Primary headache disorders are the most common complaints worldwide. The socioeconomic and personal impact of headache disorders is enormous, as it is the leading cause of workplace absence. Headache patients’ consultations are increasing as the population has increased in size, live longer and many people have multiple conditions, however, access to specialist services across the UK is currently inequitable because the numbers of trained consultant neurologists in the UK are 10 times lower than other European countries. Additionally, more than two third of headache cases presented to primary care were labelled with unspecified headache. Therefore, an alternative pathway to diagnose and manage patients with primary headache could be crucial to reducing the need for specialist assessment and increase capacity within the current service model. Several recent studies have targeted this issue through the development of clinical decision support systems, which can help non-specialist doctors and general practitioners to diagnose patients with primary headache disorders in primary clinics. However, the majority of these studies were following a rule-based system style, in which the rules were summarised and expressed by a computer engineer. This style carries many downsides, and we will discuss them later on in this dissertation. In this study, we are adopting a completely different approach. The use of machine learning is recruited for the classification of primary headache disorders, for which a dataset of 832 records of patients with primary headaches was considered, originating from three medical centres located in Turkey. Three main types of primary headaches were derived from the data set including Tension Type Headache in both episodic and chronic forms, Migraine with and without Aura, followed by Trigeminal Autonomic Cephalalgia that further subdivided into Cluster headache, paroxysmal hemicrania and short-lasting unilateral neuralgiform headache attacks with conjunctival injection and tearing. Six popular machine-learning based classifiers, including linear and non-linear ensemble learning, in addition to one regression based procedure, have been evaluated for the classification of primary headaches within a supervised learning setting, achieving highest aggregate performance outcomes of AUC 0.923, sensitivity 0.897, and overall classification accuracy of 0.843. This study also introduces the proposed HydroApp system, which is an M-health based personalised application for the follow-up of patients with long-term conditions such as chronic headache and hydrocephalus. We managed to develop this system with the supervision of headache specialists at Ashford hospital, London, and neurology experts at Walton Centre and Alder Hey hospital Liverpool. We have successfully investigated the acceptance of using such an M-health based system via an online questionnaire, where 86% of paediatric patients and 60% of adult patients were interested in using HydroApp system to manage their conditions. Features and functions offered by HydroApp system such as recording headache score, recording of general health and well-being as well as alerting the treating team, have been perceived as very or extremely important aspects from patients’ point of view. The study concludes that the advances in intelligent systems and M-health applications represent a promising atmosphere through which to identify alternative solutions, which in turn increases the capacity in the current service model and improves diagnostic capability in the primary headache domain and beyond.

The area of computer-aided diagnosis (CAD) has undergone tremendous growth in recent years. In CAD, the computer output is used as a second opinion for cancer diagnosis. Development of cancer is a multiphase process and mutation of genes is involved over the years. Cancer grows out of normal cells in the body and it usually occurs when growth of the cells in the body is out of control. This phenomenon changes the shape and structure of the tissue glands. In this thesis, we have developed three algorithms for classification of colon and prostate biopsy samples. First, we computed morphological and shape based parameters from hyperspectral images of colon samples and used linear and non-linear classifiers for the identification of cancerous regions. To investigate the importance of hyperspectral imagery in histopathology, we selected a single spectral band from its hyperspectral cube and performed an analysis based on texture of the images. Texture refers to an arrangement of the basic constituents of the material and it is represented by the interrelationships between the spatial arrangements of the image pixels. A novel feature selection method based on the quality of clustering is developed to discard redundant information. In the third algorithm, we used Bayesian inference for segmentation of glands in colon and prostate biopsy samples. In this approach, glands in a tissue are represented by polygonal models with variuos number of vertices depending on the size of glands. An appropriate set of proposals for Metropolis- Hastings-Green algorithm is formulated and a series of Markov chain Monte Carlo (MCMC) simulations are run to extract polygonal models for the glands. We demonstrate the performance of 3D spectral and spatial and 2D spatial analyses with over 90% classification accuracies and less than 10% average segmentation error for the polygonal models.

