Postmortem quantitative 1.5-T MRI for the differentiation and characterization of serous fluids, blood, CSF, and putrefied CSF.

The purpose of the present study was to investigate whether serous fluids, blood, cerebrospinal fluid (CSF), and putrefied CSF can be characterized and differentiated in synthetically calculated magnetic resonance (MR) images based on their quantitative T 1, T 2, and proton density (PD) values. Images from 55 postmortem short axis cardiac and 31 axial brain 1.5-T MR examinations were quantified using a quantification sequence. Serous fluids, fluid blood, sedimented blood, blood clots, CSF, and putrefied CSF were analyzed for their mean T 1, T 2, and PD values. Body core temperature was measured during the MRI scans. The fluid-specific quantitative values were related to the body core temperature. Equations to correct for temperature differences were generated. In a 3D plot as well as in statistical analysis, the quantitative T 1, T 2 and PD values of serous fluids, fluid blood, sedimented blood, blood clots, CSF, and putrefied CSF could be well differentiated from each other. The quantitative T 1 and T 2 values were temperature-dependent. Correction of quantitative values to a temperature of 37 °C resulted in significantly better discrimination between all investigated fluid mediums. We conclude that postmortem 1.5-T MR quantification is feasible to discriminate between blood, serous fluids, CSF, and putrefied CSF. This finding provides a basis for the computer-aided diagnosis and detection of fluids and hemorrhages.

Lung Pattern Classification for Interstitial Lung Diseases Using a Deep Convolutional Neural Network

Automated tissue characterization is one of the most crucial components of a computer aided diagnosis (CAD) system for interstitial lung diseases (ILDs). Although much research has been conducted in this field, the problem remains challenging. Deep learning techniques have recently achieved impressive results in a variety of computer vision problems, raising expectations that they might be applied in other domains, such as medical image analysis. In this paper, we propose and evaluate a convolutional neural network (CNN), designed for the classification of ILD patterns. The proposed network consists of 5 convolutional layers with 2×2 kernels and LeakyReLU activations, followed by average pooling with size equal to the size of the final feature maps and three dense layers. The last dense layer has 7 outputs, equivalent to the classes considered: healthy, ground glass opacity (GGO), micronodules, consolidation, reticulation, honeycombing and a combination of GGO/reticulation. To train and evaluate the CNN, we used a dataset of 14696 image patches, derived by 120 CT scans from different scanners and hospitals. To the best of our knowledge, this is the first deep CNN designed for the specific problem. A comparative analysis proved the effectiveness of the proposed CNN against previous methods in a challenging dataset. The classification performance (~85.5%) demonstrated the potential of CNNs in analyzing lung patterns. Future work includes, extending the CNN to three-dimensional data provided by CT volume scans and integrating the proposed method into a CAD system that aims to provide differential diagnosis for ILDs as a supportive tool for radiologists.

Artificial Neural Nets with Interaction of Afferents

The aim is to obtain computationally more powerful, neuro physiologically founded, artificial neurons and neural nets. Artificial Neural Nets (ANN) of the Perceptron type evolved from the original proposal by McCulloch an Pitts classical paper [1]. Essentially, they keep the computing structure of a linear machine followed by a non linear operation. The McCulloch-Pitts formal neuron (which was never considered by the author’s to be models of real neurons) consists of the simplest case of a linear computation of the inputs followed by a threshold. Networks of one layer cannot compute anylogical function of the inputs, but only those which are linearly separable. Thus, the simple exclusive OR (contrast detector) function of two inputs requires two layers of formal neurons

Dendritic-Like Reliable Computation in Artificial Neurons

Dendritic computation is a term that has been in neuro physiological research for a long time [1]. It is still controversial and far for been clarified within the concepts of both computation and neurophysiology [2], [3]. In any case, it hasnot been integrated neither in a formal computational scheme or structure, nor into formulations of artificial neural nets. Our objective here is to formulate a type of distributed computation that resembles dendritic trees, in such a way that it shows the advantages of neural network distributed computation, mostly the reliability that is shown under the existence of holes (scotomas) in the computing net, without ?blind spots?.

Self-learning of fault diagnosis identification

A good and early fault detection and isolation system along with efficient alarm management and fine sensor validation systems are very important in today¿s complex process plants, specially in terms of safety enhancement and costs reduction. This paper presents a methodology for fault characterization. This is a self-learning approach developed in two phases. An initial, learning phase, where the simulation of process units, without and with different faults, will let the system (in an automated way) to detect the key variables that characterize the faults. This will be used in a second (on line) phase, where these key variables will be monitored in order to diagnose possible faults. Using this scheme the faults will be diagnosed and isolated in an early stage where the fault still has not turned into a failure.

