La perception est le moyen par lequel nous connaissons le monde extérieur. C’est grâce à nos perceptions que nous sommes capables d’interagir avec notre environnement et d’accomplir de nombreuses actions du quotidien comme se repérer, se déplacer, reconnaître des objets et autres. Cette perception n’est pas juste passive comme peut l’être une sensation, elle comporte des aspects actifs. En particulier, elle peut être orientée dans un but précis, permettant de filtrer les données pour ne traiter que les plus pertinentes. Si la perception humaine est particulièrement efficace, la perception artificielle, elle, demeure un problème complexe qui se heurte à de nombreuses difficultés. Ainsi, les changements de conditions de perception comme des modifications de l’illumination ou des occultations partielles de l’objet à percevoir doivent pouvoir être gérées efficacement. Pour résoudre ces difficultés, s’inspirer de la perception humaine semble être une piste intéressante. Ce manuscrit propose un système de perception polyvalent et générique reposant sur une stratégie de perception active. Pour ce faire, nous proposons un algorithme Top-Down utilisant un modèle en parties. Le problème de perception est transformé en un problème d’estimation d’un vecteur de caractéristiques. La détection des différentes parties permet de réaliser cette estimation. Le système de perception proposé est un algorithme itératif multi-capteurs. À chaque itération, il sélectionne au mieux, en fonction des objectifs fixés par l’application, la partie à détecter ainsi que les meilleurs capteur et détecteur compatibles. Un réseau bayésien est utilisé pour prendre en compte les événements incertains pouvant survenir lors de ce processus comme la défaillance d’un détecteur ou la non existence potentielle d’une partie donnée. Un processus de focalisation à la fois spatiale et de caractéristiques permet d’améliorer la détection en augmentant le rapport signal sur bruit, en restreignant la zone de recherche pour une partie et en éliminant certains des candidats trouvés. Ce processus de focalisation permet aussi de réduire les temps de calcul et de restreindre l’influence des distracteurs. L’ajout de nouveaux capteurs, détecteurs ou parties se fait simplement. De plus, l’utilisation d’un réseau bayésien permet une grande flexibilité au niveau de la modélisation des événements pris en compte : il est facile de rajouter de nouveaux événements pour obtenir une modélisation plus réaliste. L’algorithme proposé a été utilisé pour plusieurs applications incluant de la reconnaissance d’objets, de l’estimation fine de pose et de la localisation. / Perception is the way by which we know the outside world. Thanks to our perceptions we are able to interact with our environment and to achieve various everyday life actions as locating or moving in an environment, or recognizing objects. Perception is not passive whereas sensations are, it has active components. In particular, perception can be oriented for a specific purpose allowing to filter data and to take care only of the most relevant. If human perception is particularly effective, artificial perception remains a complex problem with a lot of non solved difficulties. For example, changes of perception conditions as modification of illumination or partial occultation of the searched object must be effectively managed. This thesis proposes a system of perception based on a strategy of active perception which can adapt itself to various applications. To do it, we propose an algorithm Top-Down using a part-based model. The problem of perception is transformed into a problem of estimation of a characteristics vector. The detection of the different parts constituting the searched object allows to realize this estimation. The proposed perceptive system is an iterative and multi-sensors algorithm. In every iteration, it selects, at best, according to the application objectives, the part to detect and the best compatible sensor and detector. A bayesian network is used to take into account uncertain events which can arise during this process as detector failure or potential non existing part. A focus process consisting of a spatial focus and of a characteristics focus, improves the detection by restricting the search area, by improving the signal to noise ratio and by eliminating some erroneous candidates. This focus process also allows to reduce computation time and to restrict influence of distractors. Adding a part, a sensor or a detector is simple. Furthermore, the use of a bayesian network allows to be flexible in the events modelisation : it is easy to add new events to obtain a more realistic modelisation. The proposed algorithm has been used for several applications including object’s recognition, fine pose estimation and localization. So, it is multi-purpose and generic.

No description available.

This diploma thesis deals with 3D reconstruction of objects using image analysis methods. The work includes mathematical theory associated with this problem, a procedure for creating 2D sharp images and 3D reconstruction itself. The outputs are 2D sharp images, 3D models, stl models. Different kinds of data are analyzed.

