Outlining of the prostate boundary in ultrasound images is a very useful procedure performed and subsequently used by clinicians. The contribution of the resulting segmentation is twofold. First of all, the segmentation of the prostate glands can be used to analyze the size, geometry, and volume of the gland. Such analysis is useful as it is known that the former quantities used in conjunction with a PSA blood test can be used as an indicator of malignancy in the gland itself. The second purpose of accurate segmentation is for treatment planning purposes. In brachetherapy, commonly used to treat localized prostate cancer, the accurate location of the prostate must be found so that the radioactive seeds can be placed precisely in the malignant regions. Unfortunately, the current method of segmentation of ultrasound images is performed manually by expert radiologists. Due to the abundance of ultrasound data, the process of manual segmentation can be extremely time consuming and inefficient. A much more desirable way to perform the segmentation process is through automatic procedures, which should be able to accurately and efficiently extract the boundary of the prostate gland with minimal user intervention. This is the ultimate goal of the proposed approach. The proposed segmentation algorithm uses a probability distribution tracking framework to accurately and efficiently perform the task at hand. The basis for this methodology is to extract image and shape features from available manually segmented ultrasound images for which the actual prostate region is known. Then, the segmentation algorithm seeks a region in new ultrasound images whose features closely mirror the learned features of known prostate regions. Promising results were achieved using this method in a series of in silico and in vivo experiments.

Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal

Résumé: L’Imagerie par Résonance Magnétique fonctionnelle est devenue un des outils principaux pour mesurer l’activité cérébrale chez l’Homme de façon non-invasive. Le type de séquence plus fréquemment utilisé à ces fins est le signal dépendant du niveau d'oxvgène sanguin (BOLD signal) car il reflète les changements hémodynamiques associés à l’activité cérébrale et offr l’avantage d'être relativement facile à mesurer. Cependant, cette mesure est dépendante du niveau de désoxyhémoglobine, présent dans les veines. L'architecture structurelle de ces dernières varie considérablement d'une région du cerveau à l'autre et cela fait en sorte qu'il est ardu de savoir si les différences inter-régionales sont dues à des différences au niveau de l’activité neuronale ou de la structure vasculaire. C'est pour répondre à cette question que nous avons observé la relation entre les variations régionales de densité vasculaire (VAD) et l’amplitude du signal BOLD au repos et en présence d'un stimulus. Nous avons implémenté une méthode de segmentation des veines pour des images en pondération de susceptibilité (SWI). Nous avons également utilisé une méthode de segmentation des tissus (Morphométrie Basée sur les Voxels (VBM )) pour séparer les tissus et calculer une mesure de densité de matière. Nous avons trouvé qu'indépendamment de la taille des voxels, les variations régionales d'amplitude du signal BOLD au repos et du signal évoqué par une tâche sont mieux corrélées avec la VAD qu'avec la densité de matière grise. En utilisant un modèle linéaire général, nous avons observé qu'une grande partie des variations régionales au repos pouvait être modélisée par la VAD. En somme, nos résultats suggèrent que le signal BOLD au repos est intimement relié à la structure vasculaire. Utiliser la densité vasculaire afin de calibrer les mesures d'amplitude du signal BOLD au repos nous permettra d'interpréter de façon plus précise et d'obtenir de meilleures informations sur les différences observées entre plusieurs groupes de sujets et de patients. // Abstract: Functional Magnetic Resonance Imaging (fMRI) has become one of the primary tools used for non-invasively measuring brain activity in humans. For the most part, the blood oxygen level-dependent (BOLD) contrast is used, which reflects changes in heinodynamics associated with active brain tissue. The main advantage of the BOLD signal is that it is relatively easy to measure and thus is often used as a proxy for comparing brain function across population groups (i.e. control vs. patient). However, it is particularly weighted towards veins whose structural architecture is known to vary considerably across the brain. This makes it difficult to interpret whether differences in BOLD between cortical areas reflect true differences in neu-ral activity or vascular structure. We therefore investigated how regional variations of vascular density (VAD) relate to the amplitude of resting-state and task-evoked BOLD signals. To address this, we first developed an automated method for segmen-ting veins in images acquired with susceptibility weighted imaging (SWI), allowing us to visualize the venoms vascular tree across the brain. In 19 healthy subjects, we then applied Voxel-Based Morphometry (VBNI) to Ti-weighted images and compu-ted regional measures of gray matter density (GNID). We found that, independent of spatial scale. regional variations in resting-state and task-evoked fNIRI amplitude was better correlated to VAD compared to GRID. Using a general linear model (GLNI). it was observed that the bulk of regional variance in resting-state activity could be modelled by VAD. Taken together, our results suggest that resting-state BOLD signals are signifi-cantly related to the underlying structure of the brain vascular system. Calibrating resting BOLD activity by venous structure may result in a more accurate interpreta-tion of differences observed between cortical areas and/or individuals.

