Modelo basado en técnicas de procesamiento de lenguaje natural para extraer y anotar información de publicaciones científicas

La rápida evolución experimentada en los últimos años por las tecnologías de Internet ha estimulado la proliferación de recursos software en varias disciplinas científicas, especialmente en bioinformática. En la mayoría de los casos, la tendencia actual es publicar dichos recursos como servicios accesibles libremente a través de Internet, utilizando tecnologías y patrones de diseño definidos para la implementación de Arquitecturas Orientadas a Servicios (SOA). La combinación simultánea de múltiples servicios dentro de un mismo flujo de trabajo abre la posibilidad de crear aplicaciones potencialmente más útiles y complejas. La integración de dichos servicios plantea grandes desafíos, tanto desde un punto de vista teórico como práctico, como por ejemplo, la localización y acceso a los recursos disponibles o la coordinación entre ellos. En esta tesis doctoral se aborda el problema de la identificación, localización, clasificación y acceso a los recursos informáticos disponibles en Internet. Con este fin, se ha definido un modelo genérico para la construcción de índices de recursos software con información extraída automáticamente de artículos de la literatura científica especializada en un área. Este modelo consta de seis fases que abarcan desde la selección de las fuentes de datos hasta el acceso a los índices creados, pasando por la identificación, extracción, clasificación y “curación” de la información relativa a los recursos. Para verificar la viabilidad, idoneidad y eficiencia del modelo propuesto, éste ha sido evaluado en dos dominios científicos diferentes—la BioInformática y la Informática Médica—dando lugar a dos índices de recursos denominados BioInformatics Resource Inventory (BIRI) y electronic-Medical Informatics Repository of Resources(e-MIR2) respectivamente. Los resultados obtenidos de estas aplicaciones son presentados a lo largo de la presente tesis doctoral y han dado lugar a varias publicaciones científicas en diferentes revistas JCR y congresos internacionales. El impacto potencial y la utilidad de esta tesis doctoral podrían resultar muy importantes teniendo en cuenta que, gracias a la generalidad del modelo propuesto, éste podría ser aplicado en cualquier disciplina científica. Algunas de las líneas de investigación futuras más relevantes derivadas de este trabajo son esbozadas al final en el último capítulo de este libro. ABSTRACT The rapid evolution experimented in the last years by the Internet technologies has stimulated the proliferation of heterogeneous software resources in most scientific disciplines, especially in the bioinformatics area. In most cases, current trends aim to publish those resources as services freely available over the Internet, using technologies and design patterns defined for the implementation of Service-Oriented Architectures (SOA). Simultaneous combination of various services into the same workflow opens the opportunity of creating more complex and useful applications. Integration of services raises great challenges, both from a theoretical to a practical point of view such as, for instance, the location and access to the available resources or the orchestration among them. This PhD thesis deals with the problem of identification, location, classification and access to informatics resources available over the Internet. On this regard, a general model has been defined for building indexes of software resources, with information extracted automatically from scientific articles from the literature specialized in the area. Such model consists of six phases ranging from the selection of data sources to the access to the indexes created, covering the identification, extraction, classification and curation of the information related to the software resources. To verify the viability, feasibility and efficiency of the proposed model, it has been evaluated in two different scientific domains—Bioinformatics and Medical Informatics—producing two resources indexes named BioInformatics Resources Inventory (BIRI) and electronic-Medical Informatics Repository of Resources (e-MIR2) respectively. The results and evaluation of those systems are presented along this PhD thesis, and they have produced different scientific publications in several JCR journals and international conferences. The potential impact and utility of this PhD thesis could be of great relevance considering that, thanks to the generality of the proposed model, it could be successfully extended to any scientific discipline. Some of the most relevant future research lines derived from this work are outlined at the end of this book.

Supporting patient screening to identify suitable clinical trials.

