Desenvolvimento de um sistema multidimensional para classificação e gestão de informação em saúde: aplicabilidade à informação clínica

Introdução: A profusão de informação na área médica cria problemas de gestão, sendo necessários métodos sistematizados para armazenamento e recuperação. Quando a informação se insere no contexto do processo clínico, os métodos devem integrar terminologias biomédicas controladas e igualmente devem integrar as características desejáveis dirigidas à estrutura, conteúdo e resultados clínicos. O objectivo deste artigo é testar a aplicabilidade e capacidade de recuperação, de um sistema multidimensional desenvolvido para classificação e gestão de informação em saúde. Métodos: A partir das questões recebidas em seis anos (Serviço de Informação de Medicamentos, Serviços Farmacêuticos, Hospitais da Universidade de Coimbra), seleccionaram-se 300 questões sobre informação clínica, por método aleatório informatizado. Caracterizou-se e avaliou-se a aplicabilidade pela quantidade classificada e pela necessidade de alterações ao sistema que é constituído por várias dimensões independentes e que englobam conceitos por vezes hierarquizados. A recuperação das questões foi testada pesquisando informação numa dimensão ou cruzamento de dimensões. Resultados: Todas as questões foram classificadas: 53% são casos clínicos com incidência nas doenças geniturinárias; doenças metabólicas, nutricionais e endócrinas; neoplasias; infecções e doenças do sistema nervoso. Em 81%, o objecto é o medicamento, sobretudo anti-infecciosos e anti-neoplásicos. As áreas de terapêutica e segurança foram as mais solicitadas, incidindo principalmente sobre os assuntos: utilização, reacções adversas, identificação de medicamentos e tecnologia farmacêutica. Na aplicabilidade, foi necessário adicionar alguns conceitos e modificar alguns grupos hierárquicos que não modificaram a estrutura base, nem colidiram com as características desejáveis. As limitações prenderam-se com os sistemas de classificação externos integrados. A pesquisa na dimensão assunto, do conceito administração de medicamentos, recuperou 19 questões. O cruzamento de duas dimensões: anti-infecciosos (externa) e teratogenicidade (assunto), recuperou três questões. Nos dois exemplos recupera-se informação a partir de qualquer um dos níveis da hierarquia, do mais geral ao mais específico e mesmo a partir de dimensões externas. Conclusões: A utilização do sistema nesta amostra demonstrou aplicabilidade na classificação e arquivo de informação clínica, capacidade de recuperação e flexibilidade, sofrendo alterações sem interferir com as características desejáveis. Esta ferramenta permite a recuperação da evidência que interessa orientada para o doente. Introduction: The large amount of information in the medical area creates management problems, being necessary systematic methods for filing and retrieval. With information on the context of clinical records, methods must integrate controlled biomedical terminologies and desirable characteristics oriented to the structure, content and clinical results. The objective is to test the applicability and capacity for retrieval of a multidimensional system developed for classification and management of health information. Methods: Three hundred questions were randomly selected, by computerized method, from the questions received in six years (Medicine Information Service, Pharmaceutical Department, Coimbra University Hospitals). They were characterized and applicability evaluated by classified amount and need to alter the system, which is composed of various independent dimensions, incorporating concepts sometimes hierarchical. Questions retrieval was tested searching information in a dimension or between dimensions. Results: All questions were classified: 53% are clinical cases with illnesses incidence in the genitourinary system; metabolic, nutritional and endocrine disease; cancer; infections and nervous system. In 81%, the object is a drug, mostly anti-infectious and anti-neoplastic agents. The therapeutic and safety areas had been the most requested, regarding the subjects: use, adverse reactions, drug identification and pharmaceutical technology. As to applicability, it was necessary to add some concepts and modify same hierarchical groups, that didn’t modify the basic structure, nor had collided with the desirable characteristics. The limitations were related with the incorporated external classification systems. The search in the subject dimension of the concept drug administration retrieved 19 questions. The search between two dimensions: antiinfectious (external) and teratogenicity (subject) retrieved three questions. In the two examples, it was possible to retrieve information from any one of the levels of the hierarchy, from the most general to the most specific and even from external dimensions. Conclusions: The use of the system in this sample showed its applicability in clinical information classification and filing, retrieval capacity and flexibility, supporting modifications without interfering with desirable characteristics. This tool allows retrieval of patient-oriented evidence that matters.

