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 811 records · Page 45Linked to original sources
PubMed2002

Global exponential stability of continuous-time interval neural networks.

This paper addresses global robust stability of a class of continuous-time interval neural networks that contain time-invariant uncertain parameters with their values being unknown but bounded in given compact sets. We first introduce the concept of diagonally constrained interval neural networks and present a necessary and sufficient condition for global exponential stability of these interval neural networks irregardless of any bounds of non-diagonal uncertain parameters in connection weight matrices. Then we extend the robust stability result to general interval neural networks by giving a sufficient condition. Simulation results illustrate the characteristics of the main results.

Journal Article↗
PubMed2006

Piecewise-linear neural networks and their relationship to rule extraction from data.

This article addresses the topic of extracting logical rules from data by means of artificial neural networks. The approach based on piecewise linear neural networks is revisited, which has already been used for the extraction of Boolean rules in the past, and it is shown that this approach can be important also for the extraction of fuzzy rules. Two important theoretical properties of piecewise-linear neural networks are proved, allowing an elaboration of the basic ideas of the approach into several variants of an algorithm for the extraction of Boolean rules. That algorithm has already been used in two real-world applications. Finally, a connection to the extraction of rules of the Łukasiewicz logic is established, relying on recent results about rational McNaughton functions. Based on one of the constructive proofs of the McNaughton theorem, an algorithm is formulated that in principle allows extracting a particular kind of formulas of the Łukasiewicz predicate logic from piecewise-linear neural networks trained with rational data.

Algorithms↗
PubMed2003

HIV lipodystrophy case definition using artificial neural network modelling.

OBJECTIVE: A case definition of HIV lipodystrophy has recently been developed from a combination of clinical, metabolic and imaging/body composition variables using logistic regression methods. We aimed to evaluate whether artificial neural networks could improve the diagnostic accuracy. METHODS: The database of the case-control Lipodystrophy Case Definition Study was split into 504 subjects (265 with and 239 without lipodystrophy) used for training and 284 independent subjects (152 with and 132 without lipodystrophy) used for validation. Back-propagation neural networks with one or two middle layers were trained and validated. Results were compared against logistic regression models using the same information. RESULTS: Neural networks using clinical variables only (41 items) achieved consistently superior performance than logistic regression in terms of specificity, overall accuracy and area under the ROC curve. Their average sensitivity and specificity were 72.4 and 71.2%, as compared with 73.0 and 62.9% for logistic regression, respectively (area under the ROC curve, 0.784 vs 0.748). The discriminating performance of the neural networks was largely unaffected when built excluding 13 parameters that patients may not have readily available. The average sensitivity and specificity of the neural networks remained the same when metabolic variables were also considered (total 60 items) without a clear advantage against logistic regression (overall accuracy 71.8%). The performance of networks considering also body composition variables was similar to that of logistic regression (overall accuracy 78.5% for both). CONCLUSIONS: Neural networks may offer a means to improve the discriminating performance for HIV lipodystrophy, when only clinical data are available and a rapid approximate diagnostic decision is needed. In this context, information on metabolic parameters is apparently not helpful in improving the diagnosis of HIV lipodystrophy, unless imaging and body composition studies are also obtained.

Absorptiometry, Photon↗
PubMed1991

Neural network analysis of serial cardiac enzyme data. A clinical application of artificial machine intelligence.

There has been a recent resurgence of interest in the study and application of computerized neural networks within the broad field of artificial intelligence. These "intelligent machines" are modeled after biological nervous systems and are fundamentally different from the many computerized expert systems that previously have been introduced as clinical decision-making aids. The authors describe a neural network designed and trained to predict the probability of acute myocardial infarction (AMI) based on the analysis of paired sets of cardiac enzymes. The neural network predicted 24 of 24 (100%) AMIs and 27 of 29 (93%) No-AMIs when compared with a pathologist's interpretation of the patient's laboratory data (P less than 0.000001). The authors attempted to validate the network's diagnoses by two independent methods. When compared with echocardiogram and EKG for diagnosis of AMI, the neural network agreed with the cardiologist's interpretation in 12 of 14 (86%) AMIs and 1 of 3 (33%) No-AMIs, but the correlation was not statistically significant. Using autopsy outcome for validation, the neural network agreed with the anatomic evidence in 24 of 26 (92%) AMIs and 4 of 6 (67%) No-AMIs (P = 0.001). The authors conclude that neural networks can be successfully applied to the analysis of cardiac enzyme data and suggest that broader applications exist within the domain of clinical decision support.

