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 1,081 records · Page 60Linked to original sources
PubMed2006

Accurate prediction of the blood-brain partitioning of a large set of solutes using ab initio calculations and genetic neural network modeling.

A genetic algorithm-based artificial neural network model has been developed for the accurate prediction of the blood-brain barrier partitioning (in logBB scale) of chemicals. A data set of 123 logBB (115 old molecules and 8 new molecules) of a diverse set of chemicals was chosen in this study. The optimum 3D geometry of the molecules was estimated by the ab initio calculations at the level of RHF/STO-3G, and consequently, different electronic descriptors were calculated for each molecule. Indeed, logP as a measure of hydrophobicity and different topological indices were also calculated. A three-layered artificial neural network with backpropagation of an error-learning algorithm was employed to process the nonlinear relationship between the calculated descriptors and logBB data. Genetic algorithm was used as a feature selection method to select the most relevant set of descriptors as the input of the network. Modeling of the logBB data by the only quantum descriptors produced a 5:4:1 ANN structure with RMS error of validation and crossvalidation equal to 0.224 and 0.227, respectively. Better nonlinear model (RMS(V) and RMS(CV) equals to 0.097 and 0.099, respectively) was obtained by the incorporation of the logP and the principal components of the topological indices to electronic descriptors. The ultimate performances of the models were obtained by the application of the models to predict the logBB of 23 molecules that did not have contribution in the steps of model development. The best model produced RMS error of prediction 0.140, and could predict about 98% of variances in the logBB data.

Blood-Brain Barrier↗
PubMed2003

Stability of a neural network model with small-world connections.

Small-world networks are highly clustered networks with small distances among the nodes. There are many biological neural networks that present this kind of connection. There are no special weightings in the connections of most existing small-world network models. However, this kind of simply connected model cannot characterize biological neural networks, in which there are different weights in synaptic connections. In this paper, we present a neural network model with weighted small-world connections and further investigate the stability of this model.

Biophysical Phenomena↗
PubMedDate not supplied

Early lexical development in a self-organizing neural network.

In this paper we present a self-organizing neural network model of early lexical development called DevLex. The network consists of two self-organizing maps (a growing semantic map and a growing phonological map) that are connected via associative links trained by Hebbian learning. The model captures a number of important phenomena that occur in early lexical acquisition by children, as it allows for the representation of a dynamically changing linguistic environment in language learning. In our simulations, DevLex develops topographically organized representations for linguistic categories over time, models lexical confusion as a function of word density and semantic similarity, and shows age-of-acquisition effects in the course of learning a growing lexicon. These results match up with patterns from empirical research on lexical development, and have significant implications for models of language acquisition based on self-organizing neural networks.

Artificial Intelligence↗
PubMed2001

Fractal encoding in a chaotic neural network.

We analyze a model of a chaotic neural network consisting of three neurons, namely a chaotically forcing neuron and two neurons comprizing a stable response system with a contraction mapping property, for digital encoding with chaotic dynamics. We show that dynamics of the chaotically forcing neuron is embedded in the form of a code sequence on a fractal attractor of the two-neuron response system. We consider the relation between the state transition of the chaotically forcing neuron and the hierarchical fractal structure on the attractor in the state space of the contracting system. We also report hardware implementation of the presented model with an analog electronic circuit to investigate the fractal attractor of the chaotic neural network as a realistic system.

Journal Article↗
PubMed2001

Neural network and wavelet recognition of facial electromyographic signals.

The present aim was to explore the possibilities of using neural networks for recognizing significant changes in electrical activity of human facial muscles. We used multilayer perceptron neural networks to recognize bursts of electromyographic signals recorded with bipolar surface electrodes from two subject's facial muscles. Wavelets were applied for the detection of high frequency components of electromyographic signals. Coefficients of wavelets were used as an input to a neural network in order to differentiate bursts from the signals. The results showed that the recognition of bursts was very successful resulting to 84-97 percent total accuracies. The results were very encouraging and suggest further that the measurement of facial muscle activity may be a potentially useful computer input signal, for example, for affective computing which can be seen as a future versatile interaction between the computer and the user.

