Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Neural network”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,639 records · Page 91Linked to original sources
PubMedDate not supplied

ECG data compression using Hebbian neural networks.

Principal component analysis has long been used for a variety of signal processing applications, including signal compression. Neural network implementations of principal component analysis provide a means for unsupervised feature discovery and dimension reduction. In this paper, we describe a method for the compression of ECG data using principal component analysis. Hebbian neural networks were used for principal components computation. A variety of examples of normal and pathological ECGs obtained from the MIT ECG database demonstrate that the proposed method can provide compression ratio up to 30 with PRD% less than 5%.

Algorithms↗
PubMed2002

Use of an artificial neural network to differentiate between ECGs with IRBBB patterns of atrial septal defect and healthy subjects.

Atrial septal defect (ASD) is one of the most commonly recognized congenital cardiac anomalies in adults, but its diagnosis is easily missed because about half of the patients are asymptomatic early in life. Although an electrocardiogram (ECC) diagnosis with an incomplete right bundle branch block (IRBBB) pattern is of major importance for this disease, an RSR' complex similar to an IRBBB pattern is also found in some healthy individuals. A feed-forward artificial neural network was constructed to distinguish between ASD and healthy subjects using 12-lead ECGs with IRBBB. A total of 106 clinically validated subjects, including 58 with ASD and 48 healthy subjects were used in this study. QRS and T wave measurements from I, II, and all precordial leads were used as the input parameters to a back propagation network. The leave-one-out method revealed that in the test data (106 cases), the overall accuracy, sensitivity and specificity of the artificial neural network were 91.5, 91.4, and 91.7%, respectively. This study demonstrates that the neural network technique may offer higher accuracy than computerized ECG diagnosis of IRBBB from the viewpoint of ASD.

Bundle-Branch Block↗
PubMed2003

General-purpose computation with neural networks: a survey of complexity theoretic results.

We survey and summarize the literature on the computational aspects of neural network models by presenting a detailed taxonomy of the various models according to their complexity theoretic characteristics. The criteria of classification include the architecture of the network (feedforward versus recurrent), time model (discrete versus continuous), state type (binary versus analog), weight constraints (symmetric versus asymmetric), network size (finite nets versus infinite families), and computation type (deterministic versus probabilistic), among others. The underlying results concerning the computational power and complexity issues of perceptron, radial basis function, winner-take-all, and spiking neural networks are briefly surveyed, with pointers to the relevant literature. In our survey, we focus mainly on the digital computation whose inputs and outputs are binary in nature, although their values are quite often encoded as analog neuron states. We omit the important learning issues.

Algorithms↗
PubMed2003

The capabilities of artificial neural networks in body composition research.

When estimating in vivo body composition or combining such estimates with other results, multiple variables must be taken into account (e. g. binary attributes such as gender or continuous attributes such as most biosignals). Standard statistical models, such as logistic regression and multivariate analysis, presume well-defined distributions (e. g. normal distribution); they also presume independence among all inputs and only linear relationships, yet rarely are these requirements met in real life. As an alternative to these models, artificial neural networks can be used. In the present work, we describe the pre-processing and multivariate analysis of data using neural network techniques, providing examples from the medical field and making comparisons with classic statistical approaches. We also address the criticisms raised regarding neural network techniques and discuss their potential improvement.

Body Composition↗
PubMed1996

Sleep classification in infants by decision tree-based neural networks.

This paper presents an AI-based approach to automatic sleep stage scoring. The system TBNN (Tree-Based Neural Network) uses a decision-tree generator to provide knowledge that defines the architecture of a backpropagation neural network, including feature selection and initialisation of the weights. The case study reports a successful application to the data from polygraphic all-night sleep of 8 babies aged 6 months. The teaching input was provided by a medical expert in accordance with the rules of Guilleminault and Souquet. The performance of TBNN is compared with 5 other methods and the results are discussed.

Artificial Intelligence↗
PubMed2006

Optimizing the classification of acute lymphoblastic leukemia and acute myeloid leukemia samples using artificial neural networks.