The latest advances in digital cameras combined with powerful computer software enable us to store high-quality microscopy images of specimen. Studying hundreds of images manually is very time consuming and has the problem of human subjectivity and inconsistency. Quantitative image analysis is an emerging field and has found its way into analysis of microscopy images for clinical and research purposes. When developing a pipeline, it is important that its components are simple enough to be generalized and have predictive value. This thesis addresses the automation of quantitative analysis of tissue in two different fields: pathology and plant biology. Testicular tissue is a complex structure consisting of seminiferous tubules. The epithelial layer of a seminiferous tubule contains cells that differentiate from primitive germ cells to spermatozoa in a number of steps. These steps are combined in 12 stages in the cycle of the seminiferous epithelium in the mink. The society of toxicological pathology recommends classifying the testicular epithelial into different stages when assessing tissue damage to determine if the dynamics in the spermatogenic cycle have been disturbed. This thesis presents two automated methods for fast and robust segmentation of tubules, and an automated method of staging them. For better accuracy and statistical analysis, we proposed to pool stages into 5 groups. This pooling is suggested based on the morphology of tubules. In the 5 stage case, the overall number of correctly classified tubules is 79.6%. Contextual information on the localization of fluorescence in microscopy images of plant specimen help us to better understand differentiation and maturation of stem cells into tissues. We propose a pipeline for automated segmentation and classification of the cells in a whole cross-section of Arabidopsis hypocotyl, stem, or root. As proof-of-concept that the classification provides a meaningful basis to group cells for fluorescence characterization, we probed tissues with an antibody specific to xylem vessels in the secondary cell wall. Fluorescence intensity in different classes of cells is measured by the pipeline. The measurement results clearly show that the xylem vessels are the dominant cell type that exhibit a fluorescence signal.

