Diffusion weighted imaging (DWI) is a branch within the field of magnetic resonance imaging (MRI) that relies on the diffusion of water molecules for its contrast. Its clinical applications include the early diagnosis of ischemic stroke and mapping of the nerve tracts of the brain. The recent development of filter exchange imaging (FEXI) and the introduction of the apparent exchange rate (AXR) present a new DWI based technique that uses the exchange of water between compartments as contrast. FEXI could offer new clinical possibilities in diagnosis, differentiation and treatment follow-up of conditions involving edema or altered membrane permeability, such as tumors, cerebral edema, multiple sclerosis and stroke. Necessary steps in determining the potential of AXR as a new biomarker include running comparative studies between controls and different patient groups, looking for conditions showing large AXR-changes. However, before designing such studies, the experimental protocol of FEXI should be optimized to minimize the experimental variance. Such optimization would improve the data quality, shorten the scan time and keep the required study group sizes smaller.  Here, optimization was done using an active imaging approach and the Cramer-Rao lower bound (CRLB) of Fisher information theory. Three optimal protocols were obtained, each specialized at different tissue types, and the CRLB method was verified by bootstrapping. A test-retest study of 18 volunteers was conducted in order to investigate the reproducibility of the AXR as measured by one of the protocols, adapted for the scanner. Group sizes required were calculated based on both CRLB and the variability of the test-retest data, as well as choices in data analysis such as region of interest (ROI) size. The result of this study is new protocols offering a reduction in coefficient of variation (CV) of around 30%, as compared to previously presented protocols. Calculations of group sizes required showed that they can be used to decide whether any patient group, in a given brain region, has large alterations of AXR using as few as four individuals per group, on average, while still keeping the scan time below 15 minutes. The test-retest study showed a larger than expected variability however, and uncovered artifact like changes in AXR between measurements. Reproducibility of AXR values ranged from modest to acceptable, depending on the brain region. Group size estimations based on the collected data showed that it is still possible to detect AXR difference larger than 50% in most brain regions using fewer than ten individuals. Limitations of this study include an imprecise knowledge of model priors and a possibly suboptimal modeling of the bias caused by weak signals. Future studies on FEXI methodology could improve the method further by addressing these matters and possibly also the unknown source of variability. For minimal variability, comparative studies of AXR in patient groups could use a protocol among those presented here, while choosing large ROI sizes and calculating the AXR based on averaged signals.

We conduct a deep mid-infrared census of nine massive galaxy clusters at (0 < z < 1.3) with a total of ~ 1500 spectroscopically confirmed member galaxies using Spitzer /IRAC photometry and established mid-infrared color selection techniques. Of the 949 cluster galaxies that are detected in at least three of the four IRAC channels at the >= 3 sigma level, we identify 12 that host mid-infrared selected active galactic nuclei (IR-AGN). To compare the IR-AGN across our redshift range, we define two complete samples of cluster galaxies: (1) optically-selected members with rest-frame VAB magnitude < -21.5 and (2) mid-IR selected members brighter than (M*_3.6 +0.5), i.e. essentially a stellar mass cut. In both samples, we measure f_IR-AGN ~ 1% with a strong upper limit of ~3% at z < 1. This uniformly low IR-AGN fraction at z < 1 is surprising given the fraction of 24 micrometer sources in the same galaxy clusters is observed to increase by about a factor of four from z ~ 0 to z ~ 1; this indicates that most of the detected 24 micrometer flux is due to star formation. Only in our single galaxy cluster at z = 1.24 is the IR-AGN fraction measurably higher at ~15% (all members; ~70% for late-types only). In agreement with recent studies, we find the cluster IR-AGN are predominantly hosted by late-type galaxies with blue optical colors, i.e. members with recent/ongoing star formation. The four brightest IR-AGN are also X-ray sources; these IR+X-ray AGN all lie outside the cluster core (R_proj > 0.5 Mpc) and are hosted by highly morphologically disturbed members. Although our sample is limited, our results suggest that f_IR-AGN in massive galaxy clusters is not strongly correlated with star formation at z < 1, and that IR-AGN have a more prominent role at z &gt; 1.

