In the field of medical imaging, image registration methods are useful for many applications such as inter- and intra-subject morphological comparisons, creation of population atlases, delivery of precision therapies, etc. A user may want to know which is the most suitable registration algorithm that would work best for the intended application, but the vastness of medical image registration applications makes evaluation and comparison of image registration performance a non-trivial task. In general, evaluating image registration performance is not straightforward because in most image registration applications there is an absence of “Gold Standard” or ground truth correspondence map to compare against. It is therefore the primary goal of this thesis work to provide a means for recommending the most appropriate registration algorithm for a given task. One of the contributions of this thesis is to examine image registration algorithm performance at the component level. Another contribution of this thesis is to catalog the benefits and limitations of many of the most commonly used image registration evaluation approaches. One incremental contribution of this thesis was to demonstrate how existing evaluation methods can be applied in the midpoint coordinate system to evaluate some symmetric image registration algorithms such as the SyN registration algorithm. Finally, a major contribution of this thesis was to develop tools to evaluate and visualize 2D and 3D image registration shape collapse. This thesis demonstrates that many current diffeomorphic image registration algorithms suffer from the collapse problem, provides the first visualizations of the collapse problem in 3D for simple shapes and real human brain MR images, and provides the first experiments that demonstrate how adjusting image registration parameters can mitigate the collapse problem to some extent.

There is notable heterogeneity in clinical presentation of patients with chronic obstructive pulmonary disease (COPD). Classification of COPD is usually based on the severity of airflow limitation (pre- and post- bronchodilator FEV1), which may not sensitively differentiate subpopulations with distinct phenotypes. A recent advance of quantitative medical imaging and data analysis techniques allows for deriving quantitative computed tomography (QCT) imaging-based metrics. These imaging-based metrics can be used to link structural and functional alterations at multiscale levels of human lung. We acquired QCT images of 800 former and current smokers from Subpopulations and Intermediate Outcomes in COPD Study (SPIROMICS). A GPU-based symmetric non-rigid image registration method was applied at expiration and inspiration to derived QCT-based imaging metrics at multiscale levels. With these imaging-based variables, we employed a machine learning method (an unsupervised clustering technique (K-means)) to identify imaging-based clusters. Four clusters were identified for both current and former smokers. Four clusters were identified for both current and former smokers with meaningful associations with clinical and biomarker measures. Results demonstrated that QCT imaging-based variables in patients with COPD can derive statistically stable and clinically meaningful clusters. This sub-grouping can help better categorize the disease phenotypes, ultimately leading to a development of an efficient therapy.

The goal of lung-protective mechanical ventilation is to provide life-sustaining support of gas exchange while minimizing the risk of ventilator-induced lung injury. Multi-frequency oscillatory ventilation (MFOV) was proposed as an alternative lung-protective modality, in which multiple frequencies of pressure and flow oscillations are delivered simultaneously at the airway opening and allowed to distribute throughout the lung in accordance with regional mechanical properties. The distribution of oscillatory flow is frequency-dependent, such that regions overventilated at one frequency may be underventilated at another. Thus the central thesis of this work was that ventilation heterogeneity is frequency-dependent, and therefore ventilation with multiple simultaneous frequencies can be optimized to reduce the risk of ventilator-induced lung injury. Simulations in computational models of distributed oscillatory flow and gas transport demonstrated the sensitivity of regional ventilation heterogeneity to subject size, ventilation frequency, and injury severity. Although the risk of injury in the model associated with strain or strain rate individually was minimized by single-frequency ventilation, the risk of injury associated with mechanical power in lung parenchymal tissue was minimized by MFOV. In an experimental model of acute lung injury, MFOV was associated with reductions in the magnitude and spatial gradient of regional lung strain estimated by four-dimensional CT image registration, as well as increased rates of regional gas transport estimated by wash-in of xenon tracer gas. In conclusion, computational models demonstrated the potential for optimization of MFOV waveforms, and experimental trials demonstrated evidence of improved regional ventilation during MFOV.