The aim of the master’s thesis, is to get acquainted with the dynamic focusing in conventional ultrasound systems, to modify this method for ultrasound computer tomography and to realize and interpret this modification on simulated and real data in a transparent graphic interface. The modification of the dynamic focusing was executed for the 3D ultrasound computer tomograph developed in Karlsruhe. Simulated data, that are necessary for focusing, were generated by existing program, which was adjusted. The new created program of the dynamic focusing was tested and evaluated with those simulated and real data.

This master’s thesis gives a description of fundus camera as a medical imaging system. Sub features of this system are explained in short, thus examples of certain construction variants are given. Furthermore, the work deals with image fusion and associated possibilities of digital image processing. One set of consecutive scanned images of human eye’s retina has been provided for the practical part of this work. During program processing of these data, decomposition of obtained images to single-color sequences is performed. For these partial monochromatic sequences, monomodal registration is performed, based on calculation of the brightness similarity criterion between the pairs of images. From the three created monochromatic sequences of registered images, new sequence of color images is created, using multimodal registration of each image triples. As a basis for similarity evaluation during multimodal registration, an information similarity criterion was used.

High-performance electromigration separation methods, capillary zone electrophoresis (CZE) and capillary isoelectric focusing (CIEF), have been applied to physico-chemical characterization of new synthetic low-molecular mass markers of isoelectric points. Amphoteric compounds on the basis of aminomethylnitrophenols, their derivatives and other structurally related substances were analyzed by CZE in a series of background electrolytes in a wide pH range, 1.86 - 11.18. From the measured pH dependencies of effective electrophoretic mobilities of analytes (beforehand corrected to reference temperature of 25 řC), their isoelectric points (pI) were determined. In addition, using the non-linear regression analysis of the above dependencies, acid-base dissociation constants (pKa) of ionogenic groups of selected analytes were calculated. Subsequently, the analytes with sharply defined isoelectric points were analyzed by CIEF. CIEF confirmed applicability of these compounds as markers of isoelectric points for calibration of pH gradient in CIEF in the determination of pI of amphoteric compounds, especially peptides and proteins. The determined pKa values of ionogenic groups in particular compounds will be utilized in the development of new pI markers with desired pI values.

Three-dimensional linear ray theory is used to investigate internal waves interacting with a Lamb-Chaplygin pancake vortex dipole. These interactions involve waves propagating in the same (co-propagating) and opposite (counter-propagating) horizontal directions as the dipole translation. Co-propagating internal waves in the vertical symmetry plane between the vortices of the dipole can approach critical levels where the wave energy is absorbed by the dipole or where the waves are overturned and possibly break. As wave breaking cannot be simulated with this linear model, changes in wave steepness are calculated to aid in estimating the onset of breaking. Counter-propagating internal waves in the vertical symmetry plane can experience horizontal and vertical reflections, including turning points similar to waves in two-dimensional steady shear. Wave capture is also a possible effect of either type of interaction, depending on initial wave properties and positioning relative to the vortex dipole. Away from the vertical symmetry plane, a spanwise converging (focusing) and diverging (defocusing) of wave energy is observed in co- and counter-propagating interactions as symmetric off-center rays interact with the dipole's individual vortices. Some off-center rays experience multiple horizontal refractions similar to wave trapping.

Antibodies are Y-shaped proteins. In the human body, antibodies areproduced by plasma cells, mainly T and B cells which are included inthe adaptive immune system. The production of antibodies is stimulatedby antigens. The binding between an antigen-specific antibody and itsantigen can be like the interconnect between a lock and a key.Therefore, antibodies are widely used as diagnostic tools for avariety of diseases but most importantly cancer. Some rapid diagnostictests are completely dependent on the specificity and reactivity ofantibodies such as UBC® Rapid produced by IDL Biotech AB. Therefore,the quality of these antibodies is important. This master thesis at IDL Biotech aimed to validate six anticytokeratinantibodies that are currently used in several rapid cancerdiagnostic tests produced by IDL. Antibody validation is a processwhere specificity, selectivity and reproductivity of an antibody isdemonstrated through specific laboratory investigations. During thisthesis, two laboratory methods were used to validate antibodies,namely, isoelectric focusing electrophoresis and the Attana QuartzCrystal Microbalance based biosensor. Isoelectric focusing electrophoresis (IEF) is a method that determinesproteins pI-values which can then reveal information about posttranslationalmodifications and protein sustainability during storage.IEF revealed changes in pI-values in two antibodies: AB2 and AB4. Attana biosensor analysis on AB1-5 showed that all antibodies havehigh specificity, reactivity and relatively high affinity to theircytokeratin targets. It also revealed that 4 antibodies (AB1 and AB3-5) have lower cross-reactivity with other cytokeratins than theirtarget cytokeratins compared to AB2. Keywords: Antibody validation, Isoelectric focusing, QCM, Attanabiosensor, biosensors, rapid diagnostics, epithelial carcinomas.

