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 937 records · Page 52Linked to original sources
PubMed1996

Use of neural networks to diagnose acute myocardial infarction. II. A clinical application.

We investigated the ability of neural networks to diagnose acute myocardial infarction (AMI) from laboratory data only. Several networks were trained with different combinations of data obtained at admission and within the first 12 h and 24 h after admission. The data used included the electrocardiogram (ECG) and the concentrations in serum of potassium, creatine kinase B-subunit (CKB), and lactate dehydrogenase isoenzyme 1 for 250 patients with suspected AMI. Based on admission data, the correct diagnosis was predicted for 76% of the patients in the test group from the ECG data only, and the best combination of ECG results with other variables yielded correct diagnoses for 85% of the test group. Using all of the data available within 24 h, the network predicted the correct diagnosis for 99% of the test data. Almost the same high predictability was obtained by using only two CKB values-recorded at admission and within 12 h after admission-or by using just the latter one. Neural networks and quadratic discriminant analysis performed similarly, but the neural networks were more robust for combinations with many laboratory data.

Creatine Kinase↗
PubMed1995

Neural network system for the evaluation of side-chain packing in protein structures.

An artificial neural network system is used for pattern recognition in protein side-chain-side-chain contact maps. A back-propagation network was trained on a set of patterns which are popular in side-chain contact maps of protein structures. Several neural network architectures and different training parameters were tested to decide on the best combination for the neural network. The resulting network can distinguish between original (from protein structures) and randomized patterns with an accuracy of 84.5% and a Matthews' coefficient of 0.72 for the testing set. Applications of this system for protein structure evaluation and refinement are also proposed. Examples include structures obtained after the application of molecular dynamics to crystal structures, structures obtained from X-ray crystallography at various stages of refinement, structures obtained from a de novo folding algorithm and deliberately misfolded structures.

Amino Acid Sequence↗
PubMed1995

Neural network classification of corneal topography. Preliminary demonstration.

PURPOSE: Videokeratography is a powerful tool for the diagnosis of corneal shape abnormalities. However, interpretation of the topographic map is sometimes difficult, especially when pathologies with similar topographic patterns are suspected. The neural networks model, an artificial intelligence approach, was applied for automated pattern interpretation in corneal topography, and its usefulness was assessed. METHODS: One hundred eighty-three topographic maps were selected and classified by human experts into seven categories: normal, with-the-rule astigmatism, keratoconus (mild, moderate, advanced), postphotorefractive keratectomy, and postkeratoplasty. The maps were divided into a training set (108 maps) and a test set (75 maps). For each map, 11 topography-characterizing indices calculated from the data provided by the TMS-1 videokeratoscope, plus the corresponding diagnosis category, were used to train a neural network. RESULTS: The correct classification was achieved by a trained neural network for all 108 maps in the training set. In the test set, the neural network correctly classified 60 of 75 maps (80%). For every category, accuracy and specificity were greater than 90%, whereas sensitivity ranged from 44% to 100%. CONCLUSIONS: With further testing and refinement, the neural networks paradigm for computer-assisted interpretation or objective classification of videokeratography may become a useful tool to aid the clinician in the diagnosis of corneal topographic abnormalities.

Cornea↗
PubMed2006

Prediction of significant fibrosis in hepatitis C virus infected liver transplant recipients by artificial neural network analysis of clinical factors.

