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 1,045 records · Page 58Linked to original sources
PubMed2005

Application of serum protein fingerprinting coupled with artificial neural network model in diagnosis of hepatocellular carcinoma.

BACKGROUND: Hepatocellular carcinoma tends to present at a late clinical stage with poor prognosis. Therefore, it is urgent to explore and develop a simple, rapid diagnostic method, which has high sensitivity and specificity for hepatocellular carcinoma at an early stage. In this study, the serum proteins in patients with hepatocellular carcinoma or liver cirrhosis and in normal controls were analysed. Surface enhanced laser desorption/ionization time-of-flight mass (SELDI-TOF-MS) spectrometry was used to fingerprint serum protein using the protein chip technique and explore the value of the fingerprint, coupled with artificial neural network, to diagnose hepatocellular carcinoma. METHODS: Of the 106 serum samples obtained, 52 were from patients with hepatocellular carcinoma, 22 from patients with liver cirrhosis and 32 from healthy volunteers. The samples were randomly assigned into a training group (n = 70, 35 patients with hepatocellular carcinoma, 14 with liver cirrhosis, and 21 normal controls) and a testing group (n = 36, 17 patients with hepatocellular carcinoma, 8 with liver cirrhosis, and 11 normal controls). An artificial neural network was trained on data from 70 individuals in the training group to develop an artificial neural network diagnostic model and this model was tested. The 36 sera in the testing group were analysed with blind prediction by using the same flowchart and procedure of data collection. The 36 serum protein spectra were clustered with the preset clustering method and the same mass/charge (M/Z) peak values as those in the training group. Matrix transfer was performed after data were output. Then the data were input into the previously built artificial neural network model to get the prediction value. The M/Z peaks of the samples with more than 2000 M/Z were normalized with biomarker wizard of ProteinChip Software version 3.1 for noise filtering. The first threshold for noise filtering was set at 5, and the second was set at 2. The 10% was the minimum threshold for clustering. The statistical analysis of the data of serum protein mass spectrum was performed in the groups (normal vs. hepatocellular carcinoma, and liver cirrhosis vs. hepatocellular carcinoma) with the t test. RESULTS: Comparison between the groups of hepatocellular carcinoma and normal control: The mass spectra from 56 samples (hepatocellular carcinoma and normal controls) in the training group were analysed and 241 peaks were obtained. In addition, 21 peaks from them were used for comparison between the groups of hepatocellular carcinoma and normal controls (P < 0.01). Only 2 peaks at 3015 M/Z and 5900 M/Z were selected with significant difference (P < 10 (-9)). A model was developed based on these two proteins with different M/Z. It was confirmed that this artificial neural network model can be used for comparison between the groups of hepatocellular carcinoma and normal controls. The sensitivity was 100% (17/17), and the specificity was 100% (11/11). Comparison between the groups of hepatocellular carcinoma and liver cirrhosis: The mass spectra from 49 samples in the training group (including patients with hepatocellular carcinoma and liver cirrhosis) were analysed and 208 peaks were obtained. In addition, 21 peaks from them were used for comparison between the groups of hepatocellular carcinoma and liver cirrhosis (P < 0.01). Only 2 peaks at 7759 M/Z, 13134 M/Z were selected with significant difference (P < 10 (-9)). A model was developed based on these two proteins with different M/Z. It was confirmed that this artificial neural network model can be used for comparison between the groups of hepatocellular carcinoma and liver cirrhosis. The sensitivity was 88.2% (15/17), and the specificity was 100% (8/8). CONCLUSIONS: The specific biomarkers selected with the SELDI technology could be used for early diagnosis of hepatocellular carcinoma.

Adult↗
PubMed2003

Inductive bias strength in knowledge-based neural networks: application to magnetic resonance spectroscopy of breast tissues.