The recognition of time-varying geometry (TVG) objects (in particular, humans) and their actions is a complex task due to common real-world sensing challenges, such as obstacles and environmental variations, as well as due to issues specific to TVG objects, such as self-occlusion. Herein, it is proposed that a multi-camera active-vision system, which dynamically selects camera poses in real-time, be used to improve TVG action sensing performance by selecting camera views on-line for near-optimal sensing-task performance. Active vision for TVG objects requires an on-line sensor-planning strategy that incorporates information about the object itself, including its current action, and information about the state of the environment, including obstacles, into the pose-selection process. Thus, the focus of this research is the development of a novel methodology for real-time sensing-system reconfiguration (active vision), designed specifically for the recognition of a single TVG object and its actions in a cluttered, dynamic environment, which may contain multiple other dynamic (maneuvering) obstacles. The proposed methodology was developed as a complete, customizable sensing-system framework which can be readily modified to suit a variety of specific TVG action-sensing tasks – a 10-stage pipeline real-time architecture. Sensor Agents capture and correct camera images, removing noise and lens distortion, and segment the images into regions of interest. A Synchronization Agent aligns multiple images from different cameras to a single ‘world-time.’ Point Tracking and De-Projection Agents detect, identify, and track points of interest in the resultant 2-D images, and form constraints in normalized camera coordinates using the tracked pixel coordinates. A 3-D Solver Agent combines all constraints to estimate world-coordinate positions for all visible features of the object-of-interest (OoI) 3-D articulated model. A Form-Recovery Agent uses an iterative process to combine model constraints, detected feature points, and other contextual information to produce an estimate of the OoI’s current form. This estimate is used by an Action-Recognition Agent to determine which action the OoI is performing, if any, from a library of known actions, using a feature-vector descriptor for identification. A Prediction Agent provides estimates of future OoI and obstacle poses, given past detected locations, and estimates of future OoI forms given the current action and past forms. Using all of the data accumulated in the pipeline, a Central Planning Agent implements a formal, mathematical optimization developed from the general sensing problem. The agent seeks to optimize a visibility metric, which is positively related to sensing-task performance, to select desirable, feasible, and achievable camera poses for the next sensing instant. Finally, a Referee Agent examines the complete set of chosen poses for consistency, enforces global rules not captured through the optimization, and maintains system functionality if a suitable solution cannot be determined. In order to validate the proposed methodology, rigorous experiments are also presented herein. They confirm the basic assumptions of active vision for TVG objects, and characterize the gains in sensing-task performance. Simulated experiments provide a method for rapid evaluation of new sensing tasks. These experiments demonstrate a tangible increase in single-action recognition performance over the use of a static-camera sensing system. Furthermore, they illustrate the need for feedback in the pose-selection process, allowing the system to incorporate knowledge of the OoI’s form and action. Later real-world, multi-action and multi-level action experiments demonstrate the same tangible increase when sensing real-world objects that perform multiple actions which may occur simultaneously, or at differing levels of detail. A final set of real-world experiments characterizes the real-time performance of the proposed methodology in relation to several important system design parameters, such as the number of obstacles in the environment, and the size of the action library. Overall, it is concluded that the proposed system tangibly increases TVG action-sensing performance, and can be generalized to a wide range of applications, including human-action sensing. Future research is proposed to develop similar methods to address deformable objects and multiple objects of interest.

This paper describes my participation in a MITRE Corporation sponsored computer science clinic project at Harvey Mudd College as my senior project. The goal of the project was to implement a landmark-based pose correction system as a component in a larger, existing face recognition system. The main contribution I made to the project was the implementation of the Active Shape Models (ASM) algorithm; the inner workings of ASM are explained as well as how the pose correction system makes use of it. Included is the most recent draft (as of this writing) of the final report that my teammates and I produced highlighting the year's accomplishments. Even though there are few quantitative results to show because the clinic program is ongoing, our qualitative results are quite promising.