Den 29 januari 2015 presenterade den oberoende aktören och marknadsbevakaren Ekowebför sjunde året i rad sin marknadsrapport över Sveriges ekologiska livsmedelsförsäljning. Detsom presenterades i årets rapport var en rekordsiffra som visade att den ekologiskalivsmedelsförsäljningen i Sverige ökade under 2014 med hela 38 procent. Den totalamarknadsandelen är dock relativt låg och prognoser visar att det finns en stor mängdkonsumenter som är på väg att förändra sina ekologiska inköpsvanor.Tidigare forskning har visat på svårigheter med att segmentera den ekologiskalivsmedelsmarknaden då ekologiska konsumenter varken är homogena i demografiska ellerpersonliga aspekter. Forskning gällande den ekologiska livsmedelsmarknaden visarmotsägelsefulla resultat kring huruvida olika segmenteringsvariabler är effektiva eller inte.Det finns en stor och oenhetlig grupp konsumenter som handlar ekologiska livsmedel utifrånen mer tillfällig grund och det är denna grupp som den aktuella undersökningen syftar till attstudera och segmentera. Den aktuella undersökningen avser att öka förståelsen kring desvårigheter som finns med segmentering av den ekologiska livsmedelsmarknaden.Syftet med den aktuella studien är att fördjupa diskussionen kring komplexiteten isegmentering av marknaden för ekologisk. Forskningsfrågan som ska besvaras är följande:vilka är de främsta orsakerna till svårigheterna med att segmentera den ekologiskalivsmedelsmarknaden. Forskningsmetoden har genomförts med en genomgång av tidigareforskning för att förstå fenomenet, dess bakgrund och vilka studier som tidigare gjorts påområdet. Den valda forskningsmetoden är kvalitativ och för att förklara syftet med denaktuella studien har en enkätundersökning genomförts. Enkätundersökningen har sedananalyserats och tolkats med hjälp av vald teori.Den aktuella undersökningen har bidragit till en fördjupning av komplexiteten medsegmentering av den ekologiska livsmedelsmarknaden. Det som framkommit iundersökningen är att en mycket omfattande förståelse för konsumenterna krävs för att lyckasmed en effektiv segmentering. Därmed enbart de traditionella segmenteringsvariablernatillräckliga. Resultaten av den aktuella undersökningen visar att skillnader i konsumentersupplevda och faktiska kunskap kan påverka effektiviteten av segmenteringen då en högupplevd kunskap minskar mottagligheten för ny kunskap och information. Segmentet somidentifierats i studien har en hög subjektiv kunskap men en låg objektiv kunskap som kanhindra dem från att ta del av budskap utformade med hjälp av segmentering. / In January 2015, Ekoweb presented their yearly market report of the Swedish organic foodmarket. The results that were presented in the report showed a remarkable increase in organicfood sales during 2014. According to what was presented by Ekoweb, the organic food salesin Sweden increased by, as much as 38 percent during the last year. Although the marketingshare for organic food in Sweden is still relatively small. Forecasts show that there are a greatgroup of consumers who are about to change their consumer behavior concerning organicfood and that the increase of organic food sales is in the beginning of its progress.Previous studies made in the organic food market have shown difficulties in segmenting theorganic food consumer. One explanation is that organic food buyers tend to differ in bothdemographic and personal aspects. Studies shows contradictory results regarding whetherdifferent segmentation variables are effective or not in the organic food market. There are agreat group of consumers that are homogeneous in both demographic and personal aspectsand who have potential in start buying more organic food. This group of consumers is infocus in the following study. The current study wishes to enhance the understandingregarding the difficulties that exist in segmentation of the organic food market.The purpose of this study is to improve the discussion considering the complexity insegmentation of the organic food market. The research question that this study aims toanswer is; which are the main reasons to the difficulties with segmentation of the organicfood market? In order to answer the research question and to understand the phenomena areview over previous research has been made. The research method in the current study isqualitative and the empirical material has been collected using a survey, which has beenanalysed with the theories presented in the report.The results of the study have contributed with a deeper understanding of the complexity insegmenting the organic food market. The most important conclusion is that a comprehensiveunderstanding of the consumers is required in creating an effective segmentation on theorganic food market. A segmentation using only the traditional segmentation variables istherefore not always enough. The results show variances in consumers’ subjective andobjective knowledge concerning organic food, which can aggravate the segmentation. This,since a high subjective knowledge decreases the consumer’s willingness to receive newknowledge and learn more about what differs organic food from non-organic food. Thesegment that was identified in the study, consumers that have potential in consuming organicfood more frequently, has a high subjective knowledge but a low objective knowledge, whichcan prevent them from becoming involved in marketing actions formed using segmentation.

No description available.

This thesis proposes a control strategy for small-signal dynamic stability enhancement of a DC-segmented AC power system. This control strategy provides four control schemes based on HVDC supplementary control or modification of the operational condition of the HVDC control system to improve the system stability by (i) damping the oscillations within a segment using supplementary current control of a line-commutated HVDC link, based on the model predictive control (MPC) method (control scheme 1), (ii) minimizing the propagation of dynamics among the segments based on a coordinated linear quadratic Gaussian (LQG)-based supplementary control (control scheme 2), (iii) selectively distributing the oscillations among the segments based on a coordinated LQG-based supplementary control (control scheme 3) and (iv) changing the set-points of the HVDC control system in the direction determined based on the sensitivities of the Hopf stability margin to the HVDC links set-points (control scheme 4). Depending on the system characteristics, one or more of the proposed control schemes may be effective for mitigating the system oscillations. Study results show that (i) control scheme 1 leads to damped low-frequency oscillations and provides fast recovery times after faults, (ii) under control scheme 2, each segment in a DC-segmented system can experience major disturbances without causing adjacent segments to experience the disturbances with the same degree of severity, (iii) control scheme 3 enables the controlled propagation of the oscillations among segments and damps out the oscillatory dynamics in the faulted segment, and (iv) control scheme 4 improves the stability margin for Hopf bifurcations caused by various events. Since power system software tools exhibit limitations for advanced control design, this thesis also presents a methodology based on MATLAB/Simulink software to (i) systematically construct the nonlinear differential-algebraic model of an AC-DC system, and (ii) automatically extract a linearized state space model of the system for the design of the proposed control schemes. The nonlinear model also serves as a platform for the time-domain simulation of power system dynamics. The accuracy of the MATLAB/Simulink-based AC-DC power system model and time-domain simulation platform is validated by comparison against PSS/E.