The integration of symbolic knowledge with artificial neural networks is becoming an increasingly popular paradigm for solving real-world applications. The paradigm provides means for using prior knowledge to determine the network architecture, to program a subset of weights to induce a learning bias which guide network training, and to extract knowledge from trained networks. The role of neural networks then becomes that of knowledge refinement. It thus provides a methodology for dealing with uncertainty in the prior knowledge. We address the open question of how to determine the strength of the inductive bias of programmed weights; we present a quantitative solution which takes the network architecture, the prior knowledge, and the training data into consideration. We apply our solution to the difficult problem of analyzing breast tissue from magnetic resonance spectroscopy (MRS); the available database is extremely limited and cannot be adequately explained by expert knowledge alone.

Adult↗
PubMed1994

Visual field analysis using artificial neural networks.

There have been several reports on the application of artificial neural networks (ANNs) to visual field classification. While these have demonstrated that neural networks can be used with good results they have not explored the effects that the training set can have upon network performance nor emphasized the unique value of ANNs in visual field analysis. This paper considers the problem of differentiating normal and glaucomatous visual fields and explores different training set characteristics using field data collected from a Henson CFS2000 perimeter. Training set properties including size, balance between normals and glaucomas, extent of field loss and the spatial location of glaucomatous defects are explored. A multilayer network with 132 input nodes, 20 hidden layer nodes and 2 output nodes in trained using an error backpropagation algorithm. Both sensitivity and specificity are measured during testing. The results demonstrate that large random sets are better than small random sets since sensitivity improves with size and specificity is not adversely affected. The variability in performance also reduces as training set size increases. In addition, sets that are biased towards glaucoma examples are more sensitive and less specific, while sets biased with normal examples are more specific and less sensitive than balanced sets. Thus large training sets with class balance are generally desirable for good sensitivities and specificities. The actual glaucoma examples contained in the set are also important. A training set deficient in examples has no detrimental effect on sensitivity or specificity. The spatial distribution of defects is also crucial. Spatially biased sets are unable to recognize defects that occur in locations where no previous defect has been presented while more balanced sets lead to better performance. In conclusion the 'ideal' training set should contain many examples of early defects that represent the full range of locations where these defects may occur.

Algorithms↗
PubMed2004

Melanoma diagnosis by Raman spectroscopy and neural networks: structure alterations in proteins and lipids in intact cancer tissue.

Melanoma is the most aggressive skin cancer. The specificity and sensitivity of clinical diagnosis varies from around 40% to 80%. Here, we investigated whether the chemical changes in the melanoma tissue detected by Raman spectroscopy and neural networks can be used for diagnostic purposes. Near-infrared Fourier transform Raman spectra were obtained from samples of melanoma (n=22) and other skin tumors that can be clinically confused with melanoma: pigmented nevi (n=41), basal cell carcinoma (n=48), seborrheic keratoses (n=23), and normal skin (n=89). A sensitivity analysis of spectral frequencies used by a neural network was performed to determine the importance of the individual components in the Raman spectra. Visual inspection of the Raman spectra suggested that melanoma could be differentiated from pigmented nevi, basal cell carcinoma, seborrheic keratoses, and normal skin due to the decrease in the intensity of the amide I protein band around 1660 cm-1. Moreover, melanoma and basal cell carcinoma showed an increase in the intensity of the lipid-specific band peaks around 1310 cm-1 and 1330 cm-1, respectively. Band alterations used in the visual inspection were also independently identified by a neural network for melanoma diagnosis. The sensitivity and specificity for diagnosis of melanoma achieved by neural network analysis of Raman spectra were 85% and 99%, respectively. We propose that neural network analysis of near-infrared Fourier transform Raman spectra could provide a novel method for rapid, automated skin cancer diagnosis on unstained skin samples.

Diagnosis, Differential↗
PubMed2002

PatPho: a phonological pattern generator for neural networks.

Much of the power of neural network modeling for language use and acquisition derives from a reliance on statistical regularities implicit in the phonological properties of words. Researchers have devised several methods for representing the phonology of words, but these methods are often either unable to represent realistically sized lexicons or inadequate in the ways they represent individual words. In this paper, we present a new phonological pattern generator (PatPho) that allows connectionist modelers to derive accurate phonological representations of the English lexicon. PatPho not only generates phonological patterns that can scale up to realistically sized lexicons, but also accurately and parsimoniously captures the similarity structures of the phonology of monosyllabic and multisyllabic words.