The recognition of time-varying geometry (TVG) objects (in particular, humans) and their actions is a complex task due to common real-world sensing challenges, such as obstacles and environmental variations, as well as due to issues specific to TVG objects, such as self-occlusion. Herein, it is proposed that a multi-camera active-vision system, which dynamically selects camera poses in real-time, be used to improve TVG action sensing performance by selecting camera views on-line for near-optimal sensing-task performance. Active vision for TVG objects requires an on-line sensor-planning strategy that incorporates information about the object itself, including its current action, and information about the state of the environment, including obstacles, into the pose-selection process. Thus, the focus of this research is the development of a novel methodology for real-time sensing-system reconfiguration (active vision), designed specifically for the recognition of a single TVG object and its actions in a cluttered, dynamic environment, which may contain multiple other dynamic (maneuvering) obstacles. The proposed methodology was developed as a complete, customizable sensing-system framework which can be readily modified to suit a variety of specific TVG action-sensing tasks – a 10-stage pipeline real-time architecture. Sensor Agents capture and correct camera images, removing noise and lens distortion, and segment the images into regions of interest. A Synchronization Agent aligns multiple images from different cameras to a single ‘world-time.’ Point Tracking and De-Projection Agents detect, identify, and track points of interest in the resultant 2-D images, and form constraints in normalized camera coordinates using the tracked pixel coordinates. A 3-D Solver Agent combines all constraints to estimate world-coordinate positions for all visible features of the object-of-interest (OoI) 3-D articulated model. A Form-Recovery Agent uses an iterative process to combine model constraints, detected feature points, and other contextual information to produce an estimate of the OoI’s current form. This estimate is used by an Action-Recognition Agent to determine which action the OoI is performing, if any, from a library of known actions, using a feature-vector descriptor for identification. A Prediction Agent provides estimates of future OoI and obstacle poses, given past detected locations, and estimates of future OoI forms given the current action and past forms. Using all of the data accumulated in the pipeline, a Central Planning Agent implements a formal, mathematical optimization developed from the general sensing problem. The agent seeks to optimize a visibility metric, which is positively related to sensing-task performance, to select desirable, feasible, and achievable camera poses for the next sensing instant. Finally, a Referee Agent examines the complete set of chosen poses for consistency, enforces global rules not captured through the optimization, and maintains system functionality if a suitable solution cannot be determined. In order to validate the proposed methodology, rigorous experiments are also presented herein. They confirm the basic assumptions of active vision for TVG objects, and characterize the gains in sensing-task performance. Simulated experiments provide a method for rapid evaluation of new sensing tasks. These experiments demonstrate a tangible increase in single-action recognition performance over the use of a static-camera sensing system. Furthermore, they illustrate the need for feedback in the pose-selection process, allowing the system to incorporate knowledge of the OoI’s form and action. Later real-world, multi-action and multi-level action experiments demonstrate the same tangible increase when sensing real-world objects that perform multiple actions which may occur simultaneously, or at differing levels of detail. A final set of real-world experiments characterizes the real-time performance of the proposed methodology in relation to several important system design parameters, such as the number of obstacles in the environment, and the size of the action library. Overall, it is concluded that the proposed system tangibly increases TVG action-sensing performance, and can be generalized to a wide range of applications, including human-action sensing. Future research is proposed to develop similar methods to address deformable objects and multiple objects of interest.