Regularization for sparsity in statistical analysis and machine learning

Pragmatism is the leading motivation of regularization. We can understand regularization as a modification of the maximum-likelihood estimator so that a reasonable answer could be given in an unstable or ill-posed situation. To mention some typical examples, this happens when fitting parametric or non-parametric models with more parameters than data or when estimating large covariance matrices. Regularization is usually used, in addition, to improve the bias-variance tradeoff of an estimation. Then, the definition of regularization is quite general, and, although the introduction of a penalty is probably the most popular type, it is just one out of multiple forms of regularization. In this dissertation, we focus on the applications of regularization for obtaining sparse or parsimonious representations, where only a subset of the inputs is used. A particular form of regularization, L1-regularization, plays a key role for reaching sparsity. Most of the contributions presented here revolve around L1-regularization, although other forms of regularization are explored (also pursuing sparsity in some sense). In addition to present a compact review of L1-regularization and its applications in statistical and machine learning, we devise methodology for regression, supervised classification and structure induction of graphical models. Within the regression paradigm, we focus on kernel smoothing learning, proposing techniques for kernel design that are suitable for high dimensional settings and sparse regression functions. We also present an application of regularized regression techniques for modeling the response of biological neurons. Supervised classification advances deal, on the one hand, with the application of regularization for obtaining a na¨ıve Bayes classifier and, on the other hand, with a novel algorithm for brain-computer interface design that uses group regularization in an efficient manner. Finally, we present a heuristic for inducing structures of Gaussian Bayesian networks using L1-regularization as a filter. El pragmatismo es la principal motivación de la regularización. Podemos entender la regularización como una modificación del estimador de máxima verosimilitud, de tal manera que se pueda dar una respuesta cuando la configuración del problema es inestable. A modo de ejemplo, podemos mencionar el ajuste de modelos paramétricos o no paramétricos cuando hay más parámetros que casos en el conjunto de datos, o la estimación de grandes matrices de covarianzas. Se suele recurrir a la regularización, además, para mejorar el compromiso sesgo-varianza en una estimación. Por tanto, la definición de regularización es muy general y, aunque la introducción de una función de penalización es probablemente el método más popular, éste es sólo uno de entre varias posibilidades. En esta tesis se ha trabajado en aplicaciones de regularización para obtener representaciones dispersas, donde sólo se usa un subconjunto de las entradas. En particular, la regularización L1 juega un papel clave en la búsqueda de dicha dispersión. La mayor parte de las contribuciones presentadas en la tesis giran alrededor de la regularización L1, aunque también se exploran otras formas de regularización (que igualmente persiguen un modelo disperso). Además de presentar una revisión de la regularización L1 y sus aplicaciones en estadística y aprendizaje de máquina, se ha desarrollado metodología para regresión, clasificación supervisada y aprendizaje de estructura en modelos gráficos. Dentro de la regresión, se ha trabajado principalmente en métodos de regresión local, proponiendo técnicas de diseño del kernel que sean adecuadas a configuraciones de alta dimensionalidad y funciones de regresión dispersas. También se presenta una aplicación de las técnicas de regresión regularizada para modelar la respuesta de neuronas reales. Los avances en clasificación supervisada tratan, por una parte, con el uso de regularización para obtener un clasificador naive Bayes y, por otra parte, con el desarrollo de un algoritmo que usa regularización por grupos de una manera eficiente y que se ha aplicado al diseño de interfaces cerebromáquina. Finalmente, se presenta una heurística para inducir la estructura de redes Bayesianas Gaussianas usando regularización L1 a modo de filtro.