Data mining techniques to support the classification of MV electricity customers

This paper describes a methodology that was developed for the classification of Medium Voltage (MV) electricity customers. Starting from a sample of data bases, resulting from a monitoring campaign, Data Mining (DM) techniques are used in order to discover a set of a MV consumer typical load profile and, therefore, to extract knowledge regarding to the electric energy consumption patterns. In first stage, it was applied several hierarchical clustering algorithms and compared the clustering performance among them using adequacy measures. In second stage, a classification model was developed in order to allow classifying new consumers in one of the obtained clusters that had resulted from the previously process. Finally, the interpretation of the discovered knowledge are presented and discussed.

Demand response programs definition supported by clustering and classification techniques

The growing importance and influence of new resources connected to the power systems has caused many changes in their operation. Environmental policies and several well know advantages have been made renewable based energy resources largely disseminated. These resources, including Distributed Generation (DG), are being connected to lower voltage levels where Demand Response (DR) must be considered too. These changes increase the complexity of the system operation due to both new operational constraints and amounts of data to be processed. Virtual Power Players (VPP) are entities able to manage these resources. Addressing these issues, this paper proposes a methodology to support VPP actions when these act as a Curtailment Service Provider (CSP) that provides DR capacity to a DR program declared by the Independent System Operator (ISO) or by the VPP itself. The amount of DR capacity that the CSP can assure is determined using data mining techniques applied to a database which is obtained for a large set of operation scenarios. The paper includes a case study based on 27,000 scenarios considering a diversity of distributed resources in a 33 bus distribution network.

A hybrid recommendation approach for a tourism system

Many current e-commerce systems provide personalization when their content is shown to users. In this sense, recommender systems make personalized suggestions and provide information of items available in the system. Nowadays, there is a vast amount of methods, including data mining techniques that can be employed for personalization in recommender systems. However, these methods are still quite vulnerable to some limitations and shortcomings related to recommender environment. In order to deal with some of them, in this work we implement a recommendation methodology in a recommender system for tourism, where classification based on association is applied. Classification based on association methods, also named associative classification methods, consist of an alternative data mining technique, which combines concepts from classification and association in order to allow association rules to be employed in a prediction context. The proposed methodology was evaluated in some case studies, where we could verify that it is able to shorten limitations presented in recommender systems and to enhance recommendation quality.

On the development of an automatic voice pleasantness classification and intensity estimation system

In the last few years, the number of systems and devices that use voice based interaction has grown significantly. For a continued use of these systems, the interface must be reliable and pleasant in order to provide an optimal user experience. However there are currently very few studies that try to evaluate how pleasant is a voice from a perceptual point of view when the final application is a speech based interface. In this paper we present an objective definition for voice pleasantness based on the composition of a representative feature subset and a new automatic voice pleasantness classification and intensity estimation system. Our study is based on a database composed by European Portuguese female voices but the methodology can be extended to male voices or to other languages. In the objective performance evaluation the system achieved a 9.1% error rate for voice pleasantness classification and a 15.7% error rate for voice pleasantness intensity estimation.

EMMON: a WSN system architecture for large scale and dense real-time embedded monitoring

Wireless sensor networks (WSNs) have attracted growing interest in the last decade as an infrastructure to support a diversity of ubiquitous computing and cyber-physical systems. However, most research work has focused on protocols or on specific applications. As a result, there remains a clear lack of effective, feasible and usable system architectures that address both functional and non-functional requirements in an integrated fashion. In this paper, we outline the EMMON system architecture for large-scale, dense, real-time embedded monitoring. EMMON provides a hierarchical communication architecture together with integrated middleware and command and control software. It has been designed to use standard commercially-available technologies, while maintaining as much flexibility as possible to meet specific applications requirements. The EMMON architecture has been validated through extensive simulation and experimental evaluation, including a 300+ node test-bed, which is, to the best of our knowledge, the largest single-site WSN test-bed in Europe to date.

