Computer-aided system for morphometric mandibular index computation (Using dental panoramic radiographs)

Objective: We propose and validate a computer aided system to measure three different mandibular indexes: cortical width, panoramic mandibular index and, mandibular alveolar bone resorption index. Study Design: Repeatability and reproducibility of the measurements are analyzed and compared to the manual estimation of the same indexes. Results: The proposed computerized system exhibits superior repeatability and reproducibility rates compared to standard manual methods. Moreover, the time required to perform the measurements using the proposed method is negligible compared to perform the measurements manually. Conclusions: We have proposed a very user friendly computerized method to measure three different morphometric mandibular indexes. From the results we can conclude that the system provides a practical manner to perform these measurements. It does not require an expert examiner and does not take more than 16 seconds per analysis. Thus, it may be suitable to diagnose osteoporosis using dental panoramic radiographs

Adaptive MCMC methods with applications in environmental and geophysical models

This work presents new, efficient Markov chain Monte Carlo (MCMC) simulation methods for statistical analysis in various modelling applications. When using MCMC methods, the model is simulated repeatedly to explore the probability distribution describing the uncertainties in model parameters and predictions. In adaptive MCMC methods based on the Metropolis-Hastings algorithm, the proposal distribution needed by the algorithm learns from the target distribution as the simulation proceeds. Adaptive MCMC methods have been subject of intensive research lately, as they open a way for essentially easier use of the methodology. The lack of user-friendly computer programs has been a main obstacle for wider acceptance of the methods. This work provides two new adaptive MCMC methods: DRAM and AARJ. The DRAM method has been built especially to work in high dimensional and non-linear problems. The AARJ method is an extension to DRAM for model selection problems, where the mathematical formulation of the model is uncertain and we want simultaneously to fit several different models to the same observations. The methods were developed while keeping in mind the needs of modelling applications typical in environmental sciences. The development work has been pursued while working with several application projects. The applications presented in this work are: a winter time oxygen concentration model for Lake Tuusulanjärvi and adaptive control of the aerator; a nutrition model for Lake Pyhäjärvi and lake management planning; validation of the algorithms of the GOMOS ozone remote sensing instrument on board the Envisat satellite of European Space Agency and the study of the effects of aerosol model selection on the GOMOS algorithm.

Communication challenges in virtual teams – a case study of implementing computer-mediated communication technologies in inter-organisational virtual teams

This thesis seeks to answer, if communication challenges in virtual teams can be overcome with the help of computer-mediated communication. Virtual teams are becoming more common work method in many global companies. In order for virtual teams to reach their maximum potential, effective asynchronous and synchronous methods for communication are needed. The thesis covers communication in virtual teams, as well as leadership and trust building in virtual environments with the help of CMC. First, the communication challenges in virtual teams are identified by using a framework of knowledge sharing barriers in virtual teams by Rosen et al. (2007) Secondly, the leadership and trust in virtual teams are defined in the context of CMC. The performance of virtual teams is evaluated in the case study by exploiting these three dimensions. With the help of a case study of two virtual teams, the practical issues related to selecting and implementing communication technologies as well as overcoming knowledge sharing barriers is being discussed. The case studies involve a complex inter-organisational setting, where four companies are working together in order to maintain a new IT system. The communication difficulties are related to inadequate communication technologies, lack of trust and the undefined relationships of the stakeholders and the team members. As a result, it is suggested that communication technologies are needed in order to improve the virtual team performance, but are not however solely capable of solving the communication challenges in virtual teams. In addition, suitable leadership and trust between team members are required in order to improve the knowledge sharing and communication in virtual teams.