Seed image analysis and taxonomy of Diplotaxis DC. (Brassicaceae, Brassiceae)

The genus Diplotaxis, comprising 32 or 34 species, plus several additional infraspecific taxa, displays a considerable degree of heterogeneity in the morphology, molecular markers, chromosome numbers and geographical amplitude of the species. The taxonomic relationships within the genus Diplotaxis were investigated by phenetic characterisation of germplasm belonging to 27 taxa of the genus, because there is an increasing interest in Diplotaxis, since some of its species (D. tenuifolia, D. muralis) are gathered or cultivated for human consumption, whereas others are frequent arable weeds (D. erucoides) in many European vineyards. Using a computer-aided vision system, 33 morpho-colorimetric features of seeds were electronically measured. The data were used to implement a statistical classifier, which is able to discriminate the taxa within the genus Diplotaxis, in order to compare the resulting species grouping with the current infrageneric systematics of this genus. Despite the high heterogeneity of the samples, due to the great intra-population variability, the stepwise Linear Discriminant Analysis method, applied to distinguish the groups, was able to reach over 80% correct identification. The results obtained allowed us to confirm the current taxonomic position of most taxa and suggested the taxonomic position of others for reconsideration.

Information integration: generating functional models from structural ones

Information integration is a very important topic. Reusing the knowledge and having common representations have been (and it is) an active research topic in the process systems community. Conventional (structural) But only structural models have been dealt with so far. In this paper the issue of integration is related with two types of different knowledge, functional and structural. Functional representation and analysis have proved very useful, but still it is developed and presented in a completely isolated way from the classic structural description of the process. This paper presents an architecture to integrate both representations.

HAZOP studies using a functional modeling framework

In this paper we present a tool to perform guided HAZOP studies using a functional modeling framework: D-higraphs. It is a formalism that gathers in a single model structural (ontological) and functional information about the process considered. In this paper it is applied to an industrial case showing that the proposed methodology fits its purposes and fulfills some of the gaps and drawbacks existing in previous reported HAZOP assistant tools.

Ontology driven description and engineering of autonomous systems: application to process system engineering

Autonomous systems refer to systems capable of operating in a real world environment without any form of external control for extended periods of time. Autonomy is a desired goal for every system as it improves its performance, safety and profit. Ontologies are a way to conceptualize the knowledge of a specific domain. In this paper an ontology for the description of autonomous systems as well as for its development (engineering) is presented and applied to a process. This ontology is intended to be applied and used to generate final applications following a model driven methodology.

Control and command systems concepts from early work on a Mars Rover

We recover and develop some robotic systems concepts (on the light of present systems tools) that were originated for an intended Mars Rover in the sixties of the last century at the Instrumentation Laboratory of MIT, where one of the authors was involved. The basic concepts came from the specifications for a type of generalized robot inspired in the structure of the vertebrate nervous systems, where the decision system was based in the structure and function of the Reticular Formation (RF). The vertebrate RF is supposed to commit the whole organism to one among various modes of behavior, so taking the decisions about the present overall task. That is, it is a kind of control and command system. In this concepts updating, the basic idea is that the RF comprises a set of computing units such that each computing module receives information only from a reduced part of the overall, little processed sensory inputs. Each computing unit is capable of both general diagnostics about overall input situations and of specialized diagnostics according to the values of a concrete subset of the input lines. Slave systems to this command and control computer, there are the sensors, the representations of external environment, structures for modeling and planning and finally, the effectors acting in the external world.

The p-Adaptive Biem Approach for Two-Dimensional Elasticity Analysis

This paper presents the development and application of the p-adaptive BIEM version in elastostatics. The basic concepts underlying the p-adaptive technique are summarized and discussed. Some Pascal pseudocodes which show the way how such a technique can be implemented easily in microcomputers are also provided. Both the applicability and the accuracy of the method proposed here are illustrated through a numerical example.

Modelos Estadísticos Dinámicos para Análisis de Evolución de Tasas de Accidentes: Metodología y Herramientas Computacionales