This research addressed a major and widespread problem in the NHS: information for operational planning. The approach has been to analyse the needs of management in this field in a particular health district and to develop a system in response to these needs of use in the particular District and generally in the NHS. The emphasis had been on the identification and quantification of relationships between elements of the District important to corporate planning, so that the feasibility and effects of planning choices can be assessed. Particular attention has been paid to the attitudes, values and concerns of senior managers and doctors in the NHS throughout the project. Research began in October 1979 on the development of a database and model of the former North East District of KCWAHA. This health district contained a large undergraduate teaching hospital (The Middlesex), two large psychiatric hospitals, and several smaller specialist hospitals. The purpose of the decision-support system is to enable management to explore rapidly the implications of operational planning optlons over several years. It does not recommend which option should be followed. Plans are tested in terms of the bed capacity of general wards and specialist units. The revenue costs of an option are estimated using a detailed analysis of which types of change cause which types of cost to vary within the District. The model then assesses the non-financial consequences for the operating theatres and service departments, and the effects on nurse training. The research has shown that it is feasible to build and maintain such a model and database with very limited clerical support. The output from the system has been found useful by management. This development has generated considerable support for further research. The assumptions of the model and the procedures for updating the database are fully documented. Procedures for implementing the system in another health district are also available. The model runs on the computer at Imperial College, University of London. The research programme continues with the extension of the model to cover the whole of the new Bloomsbury District (including University College Hospital), where the use of the system has had a substantial impact on decisionmaking at the most senior level.

This thesis describes the research and development of a decision support tool for determining a medical patient's suitability for surgical anaesthesia. At present, there is a change in the way that patients are clinically assessedp rior to surgery. The pre-operative assessment, usually conducted by a qualified anaesthetist, is being more frequently performed by nursing grade staff. The pre-operative assessmenet xists to minimise the risk of surgical complications for the patient. Nursing grade staff are often not as experienced as qualified anaesthetists, and thus are not as well suited to the role of performing the pre-operative assessment. This research project used data collected during pre-operative assessments to develop a decision support tool that would assist the nurse (or anaesthetist) in determining whether a patient is suitable for surgical anaesthesia. The three main objectives are: firstly, to research and develop an automated intelligent systems technique for classifying heart and lung sounds and hence identifying cardio-respiratory pathology. Secondly, to research and develop an automated intelligent systems technique for assessing the patient's blood oxygen level and pulse waveform. Finally, to develop a decision support tool that would combine the assessmentsa bove in forming a decision as to whether the patient is suitable for surgical anaesthesia. Clinical data were collected from hospital outpatient departments and recorded alongside the diagnoses made by a qualified anaesthetist. Heart and lung sounds were collected using an electronic stethoscope. Using this data two ensembles of artificial neural networks were trained to classify the different heart and lung sounds into different pathology groups. Classification accuracies up to 99.77% for the heart sounds, and 100% for the lung sounds has been obtained. Oxygen saturation and pulse waveform measurements were recorded using a pulse oximeter. Using this data an artificial neural network was trained to discriminate between normal and abnormal pulse waveforms. A discrimination accuracy of 98% has been obtained from the system. A fuzzy inference system was generated to classify the patient's blood oxygen level as being either an inhibiting or non-inhibiting factor in their suitability for surgical anaesthesia. When tested the system successfully classified 100% of the test dataset. A decision support tool, applying the genetic programming evolutionary technique to a fuzzy classification system was created. The decision support tool combined the results from the heart sound, lung sound and pulse oximetry classifiers in determining whether a patient was suitable for surgical anaesthesia. The evolved fuzzy system attained a classification accuracy of 91.79%. The principal conclusion from this thesis is that intelligent systems, such as artificial neural networks, genetic programming, and fuzzy inference systems, can be successfully applied to the creation of medical decision support tools.

No description available.

Open data are playing a vital role in different communities, including governments, businesses, and education. This revolution has had a high impact on the education field. Recently, new practices are being adopted for publishing and connecting data on the web, known as "Linked Data", and these are used to expose and connect data which were not previously linked. In the context of education, applying Linked Data practices to the growing amount of open data used for learning is potentially highly beneficial. The work presented in this thesis tackles the challenges of data acquisition and integration from distributed web data sources into one linked dataset. The application domain of this thesis is medical education, and the focus is on bridging the gap between articles published in online educational libraries and content published on Web 2.0 platforms that can be used for education. The integration of a collection of heterogeneous resources is to create links between data collected from distributed web data sources. To address these challenges, a system is proposed that exploits the Linked Data for building a metadata schema in XML/RDF format for describing resources and enriching it with external dataset that adds semantic to its metadata. The proposed system collects resources from distributed data sources on the web and enriches their metadata with concepts from biomedical ontologies, such as SNOMED CT, that enable its linking. The final result of building this system is a linked dataset of more than 10,000 resources collected from PubMed Library, YouTube channels, and Blogging platforms. The effectiveness of the system proposed is evaluated by validating the content of the linked dataset when accessed and retrieved. Ontology-based techniques have been developed for browsing and querying the linked dataset resulting from the system proposed. Experiments have been conducted to simulate users' access to the linked dataset and validate its content. The results were promising and have shown the effectiveness of using SNOMED CT for integrating distributed resources from diverse web data sources.

