The overwhelming majority of global trade is executed over maritime infrastructure, and port-side optimization problems are significant given that commercial maritime ports are hubs at which sea trade routes and land/rail trade routes converge. Therefore, optimizing maritime operations brings the promise of improvements with global impact. Major performance bottlenecks in maritime trade process include the handling of insurance claims on shipping containers and vessel service time at port. The former has high input dimensionality and includes data pertaining to environmental and human attributes, as well as operational attributes such as the weight balance of a shipping container; and therefore lends itself to multiple classification method- ologies, many of which are explored in this work. In order to compare their performance, a first-of-its-kind dataset was developed with carefully curated attributes. The performance of these methodologies was improved by exploring metalearning techniques to improve the collective performance of a subset of these classifiers. The latter problem formulated as a schedule optimization, solved with a fuzzy system to control port-side resource deployment; whose parameters are optimized by a multi-objective evolutionary algorithm which outperforms current industry practice (as mined from real-world data). This methodology has been applied to multiple ports across the globe to demonstrate its generalizability, and improves upon current industry practice even with synthetically increased vessel traffic.

Serine/threonine phosphatase 2A (PP2A) consists of an AC core dimer composed of catalytic (C), structural (A) subunits complexed to a variable regulatory subunit derived from three gene families (B, B', B"). My dissertation work characterized the structure and function of a neuron-specific splice variant of the Bbeta regulatory gene termed Bbeta2. I found that the divergent N-terminus of Bbeta2 does not affect phosphatase activity or holoenzyme association but encodes a mitochondrial targeting signal. Moreover, transient and stable expression of wild-type Bbeta2 but not Bbeta1, Bbeta2 mutants defective in mitochondrial targeting or a monomeric mutant unable to associate with the holoenzyme, promotes apoptosis in neurons while knock-down of endogenous Bbeta2 is neuroprotective. Furthermore, I identified the mechanisms by which Bbeta2 incorporates the PP2A holoenzyme. By performing charge reversal mutagenesis in Bgamma as a model for B family regulatory subunits, I found that holoenzyme association requires multiple electrostatic charges clustered in WD repeats 3 and 4 of the beta-propeller. To identify residues in Bbeta2 important for mitochondrial association, I performed mutagenesis of the divergent N-terminus of Bbeta2 and identified basic and hydrophobic residues that are critical for mitochondrial association. The variable N-terminal tail of Bbeta2 is a cryptic mitochondrial import sequence that promotes import of GFP, but not full-length Bbeta2, because its beta-propeller domain resists the partial unfolding step necessary for translocation. Lastly, I addressed the mechanism by which Bbeta2 promotes apoptosis in neurons. I found that overexpressing Bbeta2 fragments mitochondria while RNAi of the endogenous protein promotes mitochondrial fusion in neurons. Conversely, targeting PKA, a well characterized prosurvival kinase, to the OMM by overexpressing A kinase anchoring protein 121 (AKAP121) opposes the effects of the phosphatase by elongating mitochondria. Furthermore, downregulating the endogenous AKAP121 by RNAi, or inhibiting PKA at the OMM by overexpressing an inhibitor of PKA (OMM-PKI) fragments mitochondria. The effects of OMM-targeted PP2A or PKA on survival require remodeling of mitochondria, since blocking mitochondrial fission reversed the proapoptotic effects of Bbeta2 and OMM-PKI. My dissertation provides a novel mechanism by which kinase/phosphatase signaling determines neuronal survival.