A Clinical Support System Based on Quality of Life Estimation

Quality of life is a concept influenced by social, economic, psychological, spiritual or medical state factors. More specifically, the perceived quality of an individual's daily life is an assessment of their well-being or lack of it. In this context, information technologies may help on the management of services for healthcare of chronic patients such as estimating the patient quality of life and helping the medical staff to take appropriate measures to increase each patient quality of life. This paper describes a Quality of Life estimation system developed using information technologies and the application of data mining algorithms to access the information of clinical data of patients with cancer from Otorhinolaryngology and Head and Neck services of an oncology institution. The system was evaluated with a sample composed of 3013 patients. The results achieved show that there are variables that may be significant predictors for the Quality of Life of the patient: years of smoking (p value 0.049) and size of the tumor (p value < 0.001). In order to assign the variables to the classification of the quality of life the best accuracy was obtained by applying the John Platt's sequential minimal optimization algorithm for training a support vector classifier. In conclusion data mining techniques allow having access to patients additional information helping the physicians to be able to know the quality of life and produce a well-informed clinical decision.

Automatic Mallampati Classification Using Active Appearance Models

Difficult tracheal intubation assessment is an important research topic in anesthesia as failed intubations are important causes of mortality in anesthetic practice. The modified Mallampati score is widely used, alone or in conjunction with other criteria, to predict the difficulty of intubation. This work presents an automatic method to assess the modified Mallampati score from an image of a patient with the mouth wide open. For this purpose we propose an active appearance models (AAM) based method and use linear support vector machines (SVM) to select a subset of relevant features obtained using the AAM. This feature selection step proves to be essential as it improves drastically the performance of classification, which is obtained using SVM with RBF kernel and majority voting. We test our method on images of 100 patients undergoing elective surgery and achieve 97.9% accuracy in the leave-one-out crossvalidation test and provide a key element to an automatic difficult intubation assessment system.

Real-Time Automatic Object Classification and Tracking using Genetic Programming and NVIDIA R CUDA TM

Genetic Programming (GP) is a widely used methodology for solving various computational problems. GP's problem solving ability is usually hindered by its long execution times. In this thesis, GP is applied toward real-time computer vision. In particular, object classification and tracking using a parallel GP system is discussed. First, a study of suitable GP languages for object classification is presented. Two main GP approaches for visual pattern classification, namely the block-classifiers and the pixel-classifiers, were studied. Results showed that the pixel-classifiers generally performed better. Using these results, a suitable language was selected for the real-time implementation. Synthetic video data was used in the experiments. The goal of the experiments was to evolve a unique classifier for each texture pattern that existed in the video. The experiments revealed that the system was capable of correctly tracking the textures in the video. The performance of the system was on-par with real-time requirements.

Feature Selection and Classification Using Age Layered Population Structure Genetic Programming

The curse of dimensionality is a major problem in the fields of machine learning, data mining and knowledge discovery. Exhaustive search for the most optimal subset of relevant features from a high dimensional dataset is NP hard. Sub–optimal population based stochastic algorithms such as GP and GA are good choices for searching through large search spaces, and are usually more feasible than exhaustive and deterministic search algorithms. On the other hand, population based stochastic algorithms often suffer from premature convergence on mediocre sub–optimal solutions. The Age Layered Population Structure (ALPS) is a novel metaheuristic for overcoming the problem of premature convergence in evolutionary algorithms, and for improving search in the fitness landscape. The ALPS paradigm uses an age–measure to control breeding and competition between individuals in the population. This thesis uses a modification of the ALPS GP strategy called Feature Selection ALPS (FSALPS) for feature subset selection and classification of varied supervised learning tasks. FSALPS uses a novel frequency count system to rank features in the GP population based on evolved feature frequencies. The ranked features are translated into probabilities, which are used to control evolutionary processes such as terminal–symbol selection for the construction of GP trees/sub-trees. The FSALPS metaheuristic continuously refines the feature subset selection process whiles simultaneously evolving efficient classifiers through a non–converging evolutionary process that favors selection of features with high discrimination of class labels. We investigated and compared the performance of canonical GP, ALPS and FSALPS on high–dimensional benchmark classification datasets, including a hyperspectral image. Using Tukey’s HSD ANOVA test at a 95% confidence interval, ALPS and FSALPS dominated canonical GP in evolving smaller but efficient trees with less bloat expressions. FSALPS significantly outperformed canonical GP and ALPS and some reported feature selection strategies in related literature on dimensionality reduction.