The use of digitized information is rapidly gaining acceptance in bio-medical applications. Video compression plays an important role in the archiving and transmission of different digital diagnostic modalities. The present scheme of video compression for low bit-rate networks is not suitable for medical video sequences. The instability is the result of block artifacts resulting from the block based DCT coefficient quantization. The possibility of applying deformable motion estimation techniques to make the video compression standard (H.263) more adaptable for bio-medial applications was studied in detail. The study on the network characteristics and the behavior of various congestion control mechanisms was used to analyze the complete characteristics of existing low bit rate video compression algorithms. The study was conducted in three phases. The first phase involved the implementation and study of the present H.263 compression standard and its limitations. The second phase dealt with the analysis of an external force for active contours which was used to obtain estimates for deformable objects. The external force, which is termed Gradient Vector Flow (GVF), was computed as a diffusion of the gradient vectors associated with a gray-level or binary edge map derived from the image. The mathematical aspect of a multi-scale framework based on a medial representation for the segmentation and shape characterization of anatomical objects in medical imagery was derived in detail. The medial representations were based on a hierarchical representation of linked figural models such as protrusions, indentations, neighboring figures and included figures--which represented solid regions and their boundaries. The third phase dealt with the vital parameters for effective video streaming over the internet in the bottleneck bandwidth, which gives the upper limit for the speed of data delivery from one end point to the other in a network. If a codec attempts to send data beyond this limit, all packets above the limit will be lost. On the other hand, sending under this limit will clearly result in suboptimal video quality. During this phase the packet-drop-rate (PDR) performance of TCP(1/2) was investigated in conjunction with a few representative TCP-friendly congestion control protocols (CCP). The CCPs were TCP(1/256), SQRT(1/256) and TFRC (256), with and without self clocking. The CCPs were studied when subjected to an abrupt reduction in the available bandwidth. Additionally, the investigation studied the effect on the drop rates of TCP-Compatible algorithms by changing the queuing scheme from Random Early Detection (RED) to DropTail.

Dans cette thèse, nous nous intéressons à l'utilisation des modèles déformables surfaciques pour la segmentation d'images 3D et 4D. Dans un premier temps, nous nous sommes attachés à contraindre l'espace des déformations admissibles du modèle afin de rendre le processus de déformation plus fiable. Nous avons utilisé le formalisme des maillages simplexes pour exprimer des contraintes régularisantes de la surface. Nous avons développé un processus évolutif de déformation combinant une transformation globale ayant peu de degrés de liberté et un champ de déformations locales. Il permet de contrôler le nombre de degrés de liberté offerts au modèle surfacique de manière simple et efficace. Nous avons également cherché à enrichir la connaissance a priori des données apportée par le modèle. Nous utilisons des contraintes de forme pour faciliter la segmentation des structures à reconstruire, notamment dans les zones où les données de l'image sont bruitées ou lacunaires. Nous nous sommes également intéressés à la convergence formelle du processus de déformation. Nous avons développé un algorithme de changement de topologie des modèles discrets que nous avons comparé à l'approche par ensembles de niveaux. Dans un deuxième temps, nous nous sommes intéressés à la définition du terme d'attache aux données pour différents types d'images 3D. Nous avons envisagé plusieurs géométries rencontrées dans les images médicales. Nous avons étudié l'apport d'une information sur les régions ou sur la distribution des niveaux de gris de l'image pour la déformation ou le recalage multimodal d'un modèle. Finalement, nous nous sommes intéressés à la segmentation de séquences temporelles d'images cardiaques 2D ou 3D. La prise en compte de l'information temporelle permet d'introduire de nouvelles contraintes de déformations. Nous avons mis nos méthodes en pratique avec la segmentation d'images ou des séquences d'images cardiaques provenant de différentes modalités d'acquisition.

Les produits de contraste ultrasonore constituent un véritable apport pour l'imagerie échographique et sont aujourd'hui utilisés en clinique pour l'évaluation de la perfusion cardiaque ou encore la détection de tumeurs. Depuis quelques années, les transducteurs capacitifs micro-usinés (cMUTs) se présentent comme une alternative intéressante aux transducteurs piézoélectriques classiques et offrent certains avantages comme une large bande passante. Nous proposons dans cette thèse d'évaluer le potentiel de cette technologie pour l'imagerie de contraste. Dans un premier temps, notre étude s'est orientée vers l'adaptation des cMUTs à l'imagerie non linéaire. Ensuite, de nouvelles méthodes de détection de contraste, basées sur le comportement spécifique des microbulles, ont été développées pour exploiter les avantages de la technologie cMUT. Comparés aux méthodes conventionnelles, les résultats obtenus montrent une meilleure visualisation des agents de contraste par rapport aux tissus environnants. L'utilisation de cMUTs améliore l'efficacité de ces méthodes démontrant, ainsi, leur intérêt pour l'imagerie de contraste.