OBJECTIVES: Interest in developing noninvasive markers of liver fibrosis continues to increase, especially in recurrent hepatitis C virus infection after liver transplantation. Recently, a model for predicting significant fibrosis (bridging fibrosis and cirrhosis) on the basis of logistic regression and routine laboratory data has been proposed (logit model). The aim of the present study was to evaluate the accuracy of an artificial neural network, a technique reported to work better than logit models in complex biological situations, built on those same clinical variables and data set of patients, in predicting significant fibrosis. METHODS: The neural network was constructed on the training set of 414 protocol biopsies, from liver transplant recipients, and then tested on the remaining 96 biopsies, as validation set. Model performances of neural network and logit model were evaluated and compared by means of areas under receiver operating characteristic curves. RESULTS: With a cutoff value of >0.4 to predict significant fibrosis, the neural network provided sensitivity, specificity, positive and negative predictive values, respectively, of 100, 79.5, 60.5 and 100%, in the validation set. The performance of the neural network was significantly better than that of the logit model (in the validation set area under the curve = 0.93 vs. 0.84; P = 0.045). CONCLUSIONS: Artificial neural network provides accurate prediction of the presence or absence of significant fibrosis from clinical variables, allowing theoretically protocol liver biopsy to be avoided in several instances, a result of particular interest, given the lack of other types of reliable noninvasive indexes of fibrosis in the setting of transplantation.

Adult↗
PubMed2005

Theoretical note: the C/T ratio in artificial neural networks.

This paper describes computer simulations of the effect of the C/T ratio on acquisition rate in artificial neural networks. The networks consisted of neural processing elements that functioned according to a neurocomputational model whose learning rule is consistent with information on dopaminergic mechanisms of reinforcement. In Simulation 1, three comparisons were made: constant C and variable T, variable C and constant T, and a constant C/T with variable C and T. In the last two comparisons, C was manipulated by changing the probability of reinforcement within the intertrial interval (ITI), in the absence of the conditioned stimulus (CS). Acquisition rate tended to increase with C/T, and the invariant ratio had no effect. In Simulation 2, C was manipulated by changing the ITI, with continuous reinforcement in the presence of the CS and no reinforcements in its absence. Results were comparable to those obtained in Simulation 1. Simulation 3 further explored the effect of the invariant ratio, but with larger absolute values of C and T, which slowed acquisition significantly. The results parallel some experimental findings and theoretical implications of the Gibbon-Balsam model, showing that they can emerge from the moment-to-moment dynamics of a neural-network model. In contrast to that model, however, Simulation 3 suggests that the effect of invariant C/T ratios may be bounded.

Computer Simulation↗
PubMed2000

Generalized and partial synchronization of coupled neural networks.

Synchronization of neural signals has been proposed as a temporal coding scheme representing cooperated computation in distributed cortical networks. Previous theoretical studies in that direction mainly focused on the synchronization of coupled oscillatory subsystems and neglected more complex dynamical modes, that already exist on the single-unit level. In this paper we study the parametrized time-discrete dynamics of two coupled recurrent networks of graded neurons. Conditions for the existence of partially synchronized dynamics of these systems are derived, referring to a situation where only subsets of neurons in each sub-network are synchronous. The coupled networks can have different architectures and even a different number of neurons. Periodic as well as quasiperiodic and chaotic attractors constrained to a manifold M of synchronized components are observed. Examples are discussed for coupled 3-neuron networks having different architectures, and for coupled 2-neuron and 3-neuron networks. Partial synchronization of different degrees is demonstrated by numerical results for selected sets of parameters. In conclusion, the results show that synchronization phenomena far beyond completely synchronized oscillations can occur even in simple coupled networks. The type of the synchronization depends in an intricate way on stimuli, history and connectivity as well as other parameters of the network. Specific inputs can further switch between different operational modes in a complex way, suggesting a similarly rich spatio-temporal behaviour in real neural systems.

Neural Networks, Computer↗
PubMed1994

Use of artificial neural networks within deterministic logic for the computer ECG diagnosis of inferior myocardial infarction.