Amyloidosis↗
PubMed1992

Evaluation of a neural-network classifier for PET scans of normal and Alzheimer's disease subjects.

The value of PET as an objective diagnostic tool for dementia may depend on the degree to which abnormal metabolic patterns can be detected by quantitative classification methods. In these studies, a neural-network classifier based on coarse region of interest analyses was used to classify normal and abnormal FDG-PET scans. The performance of neural networks and of an expert reader were evaluated by cross-validation testing. When the "abnormal" class was represented by subjects with clinical diagnoses of "Probable Alzheimer's," the areas under the relative-operating-characteristic (ROC) curves were 0.85 and 0.89 for the neural network and the expert reader, respectively. When testing with abnormal subjects represented by "Possible AD" cases, ROC areas for both the network and the expert were 0.81. The neural network out-performed discriminant analysis. It is concluded that PET has potential for the detection of abnormal brain function in dementing diseases, and that the combination of neural networks and PET is a useful diagnostic tool. Despite the low-resolution "view" afforded the neural network, its performance was nearly equivalent to that of an expert reader.

Adult↗
PubMed1997

Artificial neural networks in cancer research.

The concept of artificial neural networks dates back to the early part of this century. However, their use in biological and medical research has only vastly proliferated during the last few years. It is now clear that these networks, which attempt to emulate functions of the human brain, can play a vital role in the field of cancer research, where they could be used in the diagnosis, prognosis and patient management stages of cancer evaluation. This paper presents a review of the underlying theory behind artificial neural networks and gives a broad overview of their many areas of application within the cancer field. This is achieved through the prognostic analysis of prostate cancer markers, the non-invasive diagnosis of lymph node involvement in breast cancer patients and the assessment of image cytometric data for predicting the metastatic potential of breast cancer.

Biomarkers, Tumor↗
PubMed1993

Simulation of stochastic micropopulation models--III. COGNET: an artificial neural network for visual recognition.

COGNET, based on a neural network first described by Fukushima, demonstrates the relationship between connectionist and other micropopulation models. Its success and physiological orientation led to an implementation using the SUMMERS simulation shell. After self-supervised learning, COGNET uses forward and backward propagation of signals to recognize partial and noisy patterns, and to reconstruct the originals. Stochastic features include variable thresholds for neuronal firing and occasional cell death. The successful implementation of COGNET demonstrates the generality of the concepts embodied in SUMMERS, which in turn promotes the reusability of software and facilitates the extension of computational models in biomedical research. COGNET itself forms a framework for building other physiologically oriented neural network models.

Computer Simulation↗
PubMed2003

Virgin olive oil quality classification combining neural network and MOS sensors.

A model based on neural networks has been designed to detect lampante virgin olive oils, a category of olive oil that cannot be consumed without a previous refining process according to the current regulation of the European Communities. The response of 7 metal oxide sensors analyzing 114 olive oil samples has been used in the design, training, and internal validation of the neural network with only 4.5% error in validation. The designed mathematical model, the equations of which are fully described, has been validated also with an external set of 13 samples of diverse varieties and geographical origins with 100% correct classification.

Chemistry Techniques, Analytical↗
PubMed2000

Detection of whale calls in noise: performance comparison between a beluga whale, human listeners, and a neural network.

This article examines the masking by anthropogenic noise of beluga whale calls. Results from human masking experiments and a software backpropagation neural network are compared to the performance of a trained beluga whale. The goal was to find an accurate, reliable, and fast model to replace lengthy and expensive animal experiments. A beluga call was masked by three types of noise, an icebreaker's bubbler system and propeller noise, and ambient arctic ice-cracking noise. Both the human experiment and the neural network successfully modeled the beluga data in the sense that they classified the noises in the same order from strongest to weakest masking as the whale and with similar call-detection thresholds. The neural network slightly outperformed the humans. Both models were then used to predict the masking of a fourth type of noise, Gaussian white noise. Their prediction ability was judged by returning to the aquarium to measure masked-hearing thresholds of a beluga in white noise. Both models and the whale identified bubbler noise as the strongest masker, followed by ramming, then white noise. Natural ice-cracking noise masked the least. However, the humans and the neural network slightly overpredicted the amount of masking for white noise. This is neglecting individual variation in belugas, because only one animal could be trained. Comparing the human model to the neural network model, the latter has the advantage of objectivity, reproducibility of results, and efficiency, particularly if the interference of a large number of signals and noise is to be examined.