New kernel functions and learning methods for text and data mining

Recent advances in machine learning methods enable increasingly the automatic construction of various types of computer assisted methods that have been difficult or laborious to program by human experts. The tasks for which this kind of tools are needed arise in many areas, here especially in the fields of bioinformatics and natural language processing. The machine learning methods may not work satisfactorily if they are not appropriately tailored to the task in question. However, their learning performance can often be improved by taking advantage of deeper insight of the application domain or the learning problem at hand. This thesis considers developing kernel-based learning algorithms incorporating this kind of prior knowledge of the task in question in an advantageous way. Moreover, computationally efficient algorithms for training the learning machines for specific tasks are presented. In the context of kernel-based learning methods, the incorporation of prior knowledge is often done by designing appropriate kernel functions. Another well-known way is to develop cost functions that fit to the task under consideration. For disambiguation tasks in natural language, we develop kernel functions that take account of the positional information and the mutual similarities of words. It is shown that the use of this information significantly improves the disambiguation performance of the learning machine. Further, we design a new cost function that is better suitable for the task of information retrieval and for more general ranking problems than the cost functions designed for regression and classification. We also consider other applications of the kernel-based learning algorithms such as text categorization, and pattern recognition in differential display. We develop computationally efficient algorithms for training the considered learning machines with the proposed kernel functions. We also design a fast cross-validation algorithm for regularized least-squares type of learning algorithm. Further, an efficient version of the regularized least-squares algorithm that can be used together with the new cost function for preference learning and ranking tasks is proposed. In summary, we demonstrate that the incorporation of prior knowledge is possible and beneficial, and novel advanced kernels and cost functions can be used in algorithms efficiently.

Development of a systematic computer vision-based method to analyse and compare images of false identity documents for forensic intelligence purposes-Part I: acquisition, calibration and validation issues

Following their detection and seizure by police and border guard authorities, false identity and travel documents are usually scanned, producing digital images. This research investigates the potential of these images to classify false identity documents, highlight links between documents produced by a same modus operandi or same source, and thus support forensic intelligence efforts. Inspired by previous research work about digital images of Ecstasy tablets, a systematic and complete method has been developed to acquire, collect, process and compare images of false identity documents. This first part of the article highlights the critical steps of the method and the development of a prototype that processes regions of interest extracted from images. Acquisition conditions have been fine-tuned in order to optimise reproducibility and comparability of images. Different filters and comparison metrics have been evaluated and the performance of the method has been assessed using two calibration and validation sets of documents, made up of 101 Italian driving licenses and 96 Portuguese passports seized in Switzerland, among which some were known to come from common sources. Results indicate that the use of Hue and Edge filters or their combination to extract profiles from images, and then the comparison of profiles with a Canberra distance-based metric provides the most accurate classification of documents. The method appears also to be quick, efficient and inexpensive. It can be easily operated from remote locations and shared amongst different organisations, which makes it very convenient for future operational applications. The method could serve as a first fast triage method that may help target more resource-intensive profiling methods (based on a visual, physical or chemical examination of documents for instance). Its contribution to forensic intelligence and its application to several sets of false identity documents seized by police and border guards will be developed in a forthcoming article (part II).

Accuracy of computer-aided ultrasound as compared with magnetic resonance imaging in the evaluation of nonalcoholic fatty liver disease in obese and eutrophic adolescents

AbstractObjective:To compare the accuracy of computer-aided ultrasound (US) and magnetic resonance imaging (MRI) by means of hepatorenal gradient analysis in the evaluation of nonalcoholic fatty liver disease (NAFLD) in adolescents.Materials and Methods:This prospective, cross-sectional study evaluated 50 adolescents (aged 11–17 years), including 24 obese and 26 eutrophic individuals. All adolescents underwent computer-aided US, MRI, laboratory tests, and anthropometric evaluation. Sensitivity, specificity, positive and negative predictive values and accuracy were evaluated for both imaging methods, with subsequent generation of the receiver operating characteristic (ROC) curve and calculation of the area under the ROC curve to determine the most appropriate cutoff point for the hepatorenal gradient in order to predict the degree of steatosis, utilizing MRI results as the gold-standard.Results:The obese group included 29.2% girls and 70.8% boys, and the eutrophic group, 69.2% girls and 30.8% boys. The prevalence of NAFLD corresponded to 19.2% for the eutrophic group and 83% for the obese group. The ROC curve generated for the hepatorenal gradient with a cutoff point of 13 presented 100% sensitivity and 100% specificity. As the same cutoff point was considered for the eutrophic group, false-positive results were observed in 9.5% of cases (90.5% specificity) and false-negative results in 0% (100% sensitivity).Conclusion:Computer-aided US with hepatorenal gradient calculation is a simple and noninvasive technique for semiquantitative evaluation of hepatic echogenicity and could be useful in the follow-up of adolescents with NAFLD, population screening for this disease as well as for clinical studies.