An investigation into the use of software-based artificial neural networks for the electrocardiographic (ECG) detection of inferior myocardial infarction was made. A total of 592 clinically validated subjects, including 208 with inferior myocardial infarction, 300 normal subjects, and 84 left ventricular hypertrophy cases, were used in this study. A total of 200 ECGs (100 from patients with inferior myocardial infarction and 100 from normal subjects) were fed to 66 supervised feedforward neural networks for training using a back-propagation algorithm. QRS and ST-T wave measurements were used as the input parameters for the neural networks. The best performing network using QRS measurements only and the best using QRS and ST-T data were selected by assessing a test set of 292 ECGs (108 from patients with inferior myocardial infarction, 84 from patients with left ventricular hypertrophy, and 100 from normal subjects). These two networks were then implanted separately into the deterministic Glasgow program for further study. After the implementation, it was found necessary to include a small inferior Q criterion to improve the specificity of reporting inferior myocardial infarction, thereby producing a small loss of sensitivity as compared with use of the network alone. The use of an artificial neural network within the deterministic logic performed better than either alone in the diagnosis of inferior myocardial infarction, producing a 20% gain in sensitivity with 2% loss in overall specificity compared with the original deterministic logic.

Diagnosis, Computer-Assisted↗
PubMed2002

Hybrid model building methodology using unsupervised fuzzy clustering and supervised neural networks.

This paper suggests a model building methodology for dealing with new processes. The methodology, called Hybrid Fuzzy Neural Networks (HFNN), combines unsupervised fuzzy clustering and supervised neural networks in order to create simple and flexible models. Fuzzy clustering was used to define relevant domains on the input space. Then, sets of multilayer perceptrons (MLP) were trained (one for each domain) to map input-output relations, creating, in the process, a set of specified sub-models. The estimated output of the model was obtained by fusing the different sub-model outputs weighted by their predicted possibilities. On-line reinforcement learning enabled improvement of the model. The determination of the optimal number of clusters is fundamental to the success of the HFNN approach. The effectiveness of several validity measures was compared to the generalization capability of the model and information criteria. The validity measures were tested with fermentation simulations and real fermentations of a yeast-like fungus, Aureobasidium pullulans. The results outline the criteria limitations. The learning capability of the HFNN was tested with the fermentation data. The results underline the advantages of HFNN over a single neural network.

Biotechnology↗
PubMed1994

Using feed-forward neural networks for estimation of microbial concentration in a simulated biochemical process.

This work investigated the feasibility of using feed-forward neural networks for estimation of a state variable in a process with highly non-linear characteristics. A biochemical process was considered where the microorganism Saccharomyces cerevisiae, a yeast, grows in a chemostat on a glucose substrate and produces ethanol as a product of primary energy metabolism. Three state variables for the process are the microbial concentration, substrate concentration and product concentration. The Levenberg-Marquardt Method was used to train the neural networks by minimising the sum of squares of the residuals. The inputs to the networks were the measured variable (product concentration) and the control variable (dilution rate). The output of the network was an estimate for the microbial concentration. Earlier work had shown that system identification of this biochemical process could be performed quite well using feed-forward neural networks. This work demonstrated that state estimation can also be performed successfully using feed-forward neural networks. Knowledge of the process model is not required. The method is simple, reliable and accurate enough for engineering purposes. It can save a lot of expense on sensors, their installation and maintenance.

Neural Networks, Computer↗
PubMed1997

A preliminary laboratory investigation of air embolus detection and grading using an artificial neural network.

SUMMARY STATEMENT: Processed digitized Doppler signals abstracted from recordings during continuous air infusion in dogs were used to train a neural network to estimate air embolism infusion rates. BACKGROUND: Precordial Doppler is a sensitive technique for detecting venous air embolism during anesthesia, but it requires constant attentive listening. Since neural networks are particularly well suited to the task of pattern recognition, we sought to investigate this technology for detection and grading of air embolism. METHODS: Air was infused into peripheral veins of four anesthetized dogs at rates of 0.025, 0.05, 0.10, 0.25, 0.50 and 1.0 ml-1.kg-1.min-1 while digital recordings of the precordial Doppler ultrasound signal were collected. The frequency content of the recordings was determined by Fourier analysis. The output of the Fourier transform was the input to a neural network. The network was then trained to estimate the air infusion rate. RESULTS: The correlation coefficient between the size of the air embolism and the air infusion rate was greater than r2 = 0.93 for each of the four animals in the study when the network was trained using the data for all four dogs. When the data from a dog was withheld from the training set and used only for testing the correlation coefficients ranged from r2 = 0.75 to r2 = 0.27. For frequencies below 250 Hz, the acoustic energy tended to fall as the air infusion rate increased. The opposite occurred at frequencies above 325 Hz. CONCLUSIONS: Neural network processing of the precordial Doppler signal provides a quantitative estimate of the size of an air embolism.