Disc degeneration is a natural process and is widely prevalent. The severity of disc degeneration and the type of treatment varies from person to person. Fusion is a commonly chosen treatment option. However, clinical and biomechanical studies have shown that intervertebral discs adjacent to a fusion experience increased motion and higher stress which may lead to adjacent-segment disease. Cervical disc arthroplasty achieves similar decompression of the neural elements, but preserves the motion at the operated level and may potentially decrease the occurrence of adjacent segment degeneration. Computationally, a validated intact 3D finite element model of the cervical spine (C2-T1) was modified to simulate single (C5-C6) and bi-level (C5-C7) degeneration. The single level degenerative model was modified to simulate single level fusion and arthroplasty with the Bryan and Prestige LP artificial discs. The bi-level degenerative model was modified to simulate a bi-level fusion, bi-level arthroplasty with Bryan and Prestige LP discs and a disc replacement adjacent to fusion. An in-vitro biomechanical study was also conducted to address the effects of arthroplasty and fusion on the kinematics of the cervical spine. A total of 11 specimens (C2-T1) were divided into two groups (Bryan and Prestige LP). The specimens were tested in the following order; intact, single level TDR at C5-C6, bi-level TDR C5-C6-C7, fusion at C5-C6 and TDR at C6-C7 (Hybrid construct) and finally a bi-level fusion. The intact state was tested up to a moment of 2Nm. After surgical intervention, the specimens were loaded until the primary motion (C2-T1) matched the motion of intact state (hybrid control). In all cases; computational and experimental, an arthroplasty preserved motion at the implanted level and maintained normal motion at the nonoperative levels. A fusion, on the other hand, resulted in a significant decrease in motion at the fused level and an increase in motion at the un-fused levels. In the hybrid construct, the TDR adjacent to fusion preserved motion at that level, thus reducing the demand on the other levels. The computational models were used to analyze disc stresses at the adjacent levels and facet forces at the index and adjacent levels. The disc stresses followed the same trends as motion. Facet forces though, increased considerably at the index level following a TDR. There was a decrease in facet forces however at the adjacent levels. The adjacent level facet forces increased considerably with a fusion. The hybrid construct had adjacent level facet forces between the bi-level TDR and bi-level fusion models. To conclude, this study highlighted that cervical disc replacement with both the Bryan and Prestige LP discs not only preserved the motion at the operated level, but also maintained the normal motion at the adjacent levels. Under hybrid loading, the motion pattern of the spine with a TDR was closer to the intact motion pattern, as compared to the degenerative or fusion models. Also, in the presence of a pre-existing fusion, this study shows that an adjacent segment disc replacement is preferable to a second fusion.

With the advancement of technology, medical imaging has become a fast growing area of research. Some imaging questions require little physician analysis, such as diagnosing a broken bone, using a 2-D X-ray image. More complicated questions, using 3-D scans, such as computerized tomography (CT), can be much more difficult to answer. For example, estimating tumor growth to evaluate malignancy; which informs whether intervention is necessary. This requires careful delineation of different structures in the image. For example, what is the tumor versus what is normal tissue; this is referred to as segmentation. Currently, the gold standard of segmentation is for a radiologist to manually trace structure edges in the 3-D image, however, this can be extremely time consuming. Additionally, manual segmentation results can differ drastically between and even within radiologists. A more reproducible, less variable, and more time efficient segmentation approach would drastically improve medical treatment. This potential, as well as the continued increase in computing power, has led to computationally intensive semiautomated segmentation algorithms. Segmentation algorithms' widespread use is limited due to difficulty in validating their performance. Fusion models, such as STAPLE, have been proposed as a way to combine multiple segmentations into a consensus ground truth; this allows for evaluation of both manual and semiautomated segmentation in relation to the consensus ground truth. Once a consensus ground truth is obtained, a multitude of approaches have been proposed for evaluating different aspects of segmentation performance; segmentation accuracy, between- and within -reader variability. The focus of this dissertation is threefold. First, a simulation based tool is introduced to allow for the validation of fusion models. The simulation properties closely follow a real dataset, in order to ensure that they mimic reality. Second, a statistical hierarchical Bayesian fusion model is proposed, in order to estimate a consensus ground truth within a robust statistical framework. The model is validated using the simulation tool and compared to other fusion models, including STAPLE. Additionally, the model is applied to real datasets and the consensus ground truth estimates are compared across different fusion models. Third, a statistical hierarchical Bayesian performance model is proposed in order to estimate segmentation method specific accuracy, between- and within -reader variability. An extensive simulation study is performed to validate the model’s parameter estimation and coverage properties. Additionally, the model is fit to a real data source and performance estimates are summarized.

Multispectral imaging (MSI) provides a wealth of imagery data that, together with modern signal processing techniques, facilitates the enhancement of document images. In this thesis, four topic areas are reviewed and applied to ancient documents. They are image fusion, matched filters, bleed-through removal, and shadow removal. These four areas of focus provide useful tools for papyrologists studying the digital imagery of documents. The results presented form a strong case for the utility of MSI data over the use of a single image captured at any given wavelength of light.