Feature Selection and Classification Using Age Layered Population Structure Genetic Programming

The curse of dimensionality is a major problem in the fields of machine learning, data mining and knowledge discovery. Exhaustive search for the most optimal subset of relevant features from a high dimensional dataset is NP hard. Sub–optimal population based stochastic algorithms such as GP and GA are good choices for searching through large search spaces, and are usually more feasible than exhaustive and determinis- tic search algorithms. On the other hand, population based stochastic algorithms often suffer from premature convergence on mediocre sub–optimal solutions. The Age Layered Population Structure (ALPS) is a novel meta–heuristic for overcoming the problem of premature convergence in evolutionary algorithms, and for improving search in the fitness landscape. The ALPS paradigm uses an age–measure to control breeding and competition between individuals in the population. This thesis uses a modification of the ALPS GP strategy called Feature Selection ALPS (FSALPS) for feature subset selection and classification of varied supervised learning tasks. FSALPS uses a novel frequency count system to rank features in the GP population based on evolved feature frequencies. The ranked features are translated into probabilities, which are used to control evolutionary processes such as terminal–symbol selection for the construction of GP trees/sub-trees. The FSALPS meta–heuristic continuously refines the feature subset selection process whiles simultaneously evolving efficient classifiers through a non–converging evolutionary process that favors selection of features with high discrimination of class labels. We investigated and compared the performance of canonical GP, ALPS and FSALPS on high–dimensional benchmark classification datasets, including a hyperspectral image. Using Tukey’s HSD ANOVA test at a 95% confidence interval, ALPS and FSALPS dominated canonical GP in evolving smaller but efficient trees with less bloat expressions. FSALPS significantly outperformed canonical GP and ALPS and some reported feature selection strategies in related literature on dimensionality reduction.

Modeling short-range ballistic missile defense and Israel's Iron Dome system

This paper develops a model of short-range ballistic missile defense and uses it to study the performance of Israel’s Iron Dome system. The deterministic base model allows for inaccurate missiles, unsuccessful interceptions, and civil defense. Model enhancements consider the trade-offs in attacking the interception system, the difficulties faced by militants in assembling large salvos, and the effects of imperfect missile classification by the defender. A stochastic model is also developed. Analysis shows that system performance can be highly sensitive to the missile salvo size, and that systems with higher interception rates are more “fragile” when overloaded. The model is calibrated using publically available data about Iron Dome’s use during Operation Pillar of Defense in November 2012. If the systems performed as claimed, they saved Israel an estimated 1778 casualties and $80 million in property damage, and thereby made preemptive strikes on Gaza about 8 times less valuable to Israel. Gaza militants could have inflicted far more damage by grouping their rockets into large salvos, but this may have been difficult given Israel’s suppression efforts. Counter-battery fire by the militants is unlikely to be worthwhile unless they can obtain much more accurate missiles.