Los accidentes del tráfico son un fenómeno social muy relevantes y una de las principales causas de mortalidad en los países desarrollados. Para entender este fenómeno complejo se aplican modelos econométricos sofisticados tanto en la literatura académica como por las administraciones públicas. Esta tesis está dedicada al análisis de modelos macroscópicos para los accidentes del tráfico en España. El objetivo de esta tesis se puede dividir en dos bloques: a. Obtener una mejor comprensión del fenómeno de accidentes de trafico mediante la aplicación y comparación de dos modelos macroscópicos utilizados frecuentemente en este área: DRAG y UCM, con la aplicación a los accidentes con implicación de furgonetas en España durante el período 2000-2009. Los análisis se llevaron a cabo con enfoque frecuencista y mediante los programas TRIO, SAS y TRAMO/SEATS. b. La aplicación de modelos y la selección de las variables más relevantes, son temas actuales de investigación y en esta tesis se ha desarrollado y aplicado una metodología que pretende mejorar, mediante herramientas teóricas y prácticas, el entendimiento de selección y comparación de los modelos macroscópicos. Se han desarrollado metodologías tanto para selección como para comparación de modelos. La metodología de selección de modelos se ha aplicado a los accidentes mortales ocurridos en la red viaria en el período 2000-2011, y la propuesta metodológica de comparación de modelos macroscópicos se ha aplicado a la frecuencia y la severidad de los accidentes con implicación de furgonetas en el período 2000-2009. Como resultado de los desarrollos anteriores se resaltan las siguientes contribuciones: a. Profundización de los modelos a través de interpretación de las variables respuesta y poder de predicción de los modelos. El conocimiento sobre el comportamiento de los accidentes con implicación de furgonetas se ha ampliado en este proceso. bl. Desarrollo de una metodología para selección de variables relevantes para la explicación de la ocurrencia de accidentes de tráfico. Teniendo en cuenta los resultados de a) la propuesta metodológica se basa en los modelos DRAG, cuyos parámetros se han estimado con enfoque bayesiano y se han aplicado a los datos de accidentes mortales entre los años 2000-2011 en España. Esta metodología novedosa y original se ha comparado con modelos de regresión dinámica (DR), que son los modelos más comunes para el trabajo con procesos estocásticos. Los resultados son comparables, y con la nueva propuesta se realiza una aportación metodológica que optimiza el proceso de selección de modelos, con escaso coste computacional. b2. En la tesis se ha diseñado una metodología de comparación teórica entre los modelos competidores mediante la aplicación conjunta de simulación Monte Cario, diseño de experimentos y análisis de la varianza ANOVA. Los modelos competidores tienen diferentes estructuras, que afectan a la estimación de efectos de las variables explicativas. Teniendo en cuenta el estudio desarrollado en bl) este desarrollo tiene el propósito de determinar como interpretar la componente de tendencia estocástica que un modelo UCM modela explícitamente, a través de un modelo DRAG, que no tiene un método específico para modelar este elemento. Los resultados de este estudio son importantes para ver si la serie necesita ser diferenciada antes de modelar. b3. Se han desarrollado nuevos algoritmos para realizar los ejercicios metodológicos, implementados en diferentes programas como R, WinBUGS, y MATLAB. El cumplimiento de los objetivos de la tesis a través de los desarrollos antes enunciados se remarcan en las siguientes conclusiones: 1. El fenómeno de accidentes del tráfico se ha analizado mediante dos modelos macroscópicos. Los efectos de los factores de influencia son diferentes dependiendo de la metodología aplicada. Los resultados de predicción son similares aunque con ligera superioridad de la metodología DRAG. 2. La metodología para selección de variables y modelos proporciona resultados prácticos en cuanto a la explicación de los accidentes de tráfico. La predicción y la interpretación también se han mejorado mediante esta nueva metodología. 3. Se ha implementado una metodología para profundizar en el conocimiento de la relación entre las estimaciones de los efectos de dos modelos competidores como DRAG y UCM. Un aspecto muy importante en este tema es la interpretación de la tendencia mediante dos modelos diferentes de la que se ha obtenido información muy útil para los investigadores en el campo del modelado. Los resultados han proporcionado una ampliación satisfactoria del conocimiento en torno al proceso de modelado y comprensión de los accidentes con implicación de furgonetas y accidentes mortales totales en España. ABSTRACT Road accidents are a very relevant social phenomenon and one of the main causes of death in industrialized countries. Sophisticated econometric models are applied in academic work and by the administrations for a better understanding of this very complex phenomenon. This thesis is thus devoted to the analysis of macro models for road accidents with application to the Spanish case. The objectives of the thesis may be divided in two blocks: a. To achieve a better understanding of the road accident phenomenon by means of the application and comparison of two of the most frequently used macro modelings: DRAG (demand for road use, accidents and their gravity) and UCM (unobserved components model); the application was made to van involved accident data in Spain in the period 2000-2009. The analysis has been carried out within the frequentist framework and using available state of the art software, TRIO, SAS and TRAMO/SEATS. b. Concern on the application of the models and on the relevant input variables to be included in the model has driven the research to try to improve, by theoretical and practical means, the understanding on methodological choice and model selection procedures. The theoretical developments have been applied to fatal accidents during the period 2000-2011 and van-involved road accidents in 2000-2009. This has resulted in the following contributions: a. Insight on the models has been gained through interpretation of the effect of the input variables on the response and prediction accuracy of both models. The behavior of van-involved road accidents has been explained during this process. b1. Development of an input variable selection procedure, which is crucial for an efficient choice of the inputs. Following the results of a) the procedure uses the DRAG-like model. The estimation is carried out within the Bayesian framework. The procedure has been applied for the total road accident data in Spain in the period 2000-2011. The results of the model selection procedure are compared and validated through a dynamic regression model given that the original data has a stochastic trend. b2. A methodology for theoretical comparison between the two models through Monte Carlo simulation, computer experiment design and ANOVA. The models have a different structure and this affects the estimation of the effects of the input variables. The comparison is thus carried out in terms of the effect of the input variables on the response, which is in general different, and should be related. Considering the results of the study carried out in b1) this study tries to find out how a stochastic time trend will be captured in DRAG model, since there is no specific trend component in DRAG. Given the results of b1) the findings of this study are crucial in order to see if the estimation of data with stochastic component through DRAG will be valid or whether the data need a certain adjustment (typically differencing) prior to the estimation. The model comparison methodology was applied to the UCM and DRAG models, considering that, as mentioned above, the UCM has a specific trend term while DRAG does not. b3. New algorithms were developed for carrying out the methodological exercises. For this purpose different softwares, R, WinBUGs and MATLAB were used. These objectives and contributions have been resulted in the following findings: 1. The road accident phenomenon has been analyzed by means of two macro models: The effects of the influential input variables may be estimated through the models, but it has been observed that the estimates vary from one model to the other, although prediction accuracy is similar, with a slight superiority of the DRAG methodology. 2. The variable selection methodology provides very practical results, as far as the explanation of road accidents is concerned. Prediction accuracy and interpretability have been improved by means of a more efficient input variable and model selection procedure. 3. Insight has been gained on the relationship between the estimates of the effects using the two models. A very relevant issue here is the role of trend in both models, relevant recommendations for the analyst have resulted from here. The results have provided a very satisfactory insight into both modeling aspects and the understanding of both van-involved and total fatal accidents behavior in Spain.