Accurate classification of human blood cells plays a decisive role in the diagnosis and treatment of diseases. Artificial Neural Networks (ANN) have been consistently used as a trusted classification tool for this type of analysis. In this study, a new Artificial Neural Network approach is proposed for the multidimensional classification of two of the most common forms of leukemia: Acute Lymphoblastic Leukemia (ALL) and Acute Myeloid Leukemia (AML), also sometimes called Acute Myelogenous Leukemia. Beckman-Coulter Corporation supplied flow cytometry data of 120 patients that were used in the training and testing phases. The ANN algorithm was thus developed to exploit the different features of the different blood cells provided in an optimized fashion. The goal was to establish a programming tool, supported through this new ANN development, for the identification of normal and abnormal blood samples and provide information to medical doctors in the form of diagnostic references for the specific disease state that is considered for this study. The application of the ANN algorithm produced remarkable classification accuracy results that show a 95% classification accuracy for the normal blood samples and 90% classification accuracy for the abnormal samples even under the ubiquitous problem of overlap.

Blood Cell Count↗
PubMed1992

Application of neural networks: quantitative structure-activity relationships of the derivatives of 2,4-diamino-5-(substituted-benzyl)pyrimidines as DHFR inhibitors.

A comparative study of quantitative structure-activity relationships involving diaminopyrimidines as DHFR inhibitors using regression analysis and the neural-network approach suggests that the neural network can outperform traditional methods. The technique permits the highlighting the functional form of those parameters which have an influence on the biological activity.

Drug Design↗
PubMed2001

A hybrid artificial neural network as a software sensor for optimal control of a wastewater treatment process.

For control and automation of biological treatment processes, lack of reliable on-line sensors to measure water quality parameters is one of the most important problems to overcome. Many parameters cannot be measured directly with on-line sensors. The accuracy of existing hardware sensors is also not sufficient and maintenance problems such as electrode fouling often cause trouble. This paper deals with the development of software sensor techniques that estimate the target water quality parameter from other parameters using the correlation between water quality parameters. We focus our attention on the preprocessing of noisy data and the selection of the best model feasible to the situation. Problems of existing approaches are also discussed. We propose a hybrid neural network as a software sensor inferring wastewater quality parameter. Multivariate regression, artificial neural networks (ANN), and a hybrid technique that combines principal component analysis as a preprocessing stage are applied to data from industrial wastewater processes. The hybrid ANN technique shows an enhancement of prediction capability and reduces the overfitting problem of neural networks. The result shows that the hybrid ANN technique can be used to extract information from noisy data and to describe the nonlinearity of complex wastewater treatment processes.

Automation↗
PubMed1996

Neural network computer simulation of medical aerosols.

Preliminary investigations have been conducted to assess the potential for using artificial neural networks to simulate aerosol behaviour, with a view to employing this type of methodology in the evaluation and design of pulmonary drug-delivery systems. Details are presented of the general purpose software developed for these tasks; it implements a feed-forward back-propagation algorithm with weight decay and connection pruning, the user having complete run-time control of the network architecture and mode of training. A series of exploratory investigations is then reported in which different network structures and training strategies are assessed in terms of their ability to simulate known patterns of fluid flow in simple model systems. The first of these involves simulations of cellular automata-generated data for fluid flow through a partially obstructed two-dimensional pipe. The artificial neural networks are shown to be highly successful in simulating the behaviour of this simple linear system, but with important provisos relating to the information content of the training data and the criteria used to judge when the network is properly trained. A second set of investigations is then reported in which similar networks are used to simulate patterns of fluid flow through aerosol generation devices, using training data furnished through rigorous computational fluid dynamics modelling. These more complex three-dimensional systems are modelled with equal success. It is concluded that carefully tailored, well trained networks could provide valuable tools not just for predicting but also for analysing the spatial dynamics of pharmaceutical aerosols.

Aerosols↗
PubMed2003

Global and robust stability of interval Hopfield neural networks with time-varying delays.