In the post-harvest area, a common approach to quantify the average color of fruits peel over time is the sampling of small number of points generally on its equatorial region using a colorimeter. However, when we use a colorimeter to classify an uneven-colored fruit misclassification may occur because points in the peel region may not be representative of average color of fruit. The main problem when we use this method is to determine the number of points to be sampled as well as the location of these points on the fruit\'s surface. An alternative method to evaluate measure of color is digital image analysis because it covers whole of the object surface, by using a sample of pixels taken from the image. As the colorimeter approach is faster and easier than image analysis, it may not be suitable for assessing the overall mean color of the papaya\'s peel and its performance will depend on the number of measured points and choice of sampled region. In this sense, the comparison between these approach is still necessary because we need to know if a sample on the equatorial region can reproduce a sample over the whole region, and if the colorimeter can compete with a scanner or digital camera in measuring the mean hue of papaya peel over time. Thus, we proposed a longitudinal concordance correlation (LCC) based on polynomial mixed-effects regression model to evaluate the extent of agreement among methods. The results show that ideally image analysis of whole fruit\'s region should be used to compute the mean hue and that the topography and curved surface of papaya fruit did not affect the mean hue obtained by the scanner. Since there are still no packages available to estimate the LCC in the free software environment R, we are developing a package called lcc, which provides functions for estimating the longitudinal concordance correlation (LCC) among methods based on variance components and fixed effects of polynomial mixed-effects model. Additionally, we implemented arguments in this function to estimating the longitudinal Pearson correlation (LPC), as precision measure, and longitudinal bias corrector factor (LA), as accuracy measure. Moreover, these components can be estimated using different structures for variance- covariance matrices of random effects and variance functions to model heteroscedasticity among within-group errors using or not the time as variance covariate. / No setor de pós-colheita é muito comum a utilização de colorímetros para avaliar a cor média da casca de frutos ao longo do tempo. No entanto, muitas vezes as técnicas de amostragem utilizando esse equipamento podem levar a medidas tendenciosas da média amostral. Alternativamente, a utilização de imagens digitais pode levar a um menor viés, uma vez que toda a região da casca do fruto é amostrada de forma sistemática. No entanto, ainda é necessária a comparação de ambas abordagens, pois o colorímetro tem vantagens em relação a facilidade de utilização e menor tempo para realizar a amostragem em cada fruto quando comparado a um scanner de mesa. Assim, no caso de variáveis respostas medidas em uma escala contínua, a reprodutibilidade das medidas tomadas por ambos equipamentos pode ser avaliada por meio do coeficiente de correlação de concordância. Dessa forma, para avaliar o perfil da concordância entre métodos, nós propomos uma correlação de concordância longitudinal (LCC), baseada em um modelo de regressão polinomial com efeitos mistos. Os resultados sugeriram que as técnicas por meio de imagens digitais devem ser utilizadas para a quantificação da tonalidade média de frutos. Adicionalmente, a partir do perfil de concordância estimado notamos que existe um período em que ambos os equipamentos podem ser utilizados. A performance do coeficiente de concordância longitudinal foi avaliada por meio de um estudo de simulação, o qual sugeriu que nossa metodologia é robusta a dados desbalanceados (\"dropout\") e que a probabilidade de convergência é aceitavel para uma amostra de 20 frutos e ideal para amostras a partir de 100 frutos. Uma vez que ainda não existem pacotes disponibilizados no ambiente computacional R para a estimação da correlação de concordância longitudinal, nós estamos desenvolvendo um pacote intitulado lcc, o qual será submetido ao \"Comprehensive R Archive Network\" (CRAN). Nesse pacote nós implementamos procedimentos para estimação da correlação de concordância longitudinal, da correlação de Person longitudinal e de uma medida de acurácia longitudinal. Além disso, nosso pacote foi desenvolvido para dados balanceados e desbalanceados, permitindo modelar a heteroscedasticidade entre erros dentro do grupo usando ou não o tempo como covariável, e, também, permitindo a inclusão de covariáveis no preditor linear para controlar variações sistemáticas na variável resposta.

