TME¨surgery for rectal cancer were randomised to a defunctioning stoma (n=116) or not (n=118). The overall rate symptomatic leakage was 19%. Patienst without a defunctioning stoma leaked in 28% and patients with a defunctioing stoma in 10%, a statistically significant difference (p&lt;0.001) not previously demonstrated in any randomised trial of adequate size. Postoperative monitoring with computed tomography scan (CT-scan) on postoperative day 2 and 7, and C-reactive protein (CRP) daily in 33 patients operated on with anterior resection of the rectum, demonstrated larger pelvic fluid collections in patients with leakage before the leakage was clinically diagnosed. CRP was increased from postoperative day 2 and onwards in patients in whom clinical leakage was diagnosed on median postoperative day 8. In 23 patients who underwent anterior resection of the rectum, intraperitoneal metabolism was investigated using microdialysis technique measuring the carbohydrate metabolites lactate, pyruvate and glucose. Intraperitoneal cytokines IL-6, IL-10 and TNF-α were collected through a pelvic drain and analysed. In patients who developed leakage, the latate/pyruvate ratio was increased near the anastomosis on postoperative day 5 and 6, as well as IL-6 and IL-10 which were increased postoperatively day 1 and 2, while TNF-α was higher on day 1.