In this paper, we investigate the problem of global and robust stability of a class of interval Hopfield neural networks that have time-varying delays. Some criteria for the global and robust stability of such networks are derived, by means of constructing suitable Lyapunov functionals for the networks. As a by-product, for the conventional Hopfield neural networks with time-varying delays, we also obtain some new criteria for their global and asymptotic stability.

Algorithms↗
PubMed2003

Hybrid canonical-correlation neural-network approach applied to nonnucleoside HIV-1 reverse transcriptase inhibitors (HEPT derivatives).

Beneficial antiviral HIV-1 chemotherapy is associated with adverse reactions. To optimize the desired actions and to lower the side effects of nonnucleoside HIV-1 reverse transcriptase (RT) inhibitors (NNRTIs), quantitative structure-activity relationships (QSARs) were studied by using a series of HEPT derivatives of NNRTIs. Hypothesis testing requires that certain assumptions are approximately satisfied in statistically based QSARs, however. A complementary approach is based on artificial neural network analysis. Model building can be made without the manifold assumptions of statistically based QSAR approaches but the problem is that the number of neural weights increase exponentially (danger of overfitting) under certain circumstances. A way to get more reliable results is to reduce the dimensionality of the two subsets (biological and chemical variables). A suitable method is the canonical correlation analysis. The two subsets of canonical variates are used as outputs (biologically derived variates) and inputs (chemically derived variates) of an optimized backpropagation neural network approach. The contribution summarizes the most recent results of this canonical-correlation backpropagation-neural network QSAR approach. It is shown that noncovalent interactions (lipophilic, steric, hydrogen-bonding, and inductive forces of the substituents) are responsible for the antiviral and cytotoxic actions. The outcome of this analysis produces an internally highly self-consistent result (model robustness). The predictive performance is tested. The butterfly-like conformation of the predicted compound is consistent with the butterfly-like model of other NNRTIs. Molecular simulation shows that the complexed drug interacts with the Tyr181 and Tyr188 residues of the RT. The uracil ring of the drug binds directly with Lys101, and the acyclic side chain (with an intact free hydroxyl function) binds with Lys103. The suggested noncovalent interaction forces are equivalent with that found by the QSAR analysis.

Anti-HIV Agents↗
PubMed2005

Retrieval of transverse relaxation time distribution from spin-echo data by recurrent neural network.

Inversion of transverse relaxation time decay curve from spin-echo experiments was carried out using Hopfield neural network, to obtain the transverse relaxation time distribution. The performance of this approach was tested against simulated and experimental data. The initial guess, necessary for the integration procedure, was established as the analytical Laplace inversion. Together with errors in the simulated data, inversion was also carried out with errors in this initial guess. The probability density function, calculated by the neural network, is used in multiple sclerosis diagnostics.

Humans↗
PubMed1994

Neural networks as a new method for EEG analysis. A basic introduction.

This paper discusses the general usability of artificial neural networks for the analysis of EEG data. The advantages and drawbacks of this new technology are compared, especially from the perspective of medical computer science. Furthermore this text clarifies the fundamental principles of neural networks, the relations to the biological model and how they compare to classical evaluation methods.

Electroencephalography↗
PubMed1999

Automated prozone effect detection in ferritin homogeneous immunoassays using neural network classifiers.

The application of turbidimetric homogeneous immunoassays made the determination of several plasma components widely available. The sensitivity and accuracy of these assays are appropriate enough for routine laboratory use; however, in the case of many pathologically high concentration samples, prozone effect (high dose hook effect) can be observed, that leads to false-negative determination. Up to the present there are no cost-effective algorithms available for the safe detection of the prozone effect. Pathological serum ferritin values can be elevated up to 5000 ng/ml, while the measuring range covers only the 0-440 ng/ml range by a commercial assay. The determination of samples with ferritin concentration higher than 1500 ng/ml results in false-negative values because of the overlapping measuring range and prozone effect range. The prozone effect can be recognised by analysis of reaction kinetics after measurement. We have developed a neural network classifier system to analyse reaction kinetics of the measurements and check the prozone effect. One thousand five hundred determinations and 77 patient samples were used for neural network training and test. Using the trained neural networks, false-negative results can be filtered immediately after the determination, without re-run; thus, the sensitivity of plasma ferritin determination may become reliable enough, even in the case of high concentration samples. Applying this new technology, false-negative serum ferritin determinations can be avoided, thus even a relatively high hook effect rate (5-12% in different patient groups) can be handled safely.