Classification analytique de systèmes différentiels linéaires déployant une singularité irrégulière de rang de Poincaré 1

Cette thèse traite de la classification analytique du déploiement de systèmes différentiels linéaires ayant une singularité irrégulière. Elle est composée de deux articles sur le sujet: le premier présente des résultats obtenus lors de l'étude de la confluence de l'équation hypergéométrique et peut être considéré comme un cas particulier du second; le deuxième contient les théorèmes et résultats principaux. Dans les deux articles, nous considérons la confluence de deux points singuliers réguliers en un point singulier irrégulier et nous étudions les conséquences de la divergence des solutions au point singulier irrégulier sur le comportement des solutions du système déployé. Pour ce faire, nous recouvrons un voisinage de l'origine (de manière ramifiée) dans l'espace du paramètre de déploiement $\epsilon$. La monodromie d'une base de solutions bien choisie est directement reliée aux matrices de Stokes déployées. Ces dernières donnent une interprétation géométrique aux matrices de Stokes, incluant le lien (existant au moins pour les cas génériques) entre la divergence des solutions à $\epsilon=0$ et la présence de solutions logarithmiques autour des points singuliers réguliers lors de la résonance. La monodromie d'intégrales premières de systèmes de Riccati correspondants est aussi interprétée en fonction des éléments des matrices de Stokes déployées. De plus, dans le second article, nous donnons le système complet d'invariants analytiques pour le déploiement de systèmes différentiels linéaires $x^2y'=A(x)y$ ayant une singularité irrégulière de rang de Poincaré $1$ à l'origine au-dessus d'un voisinage fixé $\mathbb{D}_r$ dans la variable $x$. Ce système est constitué d'une partie formelle, donnée par des polynômes, et d'une partie analytique, donnée par une classe d'équivalence de matrices de Stokes déployées. Pour chaque valeur du paramètre $\epsilon$ dans un secteur pointé à l'origine d'ouverture plus grande que $2\pi$, nous recouvrons l'espace de la variable, $\mathbb{D}_r$, avec deux secteurs et, au-dessus de chacun, nous choisissons une base de solutions du système déployé. Cette base sert à définir les matrices de Stokes déployées. Finalement, nous prouvons un théorème de réalisation des invariants qui satisfont une condition nécessaire et suffisante, identifiant ainsi l'ensemble des modules.

Characterization of the unfolding of a weak focus and modulus of analytic classification

La thèse présente une description géométrique d’un germe de famille générique déployant un champ de vecteurs réel analytique avec un foyer faible à l’origine et son complexifié : le feuilletage holomorphe singulier associé. On montre que deux germes de telles familles sont orbitalement analytiquement équivalents si et seulement si les germes de familles de difféomorphismes déployant la complexification de leurs fonctions de retour de Poincaré sont conjuguées par une conjugaison analytique réelle. Le “caractère réel” de la famille correspond à sa Z2-équivariance dans R^4, et cela s’exprime comme l’invariance du plan réel sous le flot du système laquelle, à son tour, entraîne que l’expansion asymptotique de la fonction de Poincaré est réelle quand le paramètre est réel. Le pullback du plan réel après éclatement par la projection monoidal standard intersecte le feuilletage en une bande de Möbius réelle. La technique d’éclatement des singularités permet aussi de donner une réponse à la question de la “réalisation” d’un germe de famille déployant un germe de difféomorphisme avec un point fixe de multiplicateur égal à −1 et de codimension un comme application de semi-monodromie d’une famille générique déployant un foyer faible d’ordre un. Afin d’étudier l’espace des orbites de l’application de Poincaré, nous utilisons le point de vue de Glutsyuk, puisque la dynamique est linéarisable auprès des points singuliers : pour les valeurs réels du paramètre, notre démarche, classique, utilise une méthode géométrique, soit un changement de coordonée (coordonée “déroulante”) dans lequel la dynamique devient beaucoup plus simple. Mais le prix à payer est que la géométrie locale du plan complexe ambiante devient une surface de Riemann, sur laquelle deux notions de translation sont définies. Après avoir pris le quotient par le relèvement de la dynamique nous obtenons l’espace des orbites, ce qui s’avère être l’union de trois tores complexes plus les points singuliers (l’espace résultant est non-Hausdorff). Les translations, le caractère réel de l’application de Poincaré et le fait que cette application est un carré relient les différentes composantes du “module de Glutsyuk”. Cette propriété implique donc le fait qu’une seule composante de l’invariant Glutsyuk est indépendante.