AbstraktInledning Det råder en omfattande hälsoproblematik idag och fysisk inaktivitet är en bidragande orsak. Stillasittandet har ökat i västvärlden och för att fånga även dessa individer erbjuder Tv-spelsmarknaden fysiskt aktiva spel. Handkontroll och stillasittande byts ut mot individens kroppsrörelser. Det finns till och med spel som räknar individens kaloriförbrukning under spelets gång.Syfte Syftet med studien är att se om det råder en signifikant skillnad gällande kaloriförbrukning mellan Your shape och indirekt kalorimetri samt hur Your shape förhåller sig mot Nintendo Wii boxing utifrån de fysiologiska variabler hjärtfrekvens och kaloriförbrukning.Metod En experimentell studiedesign genomfördes på 12 deltagare (7 kvinnor och 5 fem män med en medelvärdeskaraktärisering på 22.2 år, 67,9 kg, 171,3 cm och 23,0 BMI) med hjälp av indirekt kalorimetri i form av en portabel Oxycon Mobile Pro. VO2 och HF uppmättes under testet för att därefter räkna ut testpersonernas kaloriförbrukning. Spelkonsolen bestod av Microsoft xbox 360 med tillhörande kinect samt AVG-spelet Your shape. Av de fyra träningspassen som Your shape erbjuder valdes konditionsboxning ut till denna studie. Den statistiska beräkningen sattes till P = 0.05 och ett parat T-test genomfördes i SPSS.Resultat En signifikant skillnad existerade (P = 0.03) gällande kaloriförbrukning mellan Your shape och Oxycon Mobile Pro. Medelvärdet för testet gällande tid var 32 min och Your shape resulterade i ett medelvärdet på 139.7 kcal och Oxycon Mobile Pro 157.8. En skillnad på 12 %. Your shape underestimerar kaloriförbrukningen jämfört mot Oxycon Mobile Pro.Your shape redovisade högre resultat för både kcal/min och HF jämfört mot Nintendo Wii boxing.Slutsats Det finns en signifikant skillnad mellan Your shape och Oxycon Mobile Pro, nämligen en underskattning från Your shape med 12 %. Vilket ger en god uppskattning av kalorimätaren. Your shape är ett modernare AVG-spel jämfört mot Wii boxing men kräver liknande fysisk påfrestning.

Aquesta tesi presenta treball original i aplicat en l'àrea del control i la col·locació de sensors/actuadors (S/A) en sistemes de Control Actiu de Soroll (ANC). Primer, s'han aplicat tècniques de control i identificació robustes per a aconseguir ANC. La fase d'identificació està basada en una proposta d'identificació robusta orientada al control, considerant descripcions del sistema tant paramètriques com no-paramètriques, així com quantificant la incertesa. El disseny del controlador compara les estructures de control feedback (FB), feedforward (FF) i híbrida (FB/FF). El controlador feedback és sintetitzat i avaluat en el marc del control robust, i s'ha dissenyat utilitzant control òptim H&#8734; plantejat com un problema de sensibilitats mixtes. El controlador FF és un identificador adaptatiu, basat en l'algorisme &#963; robustament normalitzat. S'han desenvolupat dues propostes per a decidir quina de les estructures de control és més eficient, aplicades a un conducte de 4 metres amb soroll de banda ampla. A més a més, s'han mostrat de manera explícita els compromisos entre identificació i control, les limitacions inherents a un llaç de control feedback, així com qüestions relatives a la implementació de sistemes ANC. També s'han tractat altres qüestions com la relació entre acompliment, ordre del controlador, models paramètrics/no-paramètrics i implementació en processadors digitals de senyal (DSP), així com s'han comparat resultats teòrics i experimentals en el conducte. Les llacunes que encara resten entre teoria i pràctica en aquest tipus d'aplicacions també s'han resumit. D'altra banda, en aquest treball també es tracta el problema de com quantificar la col·locació de sensors i actuadors, amb la finalitat de controlar un sistema físic determinat. La mesura per a determinar la millor localització de S/A es basa en un criteri de llaç tancat orientat al control, el qual optimitza tant acompliment com qüestions pràctiques d'implementació. Aquesta mesura hauria de calcular-se abans del disseny, implementació i prova del controlador. La utilització d'aquesta mesura minimitza la prova combinatòria de controladors en totes les possibles combinacions de S/A. Per a aconseguir-ho, s'han definit diferents mesures que pesen l'acompliment potencial en llaç tancat, la robustesa, el número de condició de la planta (guanys relatius entrada/sortida (I/O)) així com altres qüestions d'implementació, com l'ordre del controlador. Aquestes poden