A Game to Promote Understanding about UCD Methods and Process

The User-centered design (UCD) game is a tool forhuman-computer interaction practitioners to demonstrate the key user-centered design methodsand how they interrelate in the design process in an interactive and participatory manner. The target audiences are departments and institutions unfamiliar with UCD but whose work is related to the definition, creation, and update of a product or service.

Methods for construction and analysis of computational models in systems biology : applications to the modelling of the heat shock response and the self-assembly of intermediate filaments

Systems biology is a new, emerging and rapidly developing, multidisciplinary research field that aims to study biochemical and biological systems from a holistic perspective, with the goal of providing a comprehensive, system- level understanding of cellular behaviour. In this way, it addresses one of the greatest challenges faced by contemporary biology, which is to compre- hend the function of complex biological systems. Systems biology combines various methods that originate from scientific disciplines such as molecu- lar biology, chemistry, engineering sciences, mathematics, computer science and systems theory. Systems biology, unlike “traditional” biology, focuses on high-level concepts such as: network, component, robustness, efficiency, control, regulation, hierarchical design, synchronization, concurrency, and many others. The very terminology of systems biology is “foreign” to “tra- ditional” biology, marks its drastic shift in the research paradigm and it indicates close linkage of systems biology to computer science. One of the basic tools utilized in systems biology is the mathematical modelling of life processes tightly linked to experimental practice. The stud- ies contained in this thesis revolve around a number of challenges commonly encountered in the computational modelling in systems biology. The re- search comprises of the development and application of a broad range of methods originating in the fields of computer science and mathematics for construction and analysis of computational models in systems biology. In particular, the performed research is setup in the context of two biolog- ical phenomena chosen as modelling case studies: 1) the eukaryotic heat shock response and 2) the in vitro self-assembly of intermediate filaments, one of the main constituents of the cytoskeleton. The range of presented approaches spans from heuristic, through numerical and statistical to ana- lytical methods applied in the effort to formally describe and analyse the two biological processes. We notice however, that although applied to cer- tain case studies, the presented methods are not limited to them and can be utilized in the analysis of other biological mechanisms as well as com- plex systems in general. The full range of developed and applied modelling techniques as well as model analysis methodologies constitutes a rich mod- elling framework. Moreover, the presentation of the developed methods, their application to the two case studies and the discussions concerning their potentials and limitations point to the difficulties and challenges one encounters in computational modelling of biological systems. The problems of model identifiability, model comparison, model refinement, model inte- gration and extension, choice of the proper modelling framework and level of abstraction, or the choice of the proper scope of the model run through this thesis.