Humans↗
PubMed2004

Prediction of biological targets using probabilistic neural networks and atom-type descriptors.

Prediction of biological targets for molecules from their chemical structures is beneficial for generating focused libraries, selecting compounds for screening, and annotating biological activities for those compounds whose activities are unknown. We studied the ability of a probabilistic neural network (PNN), a variant of normalized radial basis function (RBF) neural networks, to predict biological activities for a set of 799 compounds having activities against seven biological targets. The compounds were taken from the MDDR database, and they were carefully selected to comprise distinct biological activities and diverse structures. The structural characteristics of compounds were represented by a set of 24 atom-type descriptors defined by 2D topological chemical structures. The modeling was done in two ways: (1). compounds having one certain activity were discriminated from those not having that activity and (2). all compounds were classified into seven biological classes. In both cases, around 90% of the compounds were correctly classified. Further validation of the modeled PNNs was done with 26 317 compounds having biological activities against various targets except for the seven targets used for modeling, and 67-98% compounds were correctly classified depending upon the targets. A PNN trains much more quickly than widely used neural networks such as a feed-forward neural network with error back-propagation. Calculation of atom-type descriptors is easy even for a large-size chemical library. Combination of PNN and atom-type descriptors thus provides a powerful way to predict biological activities from structural information.

Angiotensin-Converting Enzyme Inhibitors↗
PubMed1992

Sequences of smoothly correlated patterns in neural networks with random transmission delays.

A neural network with a broad distribution of transmission delays was used to study numerically the retrieval of sequences having several types of correlations between successive patterns. In the case of sequences consisting of patterns correlated for finite time, the quality of retrieval was found to be (more or less) independent of the pattern correlation width and the delay distribution. On the other hand the quality of retrieval is dependent on these factors in the case of sequences with pattern correlation functions with long time 'tails'. Finally we have studied to what extent the storage capacity depends on the pattern correlation function and the delay distribution.

Animals↗
PubMed1999

Determining the geographic origin of potatoes with trace metal analysis using statistical and neural network classifiers.

The objective of this research was to develop a method to confirm the geographical authenticity of Idaho-labeled potatoes as Idaho-grown potatoes. Elemental analysis (K, Mg, Ca, Sr, Ba, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Mo, S, Cd, Pb, and P) of potato samples was performed using ICPAES. Six hundred eight potato samples were collected from known geographic growing sites in the U.S. and Canada. An exhaustive computational evaluation of the 608 x 18 data sets was carried out using statistical (PCA, CDA, discriminant function analysis, and k-nearest neighbors) and neural network techniques. The neural network classification of the samples into two geographic regions (defined as Idaho and non-Idaho) using a bagging technique had the highest percentage of correct classifications, with a nearly 100% degree of accuracy. We report the development of a method combining elemental analysis and neural network classification that may be widely applied to the determination of the geographical origin of unprocessed, fresh commodities.

Analysis of Variance↗
PubMed2000

A parallel algorithm for simulation of large neural networks.

The simulation of biologically realistic neural networks requires the numerical solution of very large systems of differential equations. Variables within the system can be changing at rates that vary by orders of magnitude, not only at different times of the solution, but at the same time in different parts of the network. Therefore, an efficient implementation must be able to vary the solution step size, and do so independently in different subsystems. A single processor algorithm is presented in which each neuron can be solved with its own step size by using a priority queue to integrate them in the correct order. But this leaves the problem of how communication and synchronisation between neurons should be managed when executing in parallel. The proposed solution uses an algorithm based on waveform relaxation, which allows groups of neurons on different processors to be solved independently and hence in parallel, for substantial parts of the computation. Realistic test problems were run on a distributed memory parallel computer and results show that speedups of 10 using 16 processors are achievable, and indicate that further speedups may be possible.

Algorithms↗
PubMed2000

Stability problems with artificial neural networks and the ensemble solution.