Anesthesia↗
PubMed2005

An empirical comparison of combinations of evolutionary algorithms and neural networks for classification problems.

There are numerous combinations of neural networks (NNs) and evolutionary algorithms (EAs) used in classification problems. EAs have been used to train the networks, design their architecture, and select feature subsets. However, most of these combinations have been tested on only a few data sets and many comparisons are done inappropriately measuring the performance on training data or without using proper statistical tests to support the conclusions. This paper presents an empirical evaluation of eight combinations of EAs and NNs on 15 public-domain and artificial data sets. Our objective is to identify the methods that consistently produce accurate classifiers that generalize well. In most cases, the combinations of EAs and NNs perform equally well on the data sets we tried and were not more accurate than hand-designed neural networks trained with simple backpropagation.

Algorithms↗
PubMed2001

Application of a time-delay neural network to promoter annotation in the Drosophila melanogaster genome.

Computational methods for automated genome annotation are critical to understanding and interpreting the bewildering mass of genomic sequence data presently being generated and released. A neural network model of the structural and compositional properties of a eukaryotic core promoter region has been developed and its application for analysis of the Drosophila melanogaster genome is presented. The model uses a time-delay architecture, a special case of a feed-forward neural network. The structure of this model allows for variable spacing between functional binding sites, which is known to play a key role in the transcription initiation process. Application of this model to a test set of core promoters not only gave better discrimination of potential promoter sites than previous statistical or neural network models, but also revealed indirectly subtle properties of the transcription initiation signal. When tested in the Adh region of 2.9 Mbases of the Drosophila genome, the neural network for promoter prediction (NNPP) program that incorporates the time-delay neural network model gives a recognition rate of 75% (69/92) with a false positive rate of 1/547 bases. The present work can be regarded as one of the first intensive studies that applies novel gene regulation technologies to the identification of the complex gene regulation sites in the genome of Drosophila melanogaster.

Animals↗
PubMed2006

Multiperiodicity and exponential attractivity evoked by periodic external inputs in delayed cellular neural networks.

We show that an n-neuron cellular neural network with time-varying delay can have 2(n) periodic orbits located in saturation regions and these periodic orbits are locally exponentially attractive. In addition, we give some conditions for ascertaining periodic orbits to be locally or globally exponentially attractive and allow them to locate in any designated region. As a special case of exponential periodicity, exponential stability of delayed cellular neural networks is also characterized. These conditions improve and extend the existing results in the literature. To illustrate and compare the results, simulation results are discussed in three numerical examples.

Models, Neurological↗
PubMed2005

Wavelet neural network classification of EEG signals by using AR model with MLE preprocessing.