VVER-440 Thermal Hydraulics as a Computer Code Validation Challenge

This thesis concentrates on the validation of a generic thermal hydraulic computer code TRACE under the challenges of the VVER-440 reactor type. The code capability to model the VVER-440 geometry and thermal hydraulic phenomena specific to this reactor design has been examined and demonstrated acceptable. The main challenge in VVER-440 thermal hydraulics appeared in the modelling of the horizontal steam generator. The major challenge here is not in the code physics or numerics but in the formulation of a representative nodalization structure. Another VVER-440 specialty, the hot leg loop seals, challenges the system codes functionally in general, but proved readily representable. Computer code models have to be validated against experiments to achieve confidence in code models. When new computer code is to be used for nuclear power plant safety analysis, it must first be validated against a large variety of different experiments. The validation process has to cover both the code itself and the code input. Uncertainties of different nature are identified in the different phases of the validation procedure and can even be quantified. This thesis presents a novel approach to the input model validation and uncertainty evaluation in the different stages of the computer code validation procedure. This thesis also demonstrates that in the safety analysis, there are inevitably significant uncertainties that are not statistically quantifiable; they need to be and can be addressed by other, less simplistic means, ultimately relying on the competence of the analysts and the capability of the community to support the experimental verification of analytical assumptions. This method completes essentially the commonly used uncertainty assessment methods, which are usually conducted using only statistical methods.

Data analysis tools and methods for DNA microarray and high-throughput sequencing data

The recent rapid development of biotechnological approaches has enabled the production of large whole genome level biological data sets. In order to handle thesedata sets, reliable and efficient automated tools and methods for data processingand result interpretation are required. Bioinformatics, as the field of studying andprocessing biological data, tries to answer this need by combining methods and approaches across computer science, statistics, mathematics and engineering to studyand process biological data. The need is also increasing for tools that can be used by the biological researchers themselves who may not have a strong statistical or computational background, which requires creating tools and pipelines with intuitive user interfaces, robust analysis workflows and strong emphasis on result reportingand visualization. Within this thesis, several data analysis tools and methods have been developed for analyzing high-throughput biological data sets. These approaches, coveringseveral aspects of high-throughput data analysis, are specifically aimed for gene expression and genotyping data although in principle they are suitable for analyzing other data types as well. Coherent handling of the data across the various data analysis steps is highly important in order to ensure robust and reliable results. Thus,robust data analysis workflows are also described, putting the developed tools andmethods into a wider context. The choice of the correct analysis method may also depend on the properties of the specific data setandthereforeguidelinesforchoosing an optimal method are given. The data analysis tools, methods and workflows developed within this thesis have been applied to several research studies, of which two representative examplesare included in the thesis. The first study focuses on spermatogenesis in murinetestis and the second one examines cell lineage specification in mouse embryonicstem cells.

Automated UI testing. WBCT, a computer-automated user test environment tool.

Today, the user experience and usability in software application are becoming a major design issue due to the adaptation of many processes using new technologies. Therefore, the study of the user experience and usability might be included in every software development project and, thus, they should be tested to get traceable results. As a result of different testing methods to evaluate the concepts, a non-expert on the topic might have doubts on which option he/she should opt for and how to interpret the outcomes of the process. This work aims to create a process to ease the whole testing methodology based on the process created by Seffah et al. and a supporting software tool to follow the procedure of these testing methods for the user experience and usability.

Application of Reptation Quantum Monte Carlo and Related Methods to LiH

This work investigates mathematical details and computational aspects of Metropolis-Hastings reptation quantum Monte Carlo and its variants, in addition to the Bounce method and its variants. The issues that concern us include the sensitivity of these algorithms' target densities to the position of the trial electron density along the reptile, time-reversal symmetry of the propagators, and the length of the reptile. We calculate the ground-state energy and one-electron properties of LiH at its equilibrium geometry for all these algorithms. The importance sampling is performed with a single-determinant large Slater-type orbitals (STO) basis set. The computer codes were written to exploit the efficiencies engineered into modern, high-performance computing software. Using the Bounce method in the calculation of non-energy-related properties, those represented by operators that do not commute with the Hamiltonian, is a novel work. We found that the unmodified Bounce gives good ground state energy and very good one-electron properties. We attribute this to its favourable time-reversal symmetry in its target density's Green's functions. Breaking this symmetry gives poorer results. Use of a short reptile in the Bounce method does not alter the quality of the results. This suggests that in future applications one can use a shorter reptile to cut down the computational time dramatically.