Lumbar fusion surgery is a commonly used procedure to treat severe spinal pathology and associated chronic disabling low back and leg pain. Despite the common incidence of spinal fusion surgery, few studies have examined patient outcomes or predictive correlates of this procedure. The objectives of this study were to characterize Utah workers who received lumbar fusion surgery in terms of relevant presurgical and outcome variables and to identify presurgical correlates of patient outcomes. An archival prospective research design was utilized consisting of a retrospective medical chart review and a postsurgical telephone outcome survey. Subjects were 203 workers' compensation patients from the state of Utah who have undergone spinal fusion surgery and who were at least 2 years postsurgery at time of follow-up. Outcomes were assessed for 144 of the 203 patients (71%). Presurgical measures _included demographic, work, compensation, disability, health, surgical, and physiological variables. Outcome measures included solid arthrosis, patient satisfaction, work disability status, functional disability due to back pain, and multidimensional health. Analysis of patient outcome data revealed that solid arthrosis was achieved in 71.9% of patients. Forty-six percent of subjects felt their back/leg pain problems were worse than what they had expected following the surgery, and 42 % felt that their quality of life had not changed or worsened as a result of lumbar fusion. Twenty-eight percent of fusion patients were work disabled at follow-up. Fusion patient mean outcome scores on multidimensional health measures reflected poorer health than comparative medical patient and nonpatient norms. The most consistent presurgical correlates across outcomes were lawyer involvement, number of prior low back operations, age at injury, and household income at time of injury. Results are compared to data from previous lumbar fusion research studies and reasons for varying findings are offered. Implications of the findings are discussed in terms of inadequate patient selection and insufficient assessment of patient outcomes in low back research studies. Limitations of the present research are discussed, including how placebo, natural history, and regression to the mean can lead to erroneous conclusions about the efficacy of lumber fusion surgery. Suggestions for improvements in low back surgery outcome research are offered.

This thesis makes a number of new contributions to control and sensing for unmanned vehicles. I begin by developing a non-linear simulation of a small unmanned helicopter and then proceed to develop new algorithms for control and sensing using the simulation. The work is field-tested in successful flight trials of biologically inspired vision and neural network control for an unstable rotorcraft. The techniques are more robust and more easily implemented on a small flying vehicle than previously attempted methods.¶ Experiments from biology suggest that the sensing of image motion or optic flow in insects provides a means of determining the range to obstacles and terrain. This biologically inspired approach is applied to control of height in a helicopter, leading to the World’s first optic flow based terrain following controller for an unmanned helicopter in forward flight. Another novel optic flow based controller is developed for the control of velocity in hover. Using the measurements of height from other sensors, optic flow is used to provide a measure of the helicopters lateral and longitudinal velocities relative to the ground plane. Feedback of these velocity measurements enables automated hover with a drift of only a few cm per second, which is sufficient to allow a helicopter to land autonomously in gusty conditions with no absolute measurement of position.¶ New techniques for sensor fusion using Extended Kalman Filtering are developed to estimate attitude and velocity from noisy inertial sensors and optic flow measurements. However, such control and sensor fusion techniques can be computationally intensive, rendering them difficult or impossible to implement on a small unmanned vehicle due to limitations on computing resources. Since neural networks can perform these functions with minimal computing hardware, a new technique of control using neural networks is presented. First a hybrid plant model consisting of exactly known dynamics is combined with a black-box representation of the unknown dynamics. Simulated trajectories are then calculated for the plant using an optimal controller. Finally, a neural network is trained to mimic the optimal controller. Flight test results of control of the heave dynamics of a helicopter confirm the neural network controller’s ability to operate in high disturbance conditions and suggest that the neural network outperforms a PD controller. Sensor fusion and control of the lateral and longitudinal dynamics of the helicopter are also shown to be easily achieved using computationally modest neural networks.

Lorsque plusieurs classifieurs sont amenés à concourir à une même tâche de reconnaissance, plusieurs stratégies de décisions, impliquant ces classifieurs de différents manières, sont possibles. Une première stratégie consiste à décider suite à différents avis : il s'agit de la combinaison de classifieurs. Une deuxième stratégie consiste à utiliser un ou plusieurs avis pour mieux guider d'autres classifieurs dans leurs phases d'apprentissages, et à utiliser un ou plusieurs avis pour améliorer la prise de décisions d'autres classifieurs dans la phase de classement : il s'agit de la coopération de classifieurs. Enfin, la troisième et dernière stratégie consiste à privilégier un ou plusieurs classifieurs en fonction de divers critères ou en fonction de la situation : il s'agit de la sélection de classifieurs. L'aspect temporel de la RdF, c'est-à-dire l'évolution possible des classes à reconnaître, est traité par la stratégie de la sélection. En étudiant les aspects statiques et dynamiques de la RdF, nous montrons que pour reconnaître des classes dynamiques, deux approches sont possibles. Ces deux approches sont validées sur un ensemble de test. Dans le cas où les trajectoires des classes ne s'intersectent pas et que ces classes sont multimodales, l'approche proposée consiste à transformer ces classes dynamiques en classes statiques. En intégrant l'évolution de ces classes dans le temps, les classes obtenues deviennent alors complexes. Pour traiter ce type de classes, un algorithme de coopération des classifieurs est proposé. Il met en {\oe}uvre, d'une part, une méthode de classification non-supervisée effectuant une sélection adaptative de classifieurs et, d'autre part, plusieurs méthodes de RdF supervisées. Lorsqu'il n'y a pas d'intersection et que les classes évoluent de manière continue dans le temps, l'approche proposée consiste à rendre dynamique le système de RdF. Une méthode, fondée sur la modélisation des changements d'états du système par un réseau de Petri flou, est proposée. La méthode permet de prédire le ou les états du système les mieux adaptés au problème de RdF, à l'instant considéré.