Animals↗
PubMed1992

An assessment of neural network and statistical approaches for prediction of E. coli promoter sites.

We have constructed a perceptron type neural network for E. coli promoter prediction and improved its ability to generalize with a new technique for selecting the sequence features shown during training. We have also reconstructed five previous prediction methods and compared the effectiveness of those methods and our neural network. Surprisingly, the simple statistical method of Mulligan et al. performed the best amongst the previous methods. Our neural network was comparable to Mulligan's method when false positives were kept low and better than Mulligan's method when false negatives were kept low. We also showed the correlation between the prediction rates of neural networks achieved by previous researchers and the information content of their data sets.

Base Sequence↗
PubMed1995

The target switch algorithm: a constructive learning procedure for feed-forward neural networks.

We propose an efficient procedure for constructing and training a feed-forward neural network. The network can perform binary classification for binary or analogue input data. We show that the procedure can also be used to construct feedforward neural networks with binary-valued weights. Neural networks with binary-valued weights are potentially straightforward to implement using microelectronic or optical devices and they can also exhibit good generalization.

Algorithms↗
PubMed1991

Use of an artificial neural network for the diagnosis of myocardial infarction.

OBJECTIVE: To validate prospectively the use of an artificial neural network to identify myocardial infarction in patients presenting to an emergency department with anterior chest pain. DESIGN: Prospective, blinded testing. SETTING: Tertiary university teaching center. PATIENTS: A total of 331 consecutive adult patients presenting with anterior chest pain. MEASUREMENTS: Diagnostic sensitivity and specificity with regard to the diagnosis of acute myocardial infarction. MAIN RESULTS: An artificial neural network was trained on clinical pattern sets retrospectively derived from the cases of 351 patients hospitalized with a high likelihood of having myocardial infarction. It was prospectively tested on 331 consecutive patients presenting to an emergency department with anterior chest pain. The ability of the network to distinguish patients with from those without acute myocardial infarction was compared with that of physicians caring for the same patients. The physicians had a diagnostic sensitivity of 77.7% (95% CI, 77.0% to 82.9%) and a diagnostic specificity of 84.7% (CI, 84.0% to 86.4%). The artificial neural network had a sensitivity of 97.2% (CI, 97.2% to 97.5%; P = 0.033) and a specificity of 96.2% (CI, 96.2% to 96.4%; P less than 0.001). CONCLUSION: An artificial neural network trained to identify myocardial infarction in adult patients presenting to an emergency department may be a valuable aid to the clinical diagnosis of myocardial infarction; however, this possibility must be confirmed through prospective testing on a larger patient sample.

Adult↗
PubMed2001

The use of neural networks and logistic regression analysis for predicting pathological stage in men undergoing radical prostatectomy: a population based study.

PURPOSE: Clinical under staging occurs in 40% to 60% of patients who undergo radical prostatectomy for prostate cancer. To decrease under staging several methods of predicting pathological stage preoperatively have been developed based on statistical logistic regression analysis and neural networks. To our knowledge none has been validated in our homogeneous regional patient population to date. We created logistic regression and neural network models, and implemented and adapted them into our practice. We also compared the 2 methods to determine their value and practicality in daily clinical practice. We present the results of our novel approach for predicting pathological staging of prostate adenocarcinoma. MATERIALS AND METHODS: Between 1986 and 1999, 600 white men from the Aragon region of Spain underwent surgery for prostate cancer; of whom 468 were selected for study. Predictive study variables included patient age, clinical stage, biopsy Gleason score and preoperative prostate specific antigen (PSA). The predicted result included in analysis was organ confined or nonorgan confined disease. Data were analyzed by multivariate logistic regression and a supervised neural network (multilayer perceptron and radial basis function). Results were compared by comparing the areas under the receiver operating characteristics curves. RESULTS: We generated 5 logistic regression models. The model created with clinical staging, Gleason biopsy score and PSA distributed in 5 categories (p <0.001) with an area under the receiver operating characteristics curve of 0.840 proved to be most predictive of pathological stage. Similarly of the 6 neural network models evaluated the radial basis function model, which included age, clinical stage, Gleason biopsy score and preoperative PSA distributed in 5 categories with an area under the curve of 0.882, proved the most predictive but not superior to the logistic regression model. The difference in the area under the curves in the 2 chosen models was 0.042 (p = 0.1). CONCLUSIONS: It is possible to generate useful predictive models of organ confined disease using logistic regression or neural networks with high indexes of clinical and statistical validity. However, using these variables neural networks did not prove to be better than logistic regression analysis. Therefore, better predictive variables must be identified, preferably nonlinear characteristics with respect to the probability of organ confined tumor, to generate better predictive models using neural networks.