calcular-se utilitzant software estàndard, tant per a models d'una-entrada-una-sortida (SISO) com per a models de múltiples-entrades-múltiples-sortides (MIMO) i poden aplicar-se a múltiples problemes d'enginyeria, ja siguin mecànics, acústics, aeroespacials, etc. En aquest treball, aquests resultats també s'han il·lustrat amb l'aplicació ANC presentada i validat amb dades experimentals. Com a resultat d'aplicar aquestes mesures, s'obté la localització de S/A que aconsegueix la millor atenuació del soroll en llaç tancat amb el menor ordre possible del controlador. / Esta tesis presenta trabajo original y aplicado en el área del control y la colocación de sensores/actuadores (S/A) en sistemas de Control Activo de Ruido (ANC). Primero, se han aplicado técnicas de control e identificación robustas para conseguir ANC. La fase de identificación está basada en una propuesta de identificación robusta orientada al control, considerando descripciones del sistema tanto paramétricas como no-paramétricas, así como cuantificando la incertidumbre. El diseño del controlador compara las estructuras de control feedback (FB), feedforward (FF) e híbrida (FB/FF). El controlador feedback es sintetizado y evaluado en el marco del control robusto, y se ha diseñado utilizando control óptimo H&#8734; planteado como un problema de sensibilidades mixtas. El controlador FF es un identificador adaptativo, basado en el algoritmo &#963; robustamente normalizado. Se han desarrollado dos propuestas para decidir cual de las estructuras de control es más eficiente, aplicadas a un conducto de 4 metros con ruido de banda ancha. Además, se han mostrado de manera explícita los compromisos entre identificación y control, las limitaciones inherentes a un lazo feedback, así como cuestiones relativas a la implementación de sistemas ANC. También se han tratado otras cuestiones como la relación entre desempeño, orden del controlador, modelos paramétricos/no-paramétricos e implementación en procesadores digitales de señal (DSP), así como se han comparado resultados teóricos y experimentales en el conducto. Las lagunas que aún quedan entre teoría y práctica en este tipo de aplicaciones también se han resumido. Por otra parte, en este trabajo se trata también el problema de como cuantificar la colocación de sensores y actuadores, con la finalidad de controlar un sistema físico determinado. La medida para determinar la mejor localización de S/A se basa en un criterio de lazo cerrado orientado al control, el cual optimiza tanto desempeño como cuestiones prácticas de implementación. Esta medida debería calcularse antes del diseño, implementación y prueba del controlador. La utilización de esta medida minimiza la prueba combinatoria de controladores en todas las posibles combinaciones de S/A. Para conseguirlo, se han definido distintas medidas que pesan el desempeño potencial en lazo cerrado, la robustez, el número de condición de la planta (ganancias relativas entrada/salida (I/O)) y otras cuestiones de implementación, como el orden del controlador. Éstas pueden calcularse utilizando software estándar, tanto para modelos de una-entrada-una-salida (SISO) como para modelos de múltiples-entradas-múltiples-salidas (MIMO) y pueden aplicarse a múltiples problemas ingenieriles, ya sean mecánicos, acústicos, aeroespaciales, etc. En este trabajo, estos resultados también son ilustrados con la aplicación ANC presentada y validados con datos experimentales. Como resultado de aplicar estas medidas, se obtiene la localización de S/A que consigue la mejor atenuación de ruido en lazo cerrado con el menor orden posible del controlador. / This thesis presents novel and applied work in the area of control and sensor/actuator (S/A) allocation in Active Noise Control (ANC) systems. First, robust identification and control techniques to perform ANC have been applied. The identification phase is based on a control-oriented robust identification approach that considers both parametric and nonparametric descriptions of the system, and quantifies the uncertainty. The controller design compares the feedback (FB), feedforward (FF) and hybrid (FB/FF) control structures. The feedback control is synthesized and evaluated in the robust control framework, and it is designed using H&#8734; optimal control as a mixed-sensitivity problem. The FF controller is an adaptive identifier, based on the robustly normalized &#963;-algorithm. Two approaches are developed to decide which control structure