Optimization methods for the operating room management under uncertainty: stochastic programming vs. decomposition approach

The operating theatres are the engine of the hospitals; proper management of the operating rooms and its staff represents a great challenge for managers and its results impact directly in the budget of the hospital. This work presents a MILP model for the efficient schedule of multiple surgeries in Operating Rooms (ORs) during a working day. This model considers multiple surgeons and ORs and different types of surgeries. Stochastic strategies are also implemented for taking into account the uncertain in surgery durations (pre-incision, incision, post-incision times). In addition, a heuristic-based methods and a MILP decomposition approach is proposed for solving large-scale ORs scheduling problems in computational efficient way. All these computer-aided strategies has been implemented in AIMMS, as an advanced modeling and optimization software, developing a user friendly solution tool for the operating room management under uncertainty.

Managing daily surgery schedules in a teaching hospital: a mixed-integer optimization approach

Background: This study examined the daily surgical scheduling problem in a teaching hospital. This problem relates to the use of multiple operating rooms and different types of surgeons in a typical surgical day with deterministic operation durations (preincision, incision, and postincision times). Teaching hospitals play a key role in the health-care system; however, existing models assume that the duration of surgery is independent of the surgeon's skills. This problem has not been properly addressed in other studies. We analyze the case of a Spanish public hospital, in which continuous pressures and budgeting reductions entail the more efficient use of resources. Methods: To obtain an optimal solution for this problem, we developed a mixed-integer programming model and user-friendly interface that facilitate the scheduling of planned operations for the following surgical day. We also implemented a simulation model to assist the evaluation of different dispatching policies for surgeries and surgeons. The typical aspects we took into account were the type of surgeon, potential overtime, idling time of surgeons, and the use of operating rooms. Results: It is necessary to consider the expertise of a given surgeon when formulating a schedule: such skill can decrease the probability of delays that could affect subsequent surgeries or cause cancellation of the final surgery. We obtained optimal solutions for a set of given instances, which we obtained through surgical information related to acceptable times collected from a Spanish public hospital. Conclusions: We developed a computer-aided framework with a user-friendly interface for use by a surgical manager that presents a 3-D simulation of the problem. Additionally, we obtained an efficient formulation for this complex problem. However, the spread of this kind of operation research in Spanish public health hospitals will take a long time since there is a lack of knowledge of the beneficial techniques and possibilities that operational research can offer for the health-care system.