Magnetic resonance imaging has developed into a powerful diagnostic technique characterized by a very high spatial resolution and an inherently relatively low sensitivity. In order to improve the contrast of MRI images, contrast agents are commonly injected into the patients before an examination. These substances are paramagnetic, superparamagnetic, or ferromagnetic compounds that shorten the relaxations times of the water hydrogen atoms. At present, most of the contrast enhanced clinical exams are performed with gadolinium complexes. They are particularly useful as their ability to change the relaxation rate (or relaxivity) can be very high. Several factors have a strong influence on the relaxivity of MRI contrast agents but the water exchange time τm and the rotational correlation time τr are particularly important for obtaining an increased relaxivity. These parameters can be adjusted by suitable chemical modifications of the Gd(III) complexes. For instance, decreasing the tumbling rate by linking a Gd(III) complex to a macromolecule leads to an increased relaxivity. This goal can be achieved through a covalent or a noncovalent linkage with synthetic polymers, particles or biomacromolecules. However, the covalent bonding has a detrimental effect on the clearance of the metal complexes thus exposing the patient to the toxicity of released Gd(III) ions and metabolites. This problem could be circumvented by using covalent links that are cleaved by endogenous biomolecules or after administration of exogenous compounds following the exam. In this context, our approach was to bind Gd(III) chelates to macromolecules through disulfide links as the latter are known to be reduced in vivo by thiols present in the body. Towards this aim, we have developed two bifunctional chelator agents bearing a methanethiosulfonate group (MTS) which reacts specifically with thiols, thus spontaneously establishing a disulfide bond between the Gd(III) chelate and the thiolated macromolecule. The first ligand that we have prepared (MTS-ADO3A) is a monoamide derivative of DOTA with an ethyl-MTS substituent. This compound is relatively easily synthesized but amide arms such as the one it features are known to have a detrimental effect on relaxivity through the lengthening of water exchange times. The conjugate obtained by binding Gd(III) chelates of this ligand to albumin or to polythiolated silica nanoparticles has been studied by nuclear magnetic relaxation dispersion (NMRD),17O NMR and luminescence analyses. These measurements confirm that the method is suitable to increase the relaxivity (20 mM-1s 1, 20 MHz, 25°C) but that this relaxivity increase (of 300%) is limited by a slow water exchange (660 ns). To overcome this limitation, a second ligand called MTS-CyDOTA has been synthesized. This ligand is a DOTA ligand grafted with a cyclohexyl ring featuring a MTS function. The synthesis is more demanding but faster water exchange times are expected because of a more sterically crowded coordination sphere. Moreover, this second ligand has a more rigid structure that could limit the independent rotation of the chelate from the macromolecule. As expected, the water exchange time of the Gd(III) chelate of this ligand (120 ns) is clearly lower than the one determined for Gd MTS-ADO3A. After binding to albumin or to silica nanoparticles a notable relaxivity increase was expected. Unfortunately, if the obtained relaxivity is higher (30 mM-1s 1, 20 MHz, 25°C), its not as high as it could have been expected in view of the size of the conjugate and of the water exchange time of the free chelate. Results obtained in this work suggest that fixation on silica nanoparticles or on albumin drastically decreases the water exchange rate which remains the limiting parameter. This effect has already been reported for Gd(III) chelates linked to albumin by non-covalent bonds and has been assigned to stable layers of water molecules on the macromolecule surface. Thanks to the high loading of the silica nanoparticles (10000 Gd(III) per particle), we have reached very high molecular relaxivities (>200000 mM-1s-1). Stability tests carried out on the disulfide links formed suggest that the small amount of free thiols in the circulation is not sufficient to cause a significant degradation of the disulfide bond in the conjugate within a reasonable length of time. An injection of glutathione would be necessary to achieve a complete degradation. To avoid the problem of water exchange lenghtening, we propose to increase the distance between Gd(III) chelates and macromolecules without loss of rigidity by developing double anchor chelates with substituents grafted on the side of the ring. Considerable synthetic efforts have devoted to the synthesis of such a system and are discussed in chapter VI. At present, this work is still in progress in the laboratory and recent results suggest that it should be possible to evaluate this double arms system in a near future. On the fringe of this synthesis, we present a relaxometric study on the interaction between HSA and a hydrophobic Gd(III) chelate obtained during the preparation of our double anchor chelate. Finally, a chapter of this work is devoted to the study of two compounds, phenEDTA and phenDTPA, which are ditopic chelates featuring a dihydro-1,10-phenanthroline unit that spontaneously self-assemble in the presence of a transition metal ion. The tris-complex generated by this process rotates more slowly in solution and thus presents an increased relaxivity (+130%). As part of this work, we have determined by potentiometric titration the acidity constants of phenEDTA and its stability constant with Gd(III). Moreover, the protonation scheme of this ligand has been studied by NMR titration. The particular behavior of Gd phenDTPA and Fe(Gd phenDTPA)3 in the presence of Zn(II) has also been studied by relaxometry, luminescence and EXAFS.