The generalized additive models for location, scale and shape (GAMLSS) developed by Rigby and Stasinopoulos (2005) are a general class of univariate regression models that do not have the response distribution restricted to the exponential family as do the generalized linear and additive models, for example. In addition, they allow all the parameters of the response variable distribution to be modeled explicitly through different sets of explanatory variables. The semiparametric subclass of GAMLSS, in particular, accepts a wide range of parametric and nonparametric terms to be included in the predictors of the parameters. Similar to the generalized linear models, the GAMLSSs link predictors to parameters through monotonic link functions, which can also change for each parameter. This dissertation describes the GAMLSSs methodology and presents two applications to data sets provenient from experiments in agronomy; exploring methods of estimation, diagnosis and comparison of these models. / Os modelos lineares generalizados para locação, escala e forma (GAMLSS) desenvolvidos por Rigby e Stasinopoulos (2005) são uma ampla classe de modelos de regressão univariados que não pressupõem que a distribuição da variável resposta pertença à família exponencial como os modelos lineares generalizados ou aditivos generalizados, por exemplo. Além do mais, eles permitem que todos os parâmetros da distribuição da variável resposta sejam modelados explicitamente por meio de diferentes conjuntos de variáveis explanatórias. A subclasse semiparamétrica dos GAMLSS, em particular, permite que uma grande variedade de termos paramétricos e não paramétricos sejam incluídos nos preditores dos parâmetros da distribuição assumida para a variável resposta. De forma análoga aos modelos lineares generalizados, os GAMLSSs ligam os preditores aos parâmetros por meio de funções de ligação monótonas, que também podem mudar de acordo com o parâmetro a ser estimado. Esta dissertação descreve a metodologia dos modelos lineares generalizados para locação, escala e forma e apresenta duas aplicações a bancos de dados provenientes de experimentos agrícolas; explorando métodos de estimação, diagnóstico e comparação desse tipo de modelos.

O mamão (Carica papaya L.) cv. \"Sunrise Solo\" é um fruto que apresenta mudança gradual e desuniforme da cor da casca, que vai de verde para amarela. Isso faz com que a metodologia instrumental para avaliação da cor, por meio de um colorímetro, seja subjetiva, devido ao número de pontos observados, bem como às localizações deles no fruto. Como alternativa, foi proposta a utilização de imagens digitalizadas de toda região da casca do fruto, obtidas por meio de um scanner de mesa. Para avaliar a precisão desses métodos, foi conduzido um experimento com 20 repetições. Cada repetição era constituída de um fruto de mamoeiro cv. \"Sunrise Solo\", mantido sob temperatura e umidade relativa controladas. A cor da casca dos frutos foi avaliada, diariamente, utilizando um colorímetro e um scanner. Com o scanner, foram digitalizadas as duas faces do fruto e, com o colorímetro, foram observados quatro pontos equidistantes, na região equatorial do mesmo. Como a avaliação para cada fruto foi feita ao longo do tempo, os dados são classificados como longitudinais. Assim, utilizaram-se modelos lineares de efeitos mistos para estudar o comportamento da tonalidade média, pois essa técnica permite o uso de diferentes estruturas de variâncias e covariâncias para as matrizes dos efeitos aleatórios e dos erros. O processo de seleção do modelo foi realizado por meio do teste da razão de verossimilhanças e dos critérios de informação AIC e BIC, resultando no mesmo preditor linear e matrizes de variâncias e covariâncias para ambas as metodologias de quantificação da cor. O modelo final apresentou um preditor linear quadrático com efeitos aleatórios para o intercepto e para os termos linear e quadrático com matriz de variâncias e covariâncias dos efeitos aleatórios não estruturada e componentes de variância com heterocedasticidade para os erros. A utilização do scanner revelou dois grupos de maturação fisiológica distintos, que podem estar relacionados ao ponto de colheita do fruto, fato que não ficou evidente ao utilizar um colorímetro. De forma geral, o uso de um scanner possibilitou obter uma avaliação precisa da maturação do fruto, além de ser mais consistentes e eficiente do que o uso de um colorímetro para estudar a tonalidade média da casca de frutos que apresentam coloração desuniforme. / Papayas (Carica papaya L.) of \"Sunrise solo\" variety are fruits that present gradual and uneven changes in the peel color, which goes from green to yellow. As a consequence, when using a colorimeter to quantify their color, the results are subjective because of the number of observed points, as well as because of their position on the fruit. A proposed alternative was to use scanned images of the whole fruit peel to quantify color. To assess the precision of these methods, an experiment with 20 replicates was carried out. Each replicate consisted of a papaya fruit, kept under controlled temperature and relative humidity. The fruits\' peel colors were assessed, daily, using a colorimeter and a scanner. With the scanner, both sides