Electromyography↗
PubMed1996

Genetic neural networks for quantitative structure-activity relationships: improvements and application of benzodiazepine affinity for benzodiazepine/GABAA receptors.

A novel tool, called a genetic neural network (GNN), has been developed for obtaining quantitative structure-activity relationships (QSAR) for high-dimensional data sets (J. Med. Chem. 1996, 39, 1521-1530). The GNN method uses a neural network to correlate activity with descriptors that are preselected by a genetic algorithm. To provide an extended test of the GNN method, the data on 57 benzodiazepines given by Maddalena and Johnston (MJ; J. Med. Chem. 1995, 38, 715-724) have been examined with an enhanced version of GNN, and the results are compared with the excellent QSAR of MJ. The problematic steepest descent training has been replaced by the scaled conjugate gradient algorithm. This leads to a substantial gain in performance in both robustness of prediction and speed of computation. The cross-validation GNN simulation and the subsequent run based on an unbiased and more efficient protocol led to the discovery of other 10-descriptor QSARs that are superior to the best model of MJ based on backward elimination selection and neural network training. Results from a series of GNNs with a different number of inputs showed that a neural network with fewer inputs can produce QSARs as good as or even better than those with higher dimensions. The top-ranking models from a GNN simulation using only six input descriptors are presented, and the chemical significance of the chosen descriptors is discussed. The statistical significance of these GNN QSARs is validated. The best QSARs are used to provide a graphical tool that aids the design of new drug analogues. By replacing functional groups at the 7- and 2'-positions with ones that have optimal substituent parameters, a number of new benzodiazepines with high potency are predicted.

Algorithms↗
PubMed1992

An artificial neural network for lesion detection on single-photon emission computed tomographic images.

RATIONALE AND OBJECTIVES: An artificial neural network (ANN) has been developed to detect nonactive circular lesions on single-slice, single-photon emission computed tomographic (SPECT) images reconstructed using filtered back projection (FBP). METHODS: The neural network is a single-layer perception which learns to identify features on the SPECT image using supervised training with a modified delta rule. The network was trained on a set of SPECT images containing clinically realistic levels of noise. The trained network was applied to a set of 120 images, and the detection performance was evaluated at several decision thresholds using receiver operating characteristic (ROC) analysis. RESULTS: The trained neural network performed better than human observers for the same detection task with the same images as reflected by a significantly larger ROC curve area. CONCLUSIONS: ANN can be trained successfully to perform lesion detection on reconstructed SPECT images.

Equipment Design↗
PubMed1997

Application of image analysis and neural networks to the pathology diagnosis of intraductal proliferative lesions of the breast.

We studied whether a computer-assisted system using a combination of data collection by image analysis and analysis by neural networks can differentiate benign and malignant breast lesions. Forty-six intraductal lesions of the breast were studied by pathologists and by the computer-assisted system. Histological evaluation was performed independently by three pathologists, and the lesions were classified into pathologically malignant (n = 12), undetermined (n = 13), and benign (n = 21). Computerized nuclear image analysis was performed using the CAS200 (Cell Analysis Systems, Elmhurst, IL) system to obtain data on nuclear morphometric and textural features. A neural network was constructed using the morphometric and texture data obtained from teaching cases of malignant and benign lesions. Then data for unknown cases were classified by the constructed neural network into neural network-malignant (n = 11), -undetermined (n = 5), and -benign (n = 30). The agreement rate between the diagnosis by pathologists and judgment by the computer-assisted system was 75%, excluding pathologically undetermined lesions. There were four false-negative but no false-positive results. False-negative cases had nuclei that were quite different from those of the teaching cases. The agreement rate obtained using either morphometric data or texture data only was lower than that using a combination of both. Selection of appropriate teaching data and incorporation of both morphometric and textural parameters seemed important for obtaining more accurate results. The present data suggest that development of a computer-assisted histopathological diagnosis system for practical use may be possible.

Breast↗
PubMed1999

Improving neural network prediction of amino acid levels in feed ingredients.