Pediatric female and young adult female patients who undergo computed tomography (CT) scanning may be at higher risk for developing radiation-induced breast cancer later in life. Thus, the purpose of this thesis was to both accurately quantify dose and explore new strategies for CT breast dose reduction. In order to determine dose reduction, dose quantification was first assessed through the development and validation of an empirical model for describing attenuation in CT and second through evaluation of the precision of dosimetry-related measurements obtained using three different models of CT scanners. Breast dose-savings was evaluated using CT dose index phantoms, anthropomorphic phantoms, and Monte Carlo computer modeling. Modifications to current scanning procedures, such as proper patient centering and beginning data acquisition with the x-ray tube facing a patient's posterior, were shown to minimize breast dose. Novel techniques, including varying the x-ray tube voltage during scanning and incorporation of a dynamic x-ray beam filter over the breasts, were also found to successfully reduce breast dose.

La squelettisation est une méthode de reconnaissance et de caractérisation des formes utilisant des notions de topologie discrète, qui permet d'extraire le squelette d'une forme binaire. Ce squelette (un sous ensemble de la forme centré, fin, topologiquement équivalent à la forme et reconstructible) se calcule grâce à plusieurs balayages de l'image. Dans ce travail de thèse nous nous sommes intéressé au calcul du squelette en une seule passe (mode " streaming "). Pour cela nous avons proposé un nouvel algorithme qui ré-ordonne les calculs sur chaque pixels dans leur ordre d'apparition dans l'image (sans changer le sens de propagation de l'information dans l'image) en calculant une carte de distance asymétrique puis qui trouve les points de l'axe médian. Selon un ordre particulier, qui garantit l'identité de nos résultats avec ceux d'un algorithme de référence, nous pouvons supprimer les points de la carte de distance en ne conservant que les points du squelette dont les points de l'axe médian. En plus du fait que nos squelettes soient identiques au pixel près à ceux obtenus par l'algorithme de référence, notre algorithme les calcule plus rapidement. Bien que notre approche soit générique, nous avons utilisé cet algorithme pour développer un outil logiciel en imagerie médicale permettant d'extraire les caractéristiques osseuses dans le but de mesurer la structure osseuse à partir du squelette obtenu sur des images de micro-scanner. L'extraction de zones d'os trabéculaire 2D sur des images binarisées, permet bien une caractérisation de l'os en accord avec la vérité terrain obtenue par d'autres logiciels.

The goal of this thesis was to produce an animal model that develops atherosclerotic plaque featuring plaque neovascularization leading to intraplaque hemorrhage and is suitable for noninvasive imaging studies. Several strategies were tested for their effectiveness in producing such plaques in the rabbit aorta, including: a high cholesterol diet, vascular endothelial growth factor injections, therapeutic contrast ultrasound, and balloon catheter injury. It was found that a combination of the high cholesterol diet and balloon injury was able to achieve plaque neovascularization in a manner dependent on circulating plasma cholesterol levels. In addition, a contrast-enhanced magnetic resonance imaging technique implemented in the animal model was able to detect plaque neovascularization and monitor its change over time in a single group of animals. In conclusion, an animal model was created where plaque neovascularization occurs in a predictable fashion and can be studied with non-invasive magnetic resonance imaging.