M. Fernández-Delgado

Intelligent telemonitoring of critical-care patients

Sutil+ is an intelligent monitoring system that is under development. The aim is to endow the system with more intelligent behaviour and to make the presentation of information to the user as flexible as possible, in space, time, and form. We present two telemonitoring solutions that enable access to information resulting from the monitoring of patients in coronary care units, independent of the location of the system user. We concentrate on aspects of data acquisition and storage and, above all, interaction with the user.

Classifying multichannel ECG patterns with an adaptive neural network

In this article the authors describe the application of a new artificial neural network model aimed at the morphological classification of heartbeats detected on a multichannel ECG signal. They emphasize the special characteristics of the algorithm as an adaptive classifier with the capacity to dynamically self-organize its response to the characteristics of the ECG input signal. They also present evaluation results based on traces from the MIT-BIH arrhythmia database.

Time-frequency analysis of heart-rate variability

We present the results of a study that shows the viability of a new technique for the diagnosis and monitoring of myocardial ischemia that is based on the utilization of heart-rate variability (HRV) information. Ischemia is understood as being the lack of oxygen supply to the heart, a situation that in an extreme and irreversible case results in acute myocardial infarction (AMI), a reason for which early detection and treatment is of great interest. The treatment of ischemia can be approached via the evolution of the ECG, and especially from one of the parameters extracted from it-the ST segment (ECG signal between S and T waves) deviation. The utility of this measure is found in its capacity for detecting abnormalities in the conduction of the cardiac impulse that are associated with the presence of ischemia.

Automatic detection and classification of grains of pollen based on shape and texture

Palynological data are used in a wide range of applications. Some studies describe the benefits of the development of a computer system to pollinic analysis. The system should involve the detection of the pollen grains on a slice, and their classification. This paper presents a system that realizes both tasks. The latter is based on the combination of shape and texture analysis. In relation to shape parameters, different ways to understand the contours are presented. The resulting system is evaluated for the discrimination of species of the Urticaceae family which are quite similar. The performance achieved is 89% of correct pollen grain classification

MART: a multichannel ART-based neural network

This paper describes MART, an ART-based neural network for adaptive classification of multichannel signal patterns without prior supervised learning. Like other ART-based classifiers, MART is especially suitable for situations in which not even the number of pattern categories to be distinguished is known a priori; its novelty lies in its truly multichannel orientation, especially its ability to quantify and take into account during pattern classification the different changing reliability of the individual signal channels. The extent to which this ability can reduce the creation of spurious or duplicate categories (a major problem for ART-based classifiers of noisy signals) is illustrated by evaluation of its performance in classifying QRS complexes in two-channel ECG traces which were taken from the MIT-BIH database and contaminated with noise.

Exhaustive comparison of colour texture features and classification methods to discriminate cells categories in histological images of fish ovary

The estimation of fecundity and reproductive cells (oocytes) development dynamic is essential for an accurate study of biology and population dynamics of fish species. This estimation can be developed using the stereometric method to analyse histological images of fish ovary. However, this method still requires specialised technicians and much time and effort to make routinary fecundity studies commonly used in fish stock assessment, because the available software does not allow an automatic analysis. The automatic fecundity estimation requires both the classification of cells depending on their stage of development and the measurement of their diameters, based on those cells that are cut through the nucleous within the histological slide. Human experts seem to use colour and texture properties of the image to classify cells, i.e., colour texture analysis from the computer vision point of view. In the current work, we provide an exhaustive statistical evaluation of a very wide variety of parallel and integrative texture analysis strategies, giving a total of 126 different feature vectors. Besides, a selection of 17 classifiers, representative of the currently available classification techniques, was used to classify the cells according to the presence/absence of nucleous and their stage of development. The Support Vector Machine (SVM) achieves the best results for nucleous (99.0% of accuracy using colour Local Binary Patterns (LPB) feature vector, integrative strategy) and for stages of development (99.6% using First Order Statistics and grey level LPB, parallel strategy) with the species Merluccius merluccius, and similar accuracies for Trisopterus luscus. These results provide a high reliability for an automatic fecundity estimation from histological images of fish ovary.

Rapid Method for Finding Faulty Elements in Antenna Arrays Using Far Field Pattern Samples

A simple and fast technique that allows a diagnosis of faulty elements in antenna arrays, that only needs to consider a small number of samples of its degraded far-field pattern is described. The method tabulates patterns radiated by the array with 1 faulty element only. Then, the pattern corresponding to the configuration of failed/unfailed elements under test is calculated using the error-free pattern and the patterns with 1 faulty element. The configuration with the lowest difference between the calculated and the degraded patterns is selected. Comparison of the performance of this method using an exhaustive search and a genetic algorithm for an equispaced linear array of 100 lambda/2-dipoles is shown. Mutual coupling as well as noise/measurement errors in the pattern samples were considered in the numerical analysis.

Direct Kernel Perceptron (DKP): Ultra-fast kernel ELM-based classification with non-iterative closed-form weight calculation

The Direct Kernel Perceptron (DKP) (Fernández-Delgado et al., 2010) is a very simple and fast kernel-based classifier, related to the Support Vector Machine (SVM) and to the Extreme Learning Machine (ELM) (Huang, Wang, & Lan, 2011), whose α-coefficients are calculated directly, without any iterative training, using an analytical closed-form expression which involves only the training patterns. The DKP, which is inspired by the Direct Parallel Perceptron, (Auer et al., 2008), uses a Gaussian kernel and a linear classifier (perceptron). The weight vector of this classifier in the feature space minimizes an error measure which combines the training error and the hyperplane margin, without any tunable regularization parameter. This weight vector can be translated, using a variable change, to the α-coefficients, and both are determined without iterative calculations. We calculate solutions using several error functions, achieving the best trade-off between accuracy and efficiency with the linear function. These solutions for the α coefficients can be considered alternatives to the ELM with a new physical meaning in terms of error and margin: in fact, the linear and quadratic DKP are special cases of the two-class ELM when the regularization parameter C takes the values C=0 and C=∞. The linear DKP is extremely efficient and much faster (over a vast collection of 42 benchmark and real-life data sets) than 12 very popular and accurate classifiers including SVM, Multi-Layer Perceptron, Adaboost, Random Forest and Bagging of RPART decision trees, Linear Discriminant Analysis, K-Nearest Neighbors, ELM, Probabilistic Neural Networks, Radial Basis Function neural networks and Generalized ART. Besides, despite its simplicity and extreme efficiency, DKP achieves higher accuracies than 7 out of 12 classifiers, exhibiting small differences with respect to the best ones (SVM, ELM, Adaboost and Random Forest), which are much slower. Thus, the DKP provides an easy and fast way to achieve classification accuracies which are not too far from the best one for a given problem. The C and Matlab code of DKP are freely available.

Fast Array Thinning using Global Optimization Methods

A simple and fast method to accelerate the global optimization approaches used in array thinning is described. This method tabulates the contribution of every array element to the far-field pattern in order to improve the numerical efficiency of the optimization algorithm employed. Experiments using our proposal alongside with a genetic algorithm reduce the search computation time about 90%.

Element failure detection in linear antenna arrays using case-based reasoning

The present work proposes a novel case-based reasoning system for fault diagnosis in moderate or large linear antenna arrays. This system identifies the set of elements that are most likely to be defective, helping to significantly reduce the computational costs of their detection (e.g., using an optimization technique such as a genetic algorithm).