Artificial neural networks (ANN) were trained to predict the amino acid (AA) profile of feed ingredients. The ANN more effectively identified the complex relationship between nutrients and feed ingredients than linear regression (LR). Three types of ANN (NeuroShell 2): three-layer backpropagation (BP3), Ward Backpropagation (WBP), a general regression neural network (GRNN); and LR (SAS Proc GLM) were used to predict the AA level in corn, soybean meal, meat and bone meal, fish meal, and wheat based on proximate analysis. In contrast to a past study, a variety of alternative ANN training parameters were examined to improve ANN performance. Predictive performance was judged on the basis of the maximum R2 value resulting from all defaults tested. Advanced selection of ANN training parameters led to further improvement in performance, especially within the GRNN architecture. In 34 of 35 ANN developed, the maximum R2 value for each individual AA in each feed ingredient was higher for GRNN than for LR, BP3, or WBP prediction methods. For example, the highest R2 value for Met in corn was 0.32 for LR, 0.40 for 3LBP, 0.51 for WBP, and 0.95 for GRNN analysis. Predictive performance was also improved overall as compared to results of a previous study. For example, corn maximum R2 values (GRNN) for Met, TSAA, and Trp were: 0.78, 0.81 and 0.44, previously, and 0.95, 0.96 and 0.88, in the current study. Current soybean meal maximum R values (GRNN) were: Met, 0.92; TSAA, 0.94; and Lys, 0.90. Current meat and bone mean maximum R2 values (GRNN) were: Met, 0.97; TSAA, 0.97; and Lys, 0.97. The ANN computation is a successful alternative to statistical regression analysis for predicting AA levels in feed ingredients.

Amino Acids↗
PubMed2006

A neural network model for combining clinical predictors of antidepressant response in mood disorders.

Artificial neural networks (ANN) represent a promising tool for combining multiple predictors in complex diseases. Antidepressant response in mood disorders is a typical complex phenomenon were a number of predictors influence outcome under non-linear interactions. In the present study we tested a neural network strategy for antidepressant outcome in subjects affected by major depression. One hundred and forty-five never reported depressed inpatients were included in this study (major depressives/bipolars: 111/34). A multi layer perceptron network composed of 1 hidden layer with 13 nodes was chosen. The network was performed on the sample of 145 cases divided as follows: train 73+verify 36+test 36. Correlation of predicted versus observed response was 0.46 in the test (independent) sample that corresponds to 21% of variance explained. Number of episodes, side effects, delusional features, baseline HAM-D, length of current episode, lithium augmentation, current medical condition and personality disorders were the main factors identified by the model. Sex, age at onset, polarity, plasma level and baseline VAS score were part of the model but with a lower rank. Overall, our findings suggest that neural networks could be a valid technique for the analysis of psychopharmacology studies. The complex interactions modelled through ANN may be eventually applied at the clinical level for the individualized therapy.

Antipsychotic Agents↗
PubMed2004

Spectroscopic quantification of bacteria using artificial neural networks.

Fourier transform-infrared spectroscopy, in conjunction with artificial neural networks, has been used for identification and classification of selected foodborne pathogens. Five bacterial species (Enterococcus faecium, Salmonella Enteritidis, Bacillus cereus, Yersinia enterocolitica, Shigella boydii) and five Escherichia coli strains (O103, O55, O121, O30, O26) suspended in phosphate-buffered saline were enumerated to provide seven different concentrations ranging from 10(9) to 10(3) CFU/ ml. The trained artificial neural networks were then validated with an independent subset of samples and compared with the traditional plate count method. It was found that the concentration-based classification of the species was 100% correct and the strain-based classification was 90 to 100% accurate.

Bacillus cereus↗
PubMed1997

Application of autonomous neural network systems to medical pattern classification tasks.

This paper presents a study of the application of autonomously learning multiple neural network systems to medical pattern classification tasks. In our earlier work, a hybrid neural network architecture has been developed for on-line learning and probability estimation tasks. The network has been shown to be capable of asymptotically achieving the Bayes optimal classification rates, on-line, in a number of benchmark classification experiments. In the context of pattern classification, however, the concept of multiple classifier systems has been proposed to improve the performance of a single classifier. Thus, three decision combination algorithms have been implemented to produce a multiple neural network classifier system. Here the applicability of the system is assessed using patient records in two medical domains. The first task is the prognosis of patients admitted to coronary care units; whereas the second is the prediction of survival in trauma patients. The results are compared with those from logistic regression models, and implications of the system as a useful clinical diagnostic tool are discussed.