Heuristic solution methods for multi-attribute vehicle routing problems

Le Problème de Tournées de Véhicules (PTV) est une clé importante pour gérér efficacement des systèmes logistiques, ce qui peut entraîner une amélioration du niveau de satisfaction de la clientèle. Ceci est fait en servant plus de clients dans un temps plus court. En terme général, il implique la planification des tournées d'une flotte de véhicules de capacité donnée basée à un ou plusieurs dépôts. Le but est de livrer ou collecter une certain quantité de marchandises à un ensemble des clients géographiquement dispersés, tout en respectant les contraintes de capacité des véhicules. Le PTV, comme classe de problèmes d'optimisation discrète et de grande complexité, a été étudié par de nombreux au cours des dernières décennies. Étant donné son importance pratique, des chercheurs dans les domaines de l'informatique, de la recherche opérationnelle et du génie industrielle ont mis au point des algorithmes très efficaces, de nature exacte ou heuristique, pour faire face aux différents types du PTV. Toutefois, les approches proposées pour le PTV ont souvent été accusées d'être trop concentrées sur des versions simplistes des problèmes de tournées de véhicules rencontrés dans des applications réelles. Par conséquent, les chercheurs sont récemment tournés vers des variantes du PTV qui auparavant étaient considérées trop difficiles à résoudre. Ces variantes incluent les attributs et les contraintes complexes observés dans les cas réels et fournissent des solutions qui sont exécutables dans la pratique. Ces extensions du PTV s'appellent Problème de Tournées de Véhicules Multi-Attributs (PTVMA). Le but principal de cette thèse est d'étudier les différents aspects pratiques de trois types de problèmes de tournées de véhicules multi-attributs qui seront modélisés dans celle-ci. En plus, puisque pour le PTV, comme pour la plupart des problèmes NP-complets, il est difficile de résoudre des instances de grande taille de façon optimale et dans un temps d'exécution raisonnable, nous nous tournons vers des méthodes approcheés à base d’heuristiques.

A brain-computer interface for navigation in virtual reality

L'interface cerveau-ordinateur (ICO) décode les signaux électriques du cerveau requise par l’électroencéphalographie et transforme ces signaux en commande pour contrôler un appareil ou un logiciel. Un nombre limité de tâches mentales ont été détectés et classifier par différents groupes de recherche. D’autres types de contrôle, par exemple l’exécution d'un mouvement du pied, réel ou imaginaire, peut modifier les ondes cérébrales du cortex moteur. Nous avons utilisé un ICO pour déterminer si nous pouvions faire une classification entre la navigation de type marche avant et arrière, en temps réel et en temps différé, en utilisant différentes méthodes. Dix personnes en bonne santé ont participé à l’expérience sur les ICO dans un tunnel virtuel. L’expérience fut a était divisé en deux séances (48 min chaque). Chaque séance comprenait 320 essais. On a demandé au sujets d’imaginer un déplacement avant ou arrière dans le tunnel virtuel de façon aléatoire d’après une commande écrite sur l'écran. Les essais ont été menés avec feedback. Trois électrodes ont été montées sur le scalp, vis-à-vis du cortex moteur. Durant la 1re séance, la classification des deux taches (navigation avant et arrière) a été réalisée par les méthodes de puissance de bande, de représentation temporel-fréquence, des modèles autorégressifs et des rapports d’asymétrie du rythme β avec classificateurs d’analyse discriminante linéaire et SVM. Les seuils ont été calculés en temps différé pour former des signaux de contrôle qui ont été utilisés en temps réel durant la 2e séance afin d’initier, par les ondes cérébrales de l'utilisateur, le déplacement du tunnel virtuel dans le sens demandé. Après 96 min d'entrainement, la méthode « online biofeedback » de la puissance de bande a atteint une précision de classification moyenne de 76 %, et la classification en temps différé avec les rapports d’asymétrie et puissance de bande, a atteint une précision de classification d’environ 80 %.