To support the efficient execution of post-genomic multi-centric clinical trials in breast cancer we propose a solution that streamlines the assessment of the eligibility of patients for available trials. The assessment of the eligibility of a patient for a trial requires evaluating whether each eligibility criterion is satisfied and is often a time consuming and manual task. The main focus in the literature has been on proposing different methods for modelling and formalizing the eligibility criteria. However the current adoption of these approaches in clinical care is limited. Less effort has been dedicated to the automatic matching of criteria to the patient data managed in clinical care. We address both aspects and propose a scalable, efficient and pragmatic patient screening solution enabling automatic evaluation of eligibility of patients for a relevant set of trials. This covers the flexible formalization of criteria and of other relevant trial metadata and the efficient management of these representations.

The information systems integration for a better caring and chronic patients´ empowerment

The technology innovation (TI) in the health system has been confirmed as the best format to share information and to facilitate the communication between all the actors involved in the chronic patients´ care.

Mixture modelling for cluster analysis

Cluster analysis via a finite mixture model approach is considered. With this approach to clustering, the data can be partitioned into a specified number of clusters g by first fitting a mixture model with g components. An outright clustering of the data is then obtained by assigning an observation to the component to which it has the highest estimated posterior probability of belonging; that is, the ith cluster consists of those observations assigned to the ith component (i = 1,..., g). The focus is on the use of mixtures of normal components for the cluster analysis of data that can be regarded as being continuous. But attention is also given to the case of mixed data, where the observations consist of both continuous and discrete variables.

Producing computer-generated tailored written information for stroke patients and their carers: System development and preliminary evaluation

Purpose: The aim of this project was to design and evaluate a system that would produce tailored information for stroke patients and their carers, customised according to their informational needs, and facilitate communication between the patient and, health professional. Method: A human factors development approach was used to develop a computer system, which dynamically compiles stroke education booklets for patients and carers. Patients and carers are able to select the topics about which they wish to receive information, the amount of information they want, and the font size of the printed booklet. The system is designed so that the health professional interacts with it, thereby providing opportunities for communication between the health professional and patient/carer at a number of points in time. Results: Preliminary evaluation of the system by health professionals, patients and carers was positive. A randomised controlled trial that examines the effect of the system on patient and carer outcomes is underway. (C) 2004 Elsevier Ireland Ltd. All rights reserved.

Estimating trends and seasonality in coronary heart disease

We present two methods of estimating the trend, seasonality and noise in time series of coronary heart disease events. In contrast to previous work we use a non-linear trend, allow multiple seasonal components, and carefully examine the residuals from the fitted model. We show the importance of estimating these three aspects of the observed data to aid insight of the underlying process, although our major focus is on the seasonal components. For one method we allow the seasonal effects to vary over time and show how this helps the understanding of the association between coronary heart disease and varying temperature patterns. Copyright (C) 2004 John Wiley Sons, Ltd.

Multiple imputation for body mass index: lessons from the Australian Longitudinal Study on Women's Health

In large epidemiological studies missing data can be a problem, especially if information is sought on a sensitive topic or when a composite measure is calculated from several variables each affected by missing values. Multiple imputation is the method of choice for 'filling in' missing data based on associations among variables. Using an example about body mass index from the Australian Longitudinal Study on Women's Health, we identify a subset of variables that are particularly useful for imputing values for the target variables. Then we illustrate two uses of multiple imputation. The first is to examine and correct for bias when data are not missing completely at random. The second is to impute missing values for an important covariate; in this case omission from the imputation process of variables to be used in the analysis may introduce bias. We conclude with several recommendations for handling issues of missing data. Copyright (C) 2004 John Wiley Sons, Ltd.

Single patient (n-of-1) trials with binary treatment preference

The use of a fully parametric Bayesian method for analysing single patient trials based on the notion of treatment 'preference' is described. This Bayesian hierarchical modelling approach allows for full parameter uncertainty, use of prior information and the modelling of individual and patient sub-group structures. It provides updated probabilistic results for individual patients, and groups of patients with the same medical condition, as they are sequentially enrolled into individualized trials using the same medication alternatives. Two clinically interpretable criteria for determining a patient's response are detailed and illustrated using data from a previously published paper under two different prior information scenarios. Copyright (C) 2005 John Wiley & Sons, Ltd.