Blood Chemical Analysis↗
PubMed2005

Autolabeling 3D tracks using neural networks.

Motion capturing systems based on monochrome video have problems assigning measured 3D marker positions to the anatomically defined positions or labels of the markers applied to the test subject. This task is usually called "labelling" and is paramount to the reconstruction of 3D trajectories from a set of video frames from multiple cameras--the tracking procedure. Labelling means sorting a set of 3D vectors by their spatial positions. Neural networks can be made to "learn" from examples of marker positions in a given marker set, i.e. previously manually tracked video sequences. Trained neural networks are able to calculate a set of sorted approximate marker positions from an unsorted set of exact marker positions. The set of sorted exact positions can be found by pairing up both sets of marker positions via a minimum distance function. The neural network is trained only once and can then be applied to any number of individuals. The algorithm is designed for cyclic motions like for locomotion analysis.

Algorithms↗
PubMed1998

Hybrid artificial neural network segmentation and classification of dynamic contrast-enhanced MR imaging (DEMRI) of osteosarcoma.

The evaluation of pediatric osteosarcoma has suffered from the lack of an accurate imaging measure of response. One major problem is that osteosarcoma do not shrink in response to chemotherapy; instead, viable tumor is replaced by necrotic tissue. Currently available techniques that use dynamic contrast-enhanced magnetic resonance imaging to quantitatively evaluate tumor response fail to assess the percentage of necrosis. At present, histopathologic evaluation of resected tissue is the only means of measuring the percentage of necrosis in treated osteosarcoma. The current study presents a non-invasive method to visualize necrotic and viable tumor and quantitatively assess the response of osteosarcoma. Our technique uses a hybrid neural network consisting of a Kohonen self-organizing map to segment dynamic contrast-enhanced magnetic resonance images and a multi-layer backpropagation neural network to classify the segmented images. Because the hybrid neural network is completely automated, our technique removes both inter- and intra-operator error. An analysis comparing the percentage of necrosis from our technique to the histopathologic analysis revealed a highly significant Spearman correlation coefficient of 0.617 with p < 0.001.

Adolescent↗
PubMed1995

Structure optimization of neural networks with A* -algorithm application in EEG pattern analysis.

The A* - Algorithm for heuristic search is applied to construct a Neural Network structure (NS) that optimally fits the structure of data to be learned. In this way, the user of Neural Networks (NN) is able to avoid the empirical testing of different structures. The method given here is applied to the recognition of different patterns derived from the EEG of an epileptic patient.

Algorithms↗
PubMed1994

Protein secondary structure from circular dichroism spectroscopy. Combining variable selection principle and cluster analysis with neural network, ridge regression and self-consistent methods.

Different approaches to improve the analysis of protein secondary structure from circular dichroism spectra are compared. Grouping proteins based on the similarity of their circular dichroism spectra, using cluster analysis methods, was utilized as a new way of implementing variable selection. The performance of three basic methods (neural networks, ridge regression and singular value decomposition) was evaluated in combination with three approaches to improve the predictions; namely, variable selection, cluster analysis and the self-consistent method. Cluster analysis performed on the basis set proteins resulted in three clusters, subanalyses of which provide a new way of performing variable selection. The neural network with two hidden layers performed better than that with one hidden layer and was combined with variable selection. Inclusion of the variable selection principle improved the performance of all three basic methods. While the neural network method performed slightly better than the other two methods at the basic level, the inclusion of variable selection led to similar performance indices for all three methods.

Circular Dichroism↗