Acoustic focusing of microscale protein crystals with acoustophoresis technology could reduce clogs during experiments with the scientific technique serial femtosecond x-ray crystallography (SFX). SFX determines molecular structures of proteins, these structures are valuable in drug discovery and fundamental biomedical research. Lysozyme crystals were focused in their own mother liquor and dilutions with PBS buffer. The aim of these tests were to study how the acoustic contrast factor Φ changes with the medium. Recorded experiments were analyzed using the particle tracking software Trackmate to extract velocities and radii. The lysozyme crystals changed morphologies in large dilutions of PBS buffert, they either became rounder or broke into fragments. The changed forms are likely caused by dissolution behaviors; some dilutions were unstable, but not unstable enough to dissolve the crystals completely.  Measured velocities during focusing of the crystals had large variance. Sinusoidal fits of the velocities had significant increases in amplitudes for larger dilutions of PBS. A change in acoustic contrast factor Φ could be the cause for the increased amplitudes, but the results do not rule out other causes. There are currently major knowledge gaps about using protein crystals as particles with acoustophoresis technologies, hence many ideas for future works have been proposed in this master thesis report. / Akustisk fokusering av mikrometers-stora proteinkristaller med hjälp av ultraljudsteknik skulle kunna reducera proppar under experiment med tekniken seriell femtosekundskristallografi (SFX). SFX kan avgöra vilken struktur proteinmolekyler har, dessa strukturer är värdefulla för industriell utveckling av nya läkemedel och fundamental biomedicinsk forskning. Proteinkristaller av lysozym har fokuserats i sin egen kristalliseringslösning och utspädningar av PBS. Målet med de här experimenten var att se om den akustiska flödeskontrasten Φ kunde bli påverkad. Fokuseringar spelades in och partiklarnas hastigheter vid fokuseringarna mättes med hjälp av Trackmate (en mjukvara för partikelspårning). I de större utspädningarna förändrades lysozymkristallernas former, de blev antingen mer runda eller så blev de små bitar. Förändringen skedde förmodligen på grund av upplösningsmekanismer i instabila utspädningar.  Uppmätta hastigheter vid testerna hade stor varians, men kurvanpassningar av data tyder på att hastigheterna ökar signifikant med utspädningar av PBS. Den akustiska flödeskontrasten Φ kan vara orsaken för ökade hastigheter, men andra hypoteser gick inte att utesluta med resultaten. Stora kunskapshål finns om akustisk fokusering av proteinkristaller, därför har många idéer för framtida experiment och arbeten föreslagits i rapporten.

<p>Microdroplets have large and varied areas of application ranging from document printing to complex lab-on-chip devices. Lab-on-chip systems often require precise volume control as well as high throughput operations. Microdroplets fulfill these requirements and have become a staple in these devices. The work presented in this thesis involves the design and characterization of two individual devices capable of droplet generation utilizing flow focusing and electrogeneration methods.</p> <p>The first design involved the generation of gel microdroplets utilizing the flow focusing technique. This device proved to be robust and reliable producing large volumes of uniformly mixed droplets. Long term operation of this device was analyzed and determined to be a feasible route for the manufacture of large quantities of droplets. The device was operated for over 30 hours creating gel droplets ranging from 40-200 μm in diameter with acceptable polydispersities for use in drug release studies.</p> <p>The second device involved the design and characterization of a system for the electrogeneration of microdroplets. This novel device involved the injection of droplets via high voltage and high frequency signals into a cross-flow of oil. The droplet generation was characterized and different droplet generation modes were observed. With the careful selection of parameters ideal conditions were obtained to generate monodisperse droplets of sizes ranging from under 5 to over 100 μm in a highly repeatable manner.</p> <p>To conclude, two separate microfluidic droplet generation devices operating in distinct modes were designed and analyzed. These devices are robust, reliable, and flexible with some applications being tested.</p> / Master of Applied Science (MASc)