Appraising the quality of randomized controlled trials: inter-rater reliability for the OTseeker evidence database

Rationale and aims 'OTseeker' is an online database of randomized controlled trials (RCTs) and systematic reviews relevant to occupational therapy. RCTs are critically appraised and rated for quality using the 'PEDro' scale. We aimed to investigate the inter-rater reliability of the PEDro scale before and after revising rating guidelines. Methods In study 1, five raters scored 100 RCTs using the original PEDro scale guidelines. In study 2, two raters scored 40 different RCTs using revised guidelines. All RCTs were randomly selected from the OTseeker database. Reliability was calculated using Kappa and intraclass correlation coefficients [ICC (model 2,1)]. Results Inter-rater reliability was 'good to excellent' in the first study (Kappas >= 0.53; ICCs >= 0.71). After revising the rating guidelines, the reliability levels were equivalent or higher to those previously obtained (Kappas >= 0.53; ICCs >= 0.89), except for the item, 'groups similar at baseline', which still had moderate reliability (Kappa = 0.53). In study 2, two PEDro scale items, which had their definitions revised, 'less than 15% dropout' and 'point measures and variability', showed higher reliability. In both studies, the PEDro items with the lowest reliability were 'groups similar at baseline' (Kappas = 0.53), 'less than 15% dropout' (Kappas

A Score Test for Zero-Inflation in Correlated Count Data

To account for the preponderance of zero counts and simultaneous correlation of observations, a class of zero-inflated Poisson mixed regression models is applicable for accommodating the within-cluster dependence. In this paper, a score test for zero-inflation is developed for assessing correlated count data with excess zeros. The sampling distribution and the power of the test statistic are evaluated by simulation studies. The results show that the test statistic performs satisfactorily under a wide range of conditions. The test procedure is further illustrated using a data set on recurrent urinary tract infections. Copyright (c) 2005 John Wiley & Sons, Ltd.

An Incremental EM-based Learning Approach for On-Line Prediction of Hospital Resource Utilization

Objective: Inpatient length of stay (LOS) is an important measure of hospital activity, health care resource consumption, and patient acuity. This research work aims at developing an incremental expectation maximization (EM) based learning approach on mixture of experts (ME) system for on-line prediction of LOS. The use of a batchmode learning process in most existing artificial neural networks to predict LOS is unrealistic, as the data become available over time and their pattern change dynamically. In contrast, an on-line process is capable of providing an output whenever a new datum becomes available. This on-the-spot information is therefore more useful and practical for making decisions, especially when one deals with a tremendous amount of data. Methods and material: The proposed approach is illustrated using a real example of gastroenteritis LOS data. The data set was extracted from a retrospective cohort study on all infants born in 1995-1997 and their subsequent admissions for gastroenteritis. The total number of admissions in this data set was n = 692. Linked hospitalization records of the cohort were retrieved retrospectively to derive the outcome measure, patient demographics, and associated co-morbidities information. A comparative study of the incremental learning and the batch-mode learning algorithms is considered. The performances of the learning algorithms are compared based on the mean absolute difference (MAD) between the predictions and the actual LOS, and the proportion of predictions with MAD < 1 day (Prop(MAD < 1)). The significance of the comparison is assessed through a regression analysis. Results: The incremental learning algorithm provides better on-line prediction of LOS when the system has gained sufficient training from more examples (MAD = 1.77 days and Prop(MAD < 1) = 54.3%), compared to that using the batch-mode learning. The regression analysis indicates a significant decrease of MAD (p-value = 0.063) and a significant (p-value = 0.044) increase of Prop(MAD