Artificial Intelligence↗
PubMed2006

The use of neural networks in evaluation of the direction and dynamics of changes in lipid parameters in kidney transplant patients on the Mediterranean diet.

OBJECTIVE: The objective of the study was to assess whether neural networks can be a tool useful in the evaluation of the effect of the Mediterranean diet (MD) on the direction and dynamics of selected parameters. DESIGN: Randomized, prospective study. SETTING: Outpatient Clinic of the Department of Nephrology, Transplantology, and Internal Medicine. PATIENTS AND INTERVENTION: The study group consisted of 21 patients after kidney transplantation whose diet complied with the MD; the control group included 16 patients (also after transplantation) on a low-fat diet, isocaloric with the study diet. MAIN OUTCOME MEASURES: Anthropometry, plasma lipids, chromatography of triacylglycerols and fatty acids, and activity of superoxide dismutase and catalase were measured in both groups. Statistical analysis was done with the SNN (Statistica Neural Networks) StatSoft software package. RESULTS: The advantage of neural networks is the possibility of the dynamic presentation of a process taking place in a biological system. In the MD group in the first months of use of the diet, the cholesterol level was reduced only in the group of young and middle-aged patients. This tendency was not observed among elderly patients, among whom a small reduction of the total cholesterol level was noted only at the end of the observation period. In control group at the beginning of the observation, the plasma total cholesterol level was proportional to the patient's age. After 6 months, the total cholesterol increased in young patients and redacted in the group of elderly patients. CONCLUSIONS: We concluded that the MD diet would be ideal for posttransplantation patients without serious pathologic dyslipidemia. In the case of patients with substantial dyslipidemia, appropriate pharmacologic treatment lowering proatherosclerotic lipid levels should be used in combination with the MD. Artificial neural networks (ANNs) were a useful tool in modeling biological parameters, showing dynamics of the studied interactions in a very detailed way. ANN is the most suitable method for investigations with many variables, interconnected nonlinearly; therefore, this method allows for a more general approach to biological problems. However, it should be noted that considerable data sets are required to obtain a satisfactory fit to the data. Moreover, to ensure the predictive power of this method for new cases, the representative database is indispensable. In spite of these demands, ANN is a prospective tool for reliable, quick assessments and predictions.

Adult↗
PubMed1993

Modeling mortality in the intensive care unit: comparing the performance of a back-propagation, associative-learning neural network with multivariate logistic regression.

The objective of this study was to compare and contrast two techniques of modeling mortality in a 30 bed multi-disciplinary ICU; neural networks and logistic regression. Fifteen physiological variables were recorded on day 3 for 422 consecutive patients whose duration of stay was over 72 hours. Two separate models were built using each technique. First, logistic and neural network models were constructed on the complete 422 patient dataset and discrimination was compared. Second, the database was randomly divided into a 284 patient developmental dataset and a 138 patient validation dataset. The developmental dataset was used to construct logistic and neural net models and the predictive power of these models was verified on the validation dataset. On the complete dataset, the neural network clearly outperformed the logistic model (sensitivity and specificity of 1 and .997 vs. .525 and .966, area under ROC curve .9993 vs. .9259), while both performed equally well on the validation dataset (area under ROC of .82). The excellent performance of the neural net on the complete dataset reveals that the problem is classifiable. Since our dataset only contained 40 mortality events, it is highly likely that the validation dataset was not representative of the developmental dataset, which led to a decreased predictive performance by both the neural net and the logistic regression models. Theoretically, given an extensive dataset, the neural network should be able to perform mortality prediction with a sensitivity and a specificity approaching 95%. Clinically, this would be an extremely important achievement.(ABSTRACT TRUNCATED AT 250 WORDS)

Hospital Mortality↗
PubMedDate not supplied

Improved backpropagation learning in neural networks with windowed momentum.