Aged↗
PubMed2004

A neural network model analysis to identify victims of intimate partner violence.

The objective of this study was to determine if a neural network model can identify victims of intimate partner violence (IPV). A custom neural network model was constructed and trained using the 1995 ED databases at Truman Medical Center of all female visits. The input vector developed was an array of 100 binary elements containing, in coded form, the patient's age, day of week, primary diagnosis (excluding 995.81), disposition, race, time, and E-code. The trained network was then presented with a series of 19,830 female patients from the 1996 ED database to determine if it could discriminate cases from control subjects. The neural network identified 231 of 297 known IPV victims (sensitivity 78%) in the 1996 database. It also categorized 2234 false-positive patients out of 19,533 IPV-negative patients (specificity 89%). A computer-based neural network model, when supplied with information commonly available in the ED medical record, can identify victims of IPV.

Case-Control Studies↗
PubMed1997

An improvement of neural networks applied to pharmaceutical problems.

In applying the neural network to the classification problem in pharmacology, we adopt an extended back-propagation (EBP) learning which adjusts the parameters appearing in an activation function, as well as the weights. The results of simulations show that such an extended learning speeds up the learning process as compared with the conventional basic back-propagation procedure, irrespective of the initial values of the parameters, which is extremely useful in the practical application of the neural network in the pharmaceutical field. We have also found that use of Morita's activation function beyond the sigmoid type further accelerates the EBP learning in some cases.

Algorithms↗
PubMed1995

Feed-forward neural networks for secondary structure prediction.

A feed-forward neural network has been employed for protein secondary structure prediction. Attempts were made to improve on previous prediction accuracies using a hierarchical mixture of experts (HME). In this method input data are clustered and used to train a series of different networks. Application of an HME to the prediction of protein secondary structure is shown to provide no advantages over a single network. We have also tried various new input representations, chosen to incorporate the effect of residues a long distance away in the one-dimensional amino acid chain. Prediction accuracy using these methods is comparable to that achieved by other neural networks.

Neural Networks, Computer↗
PubMed2003

Clustering ensembles of neural network models.

We show that large ensembles of (neural network) models, obtained e.g. in bootstrapping or sampling from (Bayesian) probability distributions, can be effectively summarized by a relatively small number of representative models. In some cases this summary may even yield better function estimates. We present a method to find representative models through clustering based on the models' outputs on a data set. We apply the method on an ensemble of neural network models obtained from bootstrapping on the Boston housing data, and use the results to discuss bootstrapping in terms of bias and variance. A parallel application is the prediction of newspaper sales, where we learn a series of parallel tasks. The results indicate that it is not necessary to store all samples in the ensembles: a small number of representative models generally matches, or even surpasses, the performance of the full ensemble. The clustered representation of the ensemble obtained thus is much better suitable for qualitative analysis, and will be shown to yield new insights into the data.

Algorithms↗
PubMed1992

Noninvasive diagnosis of coronary artery disease using a neural network algorithm.

This study examines the utility of neural networks for detecting coronary artery disease noninvasively by using the clinical examination variables and extracting useful information from the diastolic heart sounds associated with coronary occlusions. It has been widely reported that coronary stenoses produce sounds due to the turbulent blood flow in these vessels. These complex and highly attenuated signals taken from recordings made in both soundproof and noisy rooms were detected and analyzed to provide feature set based on the poles and power spectral density function (PSD) of the Autoregressive (AR) method after Adaptive Line Enhancement (ALE) method. In addition, some physical examination variables such as sex, age, body weight, smoking condition, diastolic pressure, systolic pressure and derivation from them were included in the feature vector. This feature vector was used as the input pattern to the neural network. The analysis was studied on one hundred recordings (63 abnormal, 37 normals). The network correctly identified 84% of the subjects with coronary artery disease and 89% of the normal subjects.

Algorithms↗