La fusion membranaire est un processus biologique courant que l'on retrouve notamment lors de l'exocytose, du trafic intracellulaire ou de l'entrée des virus dans les cellules. Dans le cas des rhabdovirus, une famille où l'on retrouve notamment le virus de la rage (RV) et le virus de la stomatite vésiculaire (VSV),cette fonction est assurée à pH légèrement acide par la glycoprotéine G. Cette protéine existe sous au moins trois conformations distinctes : à pH neutre, elle est présente sous une conformation native (N). Après une brève incubation à pH acide, elle est présente sous une conformation activée (A), sous laquelle elle est capable d'interagir avec une membrane cible par l'intermédiaire d'un peptide hydrophobe. Enfin, après une incubation prolongée à pH acide, elle apparaît sous une conformation inactivée (I) et est alors incapable d'induire la fusion membranaire. Contrairement à toutes les autres familles virales étudiées à ce jour, il existe un équilibre dépendant du pH entre ces conformations. De nombreuses données dans divers systèmes suggèrent qu'une protéine unique ne serait pas suffisante pour catalyser les processus de fusion membranaire, mais qu'au contraire une machinerie constituée d'un nombre plus ou moins important de protéines fusogènes serait nécessaire. Au cours de ce travail, nous avons étudié certaines propriétés du complexe de fusion des rhabdovirus. Nous avons ainsi montré qu'il était de grande taille et qu il était probablement plus grand que ce qui avait été proposé pour d autres familles virales. De plus, il est apparu que le complexe de fusion du virus rabique n avait pas une unique architecture possible, mais qu il existait au contraire divers types de complexe. Ensuite, l'observation par microscopie électronique de particules virales en train de fusionner avec des liposomes nous a montré que le processus de fusion induit par VSV se produisait toujours par la base et qu il s accompagnait d une redisposition complète des glycoprotéines à la surface du virus suivant un réseau hélicoïdal. Enfin, nous sommes parvenu à isoler l ectodomaine de G et à en obtenir des cristaux diffractant à 3,5 Å, ce qui permet d envisager à terme une résolution de la structure de G. L étude de l interaction entre l ectodomaine de G et des liposomes nous a permis de reproduire les réseaux hélicoïdaux observés à la surface du virus et d étudier leurs propriétés. L ensemble de ces données nous a permis de proposer un nouveau modèle pour le processus de fusion chez les rhabdovirus, mais il pourrait également être pertinent pour d autres familles virales.

La biométrie se réfère à la reconnaissance automatique des individus basée sur leurs caractéristiques physiologiques et/ou comportementales. Les systèmes biométriques unimodaux permettent de reconnaître une personne en utilisant une seule modalité biométrique, mais ne peuvent pas garantir avec certitude une bonne identification. De plus, ces systèmes sont sensibles au bruit introduit par l'unique capteur, à la non-universalité et au manque d'individualité de la modalité biométrique choisie ainsi qu'aux tentatives d'intrusion. La plupart de ces problèmes peuvent être réduits par la mise en place de systèmes biométriques multimodaux utilisant plusieurs signatures biométriques d'une même personne. Dans cette thèse, nous abordons plusieurs points importants concernant la biométrie multimodale. Tout d'abord, après avoir dressé un état de l'art en fusion multimodale, nous faisons le lien entre le fonctionnement du cerveau et certains algorithmes fondamentaux utilisés en reconnaissance faciale. Ensuite, nous mettons en avant l'utilisation des ondelettes à divers niveaux du système biométrique multimodal. Enfin, l'exploration de nouvelles techniques de fusion de signatures biométriques issues du visage (modalité naturelle et non intrusive) et de l'iris (une des modalités les plus précises) ainsi que des analyses statistiques à grande échelle des scores de similarité provenant de chaque modalité ont permis de mettre au point une méthode originale de fusion adaptative combinant l'utilisation des ondelettes et des moments statistiques.