Backpropagation, which is frequently used in Neural Network training, often takes a great deal of time to converge on an acceptable solution. Momentum is a standard technique that is used to speed up convergence and maintain generalization performance. In this paper we present the Windowed Momentum algorithm, which increases speedup over Standard Momentum. Windowed Momentum is designed to use a fixed width history of recent weight updates for each connection in a neural network. By using this additional information, Windowed Momentum gives significant speedup over a set of applications with same or improved accuracy. Windowed Momentum achieved an average speedup of 32% in convergence time on 15 data sets, including a large OCR data set with over 500,000 samples. In addition to this speedup, we present the consequences of sample presentation order. We show that Windowed Momentum is able to overcome these effects that can occur with poor presentation order and still maintain its speedup advantages.

Algorithms↗
PubMed2005

Noise-enhanced categorization in a recurrently reconnected neural network.

We investigate the interplay of recurrence and noise in neural networks trained to categorize spatial patterns of neural activity. We develop the following procedure to demonstrate how, in the presence of noise, the introduction of recurrence permits to significantly extend and homogenize the operating range of a feed-forward neural network. We first train a two-level perceptron in the absence of noise. Following training, we identify the input and output units of the feed-forward network, and thus convert it into a two-layer recurrent network. We show that the performance of the reconnected network has features reminiscent of nondynamic stochastic resonance: the addition of noise enables the network to correctly categorize stimuli of subthreshold strength, with optimal noise magnitude significantly exceeding the stimulus strength. We characterize the dynamics leading to this effect and contrast it to the behavior of a more simple associative memory network in which noise-mediated categorization fails.

Journal Article↗
PubMed1999

Comparison of a neural network approach with five traditional methods for predicting creatinine clearance in patients with human immunodeficiency virus infection.

STUDY OBJECTIVE: To compare the results of an artificial neural network approach with those of five published creatinine clearance (Cl(cr)) prediction equations and with the measured (true) Cl(cr) in patients infected with the human immunodeficiency virus (HIV). DESIGN: Six-month prospective study. SETTINGS: Two university medical centers. PATIENTS: Sixty-five HIV-infected patients: 18 relatively healthy outpatients and 47 inpatients. INTERVENTIONS: All subjects had urine collected for 24 hours to determine Cl(cr). MEASUREMENTS AND MAIN RESULTS: The 16 input variables were age, ideal body weight, actual body weight, body surface area, height, and the following blood chemistries: sodium, potassium, aspartate aminotransferase, alanine aminotransferase, red blood cell count, platelet count, white blood cell count, glucose, serum creatinine, blood urea nitrogen, and albumin. The only output variable was Cl(cr). A training set of 55 subjects was used to develop the relationship between input variables and the output variable. The trained neural network was then used to predict Cl(cr) of a validation set of 10 subjects. Mean differences between predicted Cl(cr) and actual Cl(cr) (bias) were 4.1, 28.7, 29.4, 26.0, 31.8, and 55.8 ml/min/1.73 m2 for the artificial neural network, Cockcroft and Gault, Jelliffe 1, Jelliffe 2, Mawer et al, and Hull et al methods, respectively. CONCLUSION: The accuracy of predicting Cl(cr) in subjects with HIV infection by the artificial neural network is superior to that of the five equations that are currently used in clinical settings.

Adult↗
PubMed2002

Determination of the elastic constants of a composite plate using wavelet transforms and neural networks.

An inverse method based on a combination of the wavelet transform and artificial neural networks is presented. The method is used to recover the elastic constants of a fiber-reinforced composite plate from experimental measurements of ultrasonic Lamb waves generated and detected with lasers. In this method, the elastic constants are not recovered from the dispersion curves but rather directly from the measured waveforms. Transient waveforms obtained by numerical simulations for different elastic constants are used as input to train the neural network. The wavelet transform is used to extract the eigenvectors from the Lamb wave signals to simplify the structure of the neutral network. The eigenvectors are then introduced into a multilayer internally recurrent neural network with a back-propagation algorithm. Finally, experimental waveforms recoded on a titanium-graphite composite plate are used as input to recover the elastic constants of the material.

Journal Article↗