Deep learning of representations and its application to computer vision

L’objectif de cette thèse par articles est de présenter modestement quelques étapes du parcours qui mènera (on espère) à une solution générale du problème de l’intelligence artificielle. Cette thèse contient quatre articles qui présentent chacun une différente nouvelle méthode d’inférence perceptive en utilisant l’apprentissage machine et, plus particulièrement, les réseaux neuronaux profonds. Chacun de ces documents met en évidence l’utilité de sa méthode proposée dans le cadre d’une tâche de vision par ordinateur. Ces méthodes sont applicables dans un contexte plus général, et dans certains cas elles on tété appliquées ailleurs, mais ceci ne sera pas abordé dans le contexte de cette de thèse. Dans le premier article, nous présentons deux nouveaux algorithmes d’inférence variationelle pour le modèle génératif d’images appelé codage parcimonieux “spike- and-slab” (CPSS). Ces méthodes d’inférence plus rapides nous permettent d’utiliser des modèles CPSS de tailles beaucoup plus grandes qu’auparavant. Nous démontrons qu’elles sont meilleures pour extraire des détecteur de caractéristiques quand très peu d’exemples étiquetés sont disponibles pour l’entraînement. Partant d’un modèle CPSS, nous construisons ensuite une architecture profonde, la machine de Boltzmann profonde partiellement dirigée (MBP-PD). Ce modèle a été conçu de manière à simplifier d’entraînement des machines de Boltzmann profondes qui nécessitent normalement une phase de pré-entraînement glouton pour chaque couche. Ce problème est réglé dans une certaine mesure, mais le coût d’inférence dans le nouveau modèle est relativement trop élevé pour permettre de l’utiliser de manière pratique. Dans le deuxième article, nous revenons au problème d’entraînement joint de machines de Boltzmann profondes. Cette fois, au lieu de changer de famille de modèles, nous introduisons un nouveau critère d’entraînement qui donne naissance aux machines de Boltzmann profondes à multiples prédictions (MBP-MP). Les MBP-MP sont entraînables en une seule étape et ont un meilleur taux de succès en classification que les MBP classiques. Elles s’entraînent aussi avec des méthodes variationelles standard au lieu de nécessiter un classificateur discriminant pour obtenir un bon taux de succès en classification. Par contre, un des inconvénients de tels modèles est leur incapacité de générer deséchantillons, mais ceci n’est pas trop grave puisque la performance de classification des machines de Boltzmann profondes n’est plus une priorité étant donné les dernières avancées en apprentissage supervisé. Malgré cela, les MBP-MP demeurent intéressantes parce qu’elles sont capable d’accomplir certaines tâches que des modèles purement supervisés ne peuvent pas faire, telles que celle de classifier des données incomplètes ou encore celle de combler intelligemment l’information manquante dans ces données incomplètes. Le travail présenté dans cette thèse s’est déroulé au milieu d’une période de transformations importantes du domaine de l’apprentissage à réseaux neuronaux profonds qui a été déclenchée par la découverte de l’algorithme de “dropout” par Geoffrey Hinton. Dropout rend possible un entraînement purement supervisé d’architectures de propagation unidirectionnel sans être exposé au danger de sur- entraînement. Le troisième article présenté dans cette thèse introduit une nouvelle fonction d’activation spécialement con ̧cue pour aller avec l’algorithme de Dropout. Cette fonction d’activation, appelée maxout, permet l’utilisation de aggrégation multi-canal dans un contexte d’apprentissage purement supervisé. Nous démontrons comment plusieurs tâches de reconnaissance d’objets sont mieux accomplies par l’utilisation de maxout. Pour terminer, sont présentons un vrai cas d’utilisation dans l’industrie pour la transcription d’adresses de maisons à plusieurs chiffres. En combinant maxout avec une nouvelle sorte de couche de sortie pour des réseaux neuronaux de convolution, nous démontrons qu’il est possible d’atteindre un taux de succès comparable à celui des humains sur un ensemble de données coriace constitué de photos prises par les voitures de Google. Ce système a été déployé avec succès chez Google pour lire environ cent million d’adresses de maisons.