Since EEG is one of the most important sources of information in therapy of epilepsy, several researchers tried to address the issue of decision support for such a data. In this paper, we introduce two fundamentally different approaches for designing classification models (classifiers); the traditional statistical method based on logistic regression and the emerging computationally powerful techniques based on artificial neural networks (ANNs). Logistic regression as well as feedforward error backpropagation artificial neural networks (FEBANN) and wavelet neural networks (WNN) based classifiers were developed and compared in relation to their accuracy in classification of EEG signals. In these methods we used FFT and autoregressive (AR) model by using maximum likelihood estimation (MLE) of EEG signals as an input to classification system with two discrete outputs: epileptic seizure or nonepileptic seizure. By identifying features in the signal we want to provide an automatic system that will support a physician in the diagnosing process. By applying AR with MLE in connection with WNN, we obtained novel and reliable classifier architecture. The network is constructed by the error backpropagation neural network using Morlet mother wavelet basic function as node activation function. The comparisons between the developed classifiers were primarily based on analysis of the receiver operating characteristic (ROC) curves as well as a number of scalar performance measures pertaining to the classification. The WNN-based classifier outperformed the FEBANN and logistic regression based counterpart. Within the same group, the WNN-based classifier was more accurate than the FEBANN-based classifier, and the logistic regression-based classifier.

Action Potentials↗
PubMed1998

Structural change and development in real and artificial neural networks.

Two related but different fields are reviewed. Initially some basic facts about developing real brains are set out and then work on dynamic neural networks is described. A dynamic neural network is defined as any artificial neural network that automatically changes its structure through exposure to input stimuli. Various models are described and evaluated and the functional correlates of both regressive and progressive structural changes are discussed. The paper concludes that, if future modelling work is to be set within a more neurally-plausible framework, then it would be fruitful to examine networks in which the connectivity between extant units is progressively embellished.

Journal Article↗
PubMed2006

Evolutionary neural networks for anomaly detection based on the behavior of a program.

The process of learning the behavior of a given program by using machine-learning techniques (based on system-call audit data) is effective to detect intrusions. Rule learning, neural networks, statistics, and hidden Markov models (HMMs) are some of the kinds of representative methods for intrusion detection. Among them, neural networks are known for good performance in learning system-call sequences. In order to apply this knowledge to real-world problems successfully, it is important to determine the structures and weights of these call sequences. However, finding the appropriate structures requires very long time periods because there are no suitable analytical solutions. In this paper, a novel intrusion-detection technique based on evolutionary neural networks (ENNs) is proposed. One advantage of using ENNs is that it takes less time to obtain superior neural networks than when using conventional approaches. This is because they discover the structures and weights of the neural networks simultaneously. Experimental results with the 1999 Defense Advanced Research Projects Agency (DARPA) Intrusion Detection Evaluation (IDEVAL) data confirm that ENNs are promising tools for intrusion detection.

Algorithms↗
PubMed1998

A new method to estimate parameters of linear compartmental models using artificial neural networks.

At present, the preferred tool for parameter estimation in compartmental analysis is an iterative procedure; weighted nonlinear regression. For a large number of applications, observed data can be fitted to sums of exponentials whose parameters are directly related to the rate constants/coefficients of the compartmental models. Since weighted nonlinear regression often has to be repeated for many different data sets, the process of fitting data from compartmental systems can be very time consuming. Furthermore the minimization routine often converges to a local (as opposed to global) minimum. In this paper, we examine the possibility of using artificial neural networks instead of weighted nonlinear regression in order to estimate model parameters. We train simple feed-forward neural networks to produce as outputs the parameter values of a given model when kinetic data are fed to the networks' input layer. The artificial neural networks produce unbiased estimates and are orders of magnitude faster than regression algorithms. At noise levels typical of many real applications, the neural networks are found to produce lower variance estimates than weighted nonlinear regression in the estimation of parameters from mono- and biexponential models. These results are primarily due to the inability of weighted nonlinear regression to converge. These results establish that artificial neural networks are powerful tools for estimating parameters for simple compartmental models.

Algorithms↗
PubMed1995

Hypnosis modelling in neural networks.

In the framework of the neural network theory effects similar to hypnotic displays are constructed. They are based on the associative paradigm involving non-linear interaction of excitatory and inhibitory channels with synaptic memory. The non-linearity of long-term memorizing processes may cause effects exhibited by blind spots, which are interpreted as the first stage of hypnosis. More complicated phenomena are discussed in terms of a two-layer network.

Humans↗