of the fruit were scanned and, with the colorimeter, four equidistant points at the equatorial region of the fruit were observed. As the assessment was made through time for a same fruit, the data are classified as longitudinal. Therefore, linear mixed effect models were used to study the behavior of the average fruit color tonality through time, as this technique allows usage of different random effects and error covariance structures. Model selection was made using likelihood-ratio tests and the Akaike and Bayesian information criteria, which resulted in the selection of the same linear predictor and covariance matrices for both color quantification methods. The final model presented a quadractic linear predictor with random effects for the intercept, linear and quadractic terms with an unstructured variance-covariance matrix for the random effects and a variance components with heterogeneity matrix for the residuals. The use of a scanner revealed two distinct phisiological maturation groups, which may be related to the harvesting time. This was not observed when using a colorimeter. In general, using a scanner made possible to obtain more consistent observations, which makes it a more efficient methodology to study the average fruit peel color tonality.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / The determination of the causes and origins of a given disease is a complex undertaking, considering that there is a large number of genes engaged that interact with each other (Watson, 2006). Bioinformatics experts working in the search for a perfect integration between biology and information, in order to understand the likely factors that trigger certain diseases (Pevzner, 2000). To achieve this, the revolutionary methodology of Microarrays (LOCKHART et al., 1996) based on the gene expression of patients, it has been widely used to simultaneously measuring changes and regulation of the genes of the genome under certain biological conditions, resulting in a list of genes that may be considered interesting from a biological point of view for a particular disease. In this thesis, we present a multi-statistic method to detect differentially expressed genetic pathways in DNA microarray data. Many statistical methods of analysis are based on the use of a single statistical test. It is believed that the use of multiple tests decreases the number of false positive discoveries. Our method can be applied to transcriptome data to investigate which pathways have changes in expression when subjected to some type of disturbance. The method determines the activity of pathways evaluated, and verifies if the changes found are statistically significant through the bootstrap, Fisher exact and Wilcoxon tests. Implemented in R language and available for download from the Comprehensive R Archive Network (CRAN) as a package called PATHChange, our method showed consistency in its results with those predicted in the literature when tested for microarray of cancer and pre-cancer colon public data. The PATHChange method offers an alternative type of analysis of differentially expressed genes pathways for researchers seeking to determine phenotypes of diseases such as cancer. / A determinação das causas e origens de uma determinada doença é uma tarefa complexa, considerando que existe um grande número de genes comprometidos que interagem entre si (WATSON, 2006). Especialistas em Bioinformática trabalham na busca de uma perfeita integração entre a biologia e a informação, com o intuito de compreender os prováveis fatores que desencadeiam determinadas doenças (PEVZNER, 2000). Para tal, a metodologia revolucionária de Microarranjos (LOCKHART et al., 1996), baseada na expressão gênica de pacientes, tem sido amplamente utilizada para medir simultaneamente as mudanças e regulação dos genes do genoma sob certas condições biológicas, resultando em uma lista de genes que podem ser considerados interessantes do ponto de vista biológico para uma determinada doença. Na presente tese, nós apresentamos um método multiestatístico destinado à detectar vias genéticas diferencialmente expressas em dados de microarranjos de DNA. Grande parte dos métodos de análise estatística são baseados no uso de apenas um teste estatístico. Acredita-se que associar métodos estatísticos baseados em testes diferentes diminui o número de falsos positivos. O método que nós desenvolvemos determina a atividade das vias avaliadas, e verifica se as alterações encontradas são estatisticamente significativas através dos testes de bootstrap, exato de Fisher e Wilcoxon. Este método pode ser aplicado à dados de transcriptoma para investigar quais vias apresentam mudanças na expressão de seus genes quando submetidos à algum tipo de perturbação. Implementado em linguagem R e disponibilizado para download no CRAN (do inglês, Comprehensive R Archive Network) como um pacote denominado PATHChange, nosso método demonstrou consistência entre os seus resultados com os previstos na literatura quando testado para dados públicos de microarranjos de câncer e pré-câncer de cólon. O método do PATHChange oferece um tipo alternativo de análise de vias de genes diferencialmente expressas para os pesquisadores que buscam apurar fenótipos de doenças, tais como o câncer.