Electrodynamic heart model construction and ECG simulation

Objectives: In this paper, we present a unified electrodynamic heart model that permits simulations of the body surface potentials generated by the heart in motion. The inclusion of motion in the heart model significantly improves the accuracy of the simulated body surface potentials and therefore also the 12-lead ECG. Methods: The key step is to construct an electromechanical heart model. The cardiac excitation propagation is simulated by an electrical heart model, and the resulting cardiac active forces are used to calculate the ventricular wall motion based on a mechanical model. The source-field point relative position changes during heart systole and diastole. These can be obtained, and then used to calculate body surface ECG based on the electrical heart-torso model. Results: An electromechanical biventricular heart model is constructed and a standard 12-lead ECG is simulated. Compared with a simulated ECG based on the static electrical heart model, the simulated ECG based on the dynamic heart model is more accordant with a clinically recorded ECG, especially for the ST segment and T wave of a V1-V6 lead ECG. For slight-degree myocardial ischemia ECG simulation, the ST segment and T wave changes can be observed from the simulated ECG based on a dynamic heart model, while the ST segment and T wave of simulated ECG based on a static heart model is almost unchanged when compared with a normal ECG. Conclusions: This study confirms the importance of the mechanical factor in the ECG simulation. The dynamic heart model could provide more accurate ECG simulation, especially for myocardial ischemia or infarction simulation, since the main ECG changes occur at the ST segment and T wave, which correspond with cardiac systole and diastole phases.

Multi-level zero-inflated Poisson regression modelling of correlated count data with excess zeros

Count data with excess zeros relative to a Poisson distribution are common in many biomedical applications. A popular approach to the analysis of such data is to use a zero-inflated Poisson (ZIP) regression model. Often, because of the hierarchical Study design or the data collection procedure, zero-inflation and lack of independence may occur simultaneously, which tender the standard ZIP model inadequate. To account for the preponderance of zero counts and the inherent correlation of observations, a class of multi-level ZIP regression model with random effects is presented. Model fitting is facilitated using an expectation-maximization algorithm, whereas variance components are estimated via residual maximum likelihood estimating equations. A score test for zero-inflation is also presented. The multi-level ZIP model is then generalized to cope with a more complex correlation structure. Application to the analysis of correlated count data from a longitudinal infant feeding study illustrates the usefulness of the approach.

Evaluating the effectiveness of Bayesian feature selection

A practical Bayesian approach for inference in neural network models has been available for ten years, and yet it is not used frequently in medical applications. In this chapter we show how both regularisation and feature selection can bring significant benefits in diagnostic tasks through two case studies: heart arrhythmia classification based on ECG data and the prognosis of lupus. In the first of these, the number of variables was reduced by two thirds without significantly affecting performance, while in the second, only the Bayesian models had an acceptable accuracy. In both tasks, neural networks outperformed other pattern recognition approaches.

Psychological cue use and implications for a clinical decision support system

Effective clinical decision making depends upon identifying possible outcomes for a patient, selecting relevant cues, and processing the cues to arrive at accurate judgements of each outcome's probability of occurrence. These activities can be considered as classification tasks. This paper describes a new model of psychological classification that explains how people use cues to determine class or outcome likelihoods. It proposes that clinicians respond to conditional probabilities of outcomes given cues and that these probabilities compete with each other for influence on classification. The model explains why people appear to respond to base rates inappropriately, thereby overestimating the occurrence of rare categories, and a clinical example is provided for predicting suicide risk. The model makes an effective representation for expert clinical judgements and its psychological validity enables it to generate explanations in a form that is comprehensible to clinicians. It is a strong candidate for incorporation within a decision support system for mental-health risk assessment, where it can link with statistical and pattern recognition tools applied to a database of patients. The symbiotic combination of empirical evidence and clinical expertise can provide an important web-based resource for risk assessment, including multi-disciplinary education and training. © 2002 Informa UK Ltd All rights reserved.