Artificial neural networks (ANNs) are very popular as classification or regression mechanisms in medical decision support systems despite the fact that they are unstable predictors. This instability means that small changes in the training data used to build the model (i.e. train the ANN) may result in very different models. A central implication of this is that different sets of training data may produce models with very different generalisation accuracies. In this paper, we show in detail how this can happen in a prediction system for use in in-vitro fertilisation. We argue that claims for the generalisation performance of ANNs used in such a scenario should only be based on k-fold cross-validation tests. We also show how the accuracy of such a predictor can be improved by aggregating the output of several predictors.

Algorithms↗
PubMed2000

The Bayesian evidence scheme for regularizing probability-density estimating neural networks.

Training probability-density estimating neural networks with the expectation-maximization (EM) algorithm aims to maximize the likelihood of the training set and therefore leads to overfitting for sparse data. In this article, a regularization method for mixture models with generalized linear kernel centers is proposed, which adopts the Bayesian evidence approach and optimizes the hyperparameters of the prior by type II maximum likelihood. This includes a marginalization over the parameters, which is done by Laplace approximation and requires the derivation of the Hessian of the log-likelihood function. The incorporation of this approach into the standard training scheme leads to a modified form of the EM algorithm, which includes a regularization term and adapts the hyperparameters on-line after each EM cycle. The article presents applications of this scheme to classification problems, the prediction of stochastic time series, and latent space models.

Acquired Immunodeficiency Syndrome↗
PubMed1991

A neural network model of the cerebellar cortex performing dynamic associations.

The present paper proposes a model which applies formal neural network modeling techniques to construct a theoretical representation of the cerebellar cortex and its performances in motor control. A schema that makes explicit use of propagation delays of neural signals, is introduced to describe the ability to store temporal sequences of patterns in the Golgi-granule cell system. A perceptron association is then performed on these sequences of patterns by the Purkinje cell layer. The model conforms with important biological constraints, such as the known excitatory or inhibitory nature of the various synapses. Also, as suggested by experimental evidence, the synaptic plasticity underlying the learning ability of the model, is confined to the parallel fiber--Purkinje cell synapses, and takes place under the control of the climbing fibers. The result is a neural network model, constructed according to the anatomy of the cerebellar cortex, and capable of learning and retrieval of temporal sequences of patterns. It provides a framework to represent and interpret properties of learning and control of movements by the cerebellum, and to assess the capacity of formal neural network techniques for modeling of real neural systems.

Animals↗
PubMed2002

A neural network approach to treatment optimization.

Typical medical diagnosis applications of neural networks for prediction and classification require training data (observations) that include the "correct" category for a number of patient records. In this paper, we borrow a technique from control systems applications of neural networks. Optimal control parameters of a system are typically not known. Instead, we only know the effect on a remote system. The correct control action drives the remote system optimally. The learning technique requires two networks: one to model the system to be controlled (here, the patient), and one to optimize the treatment (here, the treating physician). The concept was tested with artificially generated noisy data, and gives promising results.

Drug Therapy, Computer-Assisted↗
PubMed1991

Altering the synchrony of stimulus trace processes: tests of a neural-network model.

A previously described neural-network model (Desmond 1991; Desmond and Moore 1988; Moore et al. 1989) predicts that both CS-onset-evoked and CS-offset-evoked stimulus trace processes acquire associative strength during classical conditioning, and that CR waveforms can be altered by manipulating the time at which the processes are activated. In a trace conditioning paradigm, where CS offset precedes US onset, the model predicts that onset and offset traces act in synchrony to generate unimodal CR waveforms. However, if the CS duration is subsequently lengthened on CS-alone probe trials, the model predicts that onset and offset traces will asynchronously contribute to CR output and bimodal CRs will be generated. In a delay conditioning paradigm, in which US onset occurs prior to CS offset, the model predicts that only the onset process will gain associative strength, and hence, only unimodal CRs will occur. Using the rabbit conditioned nictitating membrane response preparation, we found experimental support for these predictions.

Analysis of Variance↗
PubMed1999

Pre-synaptic lateral inhibition provides a better architecture for self-organizing neural networks.

Unsupervised learning is an important ability of the brain and of many artificial neural networks. A large variety of unsupervised learning algorithms have been proposed. This paper takes a different approach in considering the architecture of the neural network rather than the learning algorithm. It is shown that a self-organizing neural network architecture using pre-synaptic lateral inhibition enables a single learning algorithm to find distributed, local, and topological representations as appropriate to the structure of the input data received. It is argued that such an architecture not only has computational advantages but is a better model of cortical self-organization.