is more efficient on a 4-m duct example with broadband noise. In addition, the compromises between identification and control, the inherent limitations of feedback and implementation issues in ANC are explicitly pointed out. Relations between performance, controller order, parametric/nonparametric models and digital signal processor (DSP) implementation are discussed. Theoretical and experimental results on the duct are compared. The gaps that still remain between theory and practice in this type of applications, are also outlined. Furthermore, this work considers the problem of quantifying the location of sensors and actuators in order to control a certain physical system. The measure to determine the best S/A location is based on a closed loop control-oriented criteria, which optimizes overall performance and practical implementation issues. In addition, it should be computed before the actual controller is designed, implemented and tested. The use of this measure minimizes the combinatorial controller testing over all possible S/A combinations. To this end, several measures have been defined which weight the potential closed-loop performance, robustness, plant condition number (input/output (I/O) relative gains) and implementation issues, such as the controller order. These may be computed with standard software, either for Single Input Single Output (SISO) models or Multiple Input Multiple Output (MIMO) models, and may be applied to many engineering problems: mechanics, acoustics, aerospace, etc. Here, these results are also illustrated with the prior ANC example and validated against experimental data. The outcome of applying these measures is the selection of the S/A location which achieves the best closed loop noise attenuation with the lowest possible controller order.

Pas de disponible

Results from several studies are presented which detail explorations of the physical and spectral properties of low luminosity active galactic nuclei. An initial Sagittarius A* general relativistic magnetohydrodynamic simulation and Monte Carlo radiation transport model suggests accretion rate changes as the dominant flaring method. A similar study on M87 introduces new methods to the Monte Carlo model for increased consistency in highly energetic sources. Again, accretion rate variation seems most appropriate to explain spectral transients. To more closely resolve the methods of particle energization in active galactic nuclei accretion disks, a series of localized shearing box simulations explores the effect of numerical resolution on the development of current sheets. A particular focus on numerically describing converged current sheet formation will provide new methods for consideration of turbulence in accretion disks.

Image segmentation is an important step in image processing, with many applications such as pattern recognition, object detection, and medical image analysis. It is a technique that separates objects of interests from the background in an image. Geometric active contour is a recent image segmentation method that overcomes previous problems with snakes. It is an attractive method for medical image segmentation as it is able to capture the object of interest in one continuous curve. The theory and implementation details of geometric active contours are discussed in this work. The robustness of the algorithm is tested through a series of tests, involving both synthetic images and medical images. Curve leaking past boundaries is a common problem in cases of non-ideal edges. Noise is also problematic for the advancement of the curve. Smoothing and parameters selection are discussed as ways to help solve these problems. This work also explores the incorporation of the multi-resolution method of Gaussian pyramids into the algorithm. Multi-resolution methods, used extensively in the areas of denoising and edge-selection, can help capture the spatial structure of an image. Results show that similar to the multi-resolution methods applied to parametric active contours, the multi-resolution can greatly increase the computation without sacrificing performance. In fact, results show that with successive smoothing and sub-sampling, performance often improves. Although smoothing and parameter adjustment help improve the performance of geometric active contours, the edge-based approach is still localized and the improvement is limited. Region-based approaches are recommended for further work on active contours.