Annotation syntaxico-sémantique des actants en corpus spécialisé

L’annotation en rôles sémantiques est une tâche qui permet d’attribuer des étiquettes de rôles telles que Agent, Patient, Instrument, Lieu, Destination etc. aux différents participants actants ou circonstants (arguments ou adjoints) d’une lexie prédicative. Cette tâche nécessite des ressources lexicales riches ou des corpus importants contenant des phrases annotées manuellement par des linguistes sur lesquels peuvent s’appuyer certaines approches d’automatisation (statistiques ou apprentissage machine). Les travaux antérieurs dans ce domaine ont porté essentiellement sur la langue anglaise qui dispose de ressources riches, telles que PropBank, VerbNet et FrameNet, qui ont servi à alimenter les systèmes d’annotation automatisés. L’annotation dans d’autres langues, pour lesquelles on ne dispose pas d’un corpus annoté manuellement, repose souvent sur le FrameNet anglais. Une ressource telle que FrameNet de l’anglais est plus que nécessaire pour les systèmes d’annotation automatisé et l’annotation manuelle de milliers de phrases par des linguistes est une tâche fastidieuse et exigeante en temps. Nous avons proposé dans cette thèse un système automatique pour aider les linguistes dans cette tâche qui pourraient alors se limiter à la validation des annotations proposées par le système. Dans notre travail, nous ne considérons que les verbes qui sont plus susceptibles que les noms d’être accompagnés par des actants réalisés dans les phrases. Ces verbes concernent les termes de spécialité d’informatique et d’Internet (ex. accéder, configurer, naviguer, télécharger) dont la structure actancielle est enrichie manuellement par des rôles sémantiques. La structure actancielle des lexies verbales est décrite selon les principes de la Lexicologie Explicative et Combinatoire, LEC de Mel’čuk et fait appel partiellement (en ce qui concerne les rôles sémantiques) à la notion de Frame Element tel que décrit dans la théorie Frame Semantics (FS) de Fillmore. Ces deux théories ont ceci de commun qu’elles mènent toutes les deux à la construction de dictionnaires différents de ceux issus des approches traditionnelles. Les lexies verbales d’informatique et d’Internet qui ont été annotées manuellement dans plusieurs contextes constituent notre corpus spécialisé. Notre système qui attribue automatiquement des rôles sémantiques aux actants est basé sur des règles ou classificateurs entraînés sur plus de 2300 contextes. Nous sommes limités à une liste de rôles restreinte car certains rôles dans notre corpus n’ont pas assez d’exemples annotés manuellement. Dans notre système, nous n’avons traité que les rôles Patient, Agent et Destination dont le nombre d’exemple est supérieur à 300. Nous avons crée une classe que nous avons nommé Autre où nous avons rassemblé les autres rôles dont le nombre d’exemples annotés est inférieur à 100. Nous avons subdivisé la tâche d’annotation en sous-tâches : identifier les participants actants et circonstants et attribuer des rôles sémantiques uniquement aux actants qui contribuent au sens de la lexie verbale. Nous avons soumis les phrases de notre corpus à l’analyseur syntaxique Syntex afin d’extraire les informations syntaxiques qui décrivent les différents participants d’une lexie verbale dans une phrase. Ces informations ont servi de traits (features) dans notre modèle d’apprentissage. Nous avons proposé deux techniques pour l’identification des participants : une technique à base de règles où nous avons extrait une trentaine de règles et une autre technique basée sur l’apprentissage machine. Ces mêmes techniques ont été utilisées pour la tâche de distinguer les actants des circonstants. Nous avons proposé pour la tâche d’attribuer des rôles sémantiques aux actants, une méthode de partitionnement (clustering) semi supervisé des instances que nous avons comparée à la méthode de classification de rôles sémantiques. Nous avons utilisé CHAMÉLÉON, un algorithme hiérarchique ascendant.