Algorithms↗
PubMedDate not supplied

Simultaneous determination of thiocyanate and salicylate by a combined UV-spectrophotometric detection principal component artificial neural network.

A modified principle component artificial neural network (PC-ANN) model is developed for simultaneous determination of thiocyanate and salycilate concentration after passing through the bulk of a liquid membrane by tri-phenyl benzyl phosphonium chloride. All calibration, and test samples data were obtained using UV-Vis spectrophotometer. In this way, a modified PC-ANN consisting of three layers of nodes was trained by combination of Bayesian-Levenberg-Marquardt as training rule. Sigmoid and liner transfer functions were used in the hidden and output layers respectively to facilitate nonlinear calibration. The model could accurately estimate the concentration of components with acceptable precision and accuracy, for mixtures. The PC-ANN model exhibits a good ability for the simultaneous determination of the thiocyanate and salycilate in concentration range 0.5 x 10(-4) mol.l(-1) up to 5.0 x 10(-4) mol.l(-1) with Root Mean square error (2.22% and 2.20%, for thiocyanate and salycilate, respectively) and high correlation coefficients (R2= 0.998 or greater). Results obtained with modified trained PC-ANN were compared with stepwise linear regression (SMLR) model. Validation of the two models shows a better ability in estimation of the modified PC-ANN as compared with the SMLR model (MSRE given are 3.12%, 6.31%.).

Neural Networks, Computer↗
PubMed2003

Sensitive quantitative predictions of peptide-MHC binding by a 'Query by Committee' artificial neural network approach.

We have generated Artificial Neural Networks (ANN) capable of performing sensitive, quantitative predictions of peptide binding to the MHC class I molecule, HLA-A*0204. We have shown that such quantitative ANN are superior to conventional classification ANN, that have been trained to predict binding vs non-binding peptides. Furthermore, quantitative ANN allowed a straightforward application of a 'Query by Committee' (QBC) principle whereby particularly information-rich peptides could be identified and subsequently tested experimentally. Iterative training based on QBC-selected peptides considerably increased the sensitivity without compromising the efficiency of the prediction. This suggests a general, rational and unbiased approach to the development of high quality predictions of epitopes restricted to this and other HLA molecules. Due to their quantitative nature, such predictions will cover a wide range of MHC-binding affinities of immunological interest, and they can be readily integrated with predictions of other events involved in generating immunogenic epitopes. These predictions have the capacity to perform rapid proteome-wide searches for epitopes. Finally, it is an example of an iterative feedback loop whereby advanced, computational bioinformatics optimize experimental strategy, and vice versa.

HLA-A Antigens↗
PubMed1997

Basic concepts of artificial neural networks (ANN) modeling in the application to pharmaceutical development.

Artificial neural networks (ANN) methodology is a new modeling method that has not been broadly applied to pharmaceutical sciences up to now. The aim of this paper is to give a detailed description of the associating networks as well as a description of less well-known networks (i.e., feature-extracting and nonadaptive networks) and their scope of application in pharmaceutical sciences. The descriptions include the historical origin and the basic concepts behind the computing. ANN are based on the attempt to model the neural networks of the brain. Learning algorithms for associating ANN use mathematical procedures usually derived from the gradient descent method whereas feature-extracting ANN map multidimensional input data sets onto two-dimensional spaces. Nonadaptive ANN map data sets and are able to reconstruct their patterns when presented with corrupted or noisy samples. Associating networks can typically be applied in the pharmaceutical field as an alternative to traditional response surface methodology, feature-extracting networks as alternative to principal component analysis, and nonadaptive networks for image recognition. Based on these abilities, the potential application fields of the ANN methodology in the pharmaceutical sciences is broad, ranging from clinical pharmacy through biopharmacy, drug and dosage form design, to interpretation of analytical data. The few applications presented in the pharmaceutical technology area seem promising and should be investigated in more detail.

Algorithms↗