Light Detection And Ranging (LiDAR) mapping has been emerging over the past few years as a mainstream tool for the dense acquisition of three dimensional point data. Besides the conventional mapping missions, LiDAR systems have proven to be very useful for a wide spectrum of applications such as forestry, structural deformation analysis, urban mapping, and reverse engineering. The wide application scope of LiDAR lead to the development of many laser scanning technologies that are mountable on multiple platforms (i.e., airborne, mobile terrestrial, and tripod mounted), this caused variations in the characteristics and quality of the generated point clouds. As a result of the increased popularity and diversity of laser scanners, one should address the heterogeneous LiDAR data post processing (i.e., registration and segmentation) problems adequately. Current LiDAR integration techniques do not take into account the varying nature of laser scans originating from various platforms. In this dissertation, the author proposes a methodology designed particularly for the registration and segmentation of heterogeneous LiDAR data. A data characterization and filtering step is proposed to populate the points’ attributes and remove non-planar LiDAR points. Then, a modified version of the Iterative Closest Point (ICP), denoted by the Iterative Closest Projected Point (ICPP) is designed for the registration of heterogeneous scans to remove any misalignments between overlapping strips. Next, a region-growing-based heterogeneous segmentation algorithm is developed to ensure the proper extraction of planar segments from the point clouds. Validation experiments show that the proposed heterogeneous registration can successfully align airborne and terrestrial datasets despite the great differences in their point density and their noise level. In addition, similar testes have been conducted to examine the heterogeneous segmentation and it is shown that one is able to identify common planar features in airborne and terrestrial data without resampling or manipulating the data in any way. The work presented in this dissertation provides a framework for the registration and segmentation of airborne and terrestrial laser scans which has a positive impact on the completeness of the scanned feature. Therefore, the derived products from these point clouds have higher accuracy as seen in the full manuscript.

The evolution of technology in the textile industry reached the design of embroidery patterns for machine embroidery. In order to create quality designs the shapes to be embroidered need to be segmented into regions that define different parts. One of the objectives of our research is to develop a method to automatically segment the shapes and by doing so making the process faster and easier. Shape analysis is necessary to find a suitable method for this purpose. It includes the study of different ways to represent shapes. In this thesis we focus on shape representation through its skeleton. We make use of a shape's skeleton and the shape's boundary through the so-called feature transform to decide how to segment a shape and where to place the segment boundaries. The direction of stitches is another important specification in an embroidery design. We develop a technique to select the stitch orientation by defining direction lines using the skeleton curves and information from the boundary. We compute the intersections of segment boundaries and direction lines with the shape boundary for the final definition of the direction line segments. We demonstrate that our shape segmentation technique and the automatic placement of direction lines produce sufficient constrains for automated embroidery designs. We show examples for lettering, basic shapes, as well as simple and complex logos.

A computer simulation study is conducted to explore the interaction of alternative segmentation strategies and the competitiveness of the market environment, a goal that can neither be tackled by purely analytic approaches nor is sufficient and undistorted real market data available to deduct findings in an empirical manner. The fundamental idea of the simulation is to increase competition in the artificial marketplace and to study the influence of segmentation strategy and varying market conditions on organizational success. Success/failure is measured using two performance criteria: number of units sold and survival of firms over 36 periods of time. Three central findings emerge: (1) the more competitive a market environment, the more successful the concentrated market segmentation strategy, (2) increased levels of marketing budgets do not favour firms following a concentrated segmentation strategy and (3) frequent rethinking and strategy modification impairs organizations that concentrate on target segments. (author's abstract) / Series: Working Papers SFB "Adaptive Information Systems and Modelling in Economics and Management Science"