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 649 records · Page 36Linked to original sources
PubMed1993

Advanced ovarian cancer. Neural network analysis to predict treatment outcome.

BACKGROUND: Quantitative methods for the analysis of prognostic information are important in order to use this knowledge optimally. The neural network is a new quantitative method where the fundamental building blocks are units which can be likened to neurons, and weighted connections which can be likened to synapses. The more the hidden units, the more complex the patterns that can be learnt. MATERIALS AND METHODS: Data from two Dutch studies in ovarian cancer were used to compare the previously reported survival rates predicted by the Cox's prognostic index with the prediction obtained by a neural network. RESULTS: Both the Cox's analysis and the neural network agreed on residual tumour size, stage, and performance status as being important for survival. The neural network identified additional predictive factors such as place of diagnosis and age. As the Cox's prognostic index has not been tested to predict survival on an independent data set a comparison with the results obtained in the neural network test set could not be performed. CONCLUSIONS: Neural networks perform at least as well as Cox's method for the prediction of survival, and prognostic factors can easily be identified. The analysis not only revealed the predictive power of some characteristics, but also the non-predictive power of the others.

Antineoplastic Combined Chemotherapy Protocols↗
PubMed1997

Artificial neural networks: application to electrical stimulation of the human nervous system.

Artificial neural networks are used increasingly in applications such as graphic pattern recognition, which are difficult to address with conventional statistical methods. In the management of chronic pain, graphic methods are used routinely; patients describe their patterns of pain using "pain drawings." The authors have previously reported an automated, computerized pain drawing methodology, which has been used by patients with implanted spinal cord stimulators to represent a technical goal of the procedure, the overlap of pain by stimulation paresthesias. Standard linear discriminant statistical methods have shown associations between stimulation parameters and electrode positions as independent variables and technical outcome and relief of pain as dependent variables. The authors have applied artificial neural networks to the problem of optimizing implanted stimulator adjustment. A data set of 3000 electrode combinations obtained in 41 patients was used to develop a linear discriminant statistical model on a mainframe computer and to train artificial neural networks on a personal computer. The performance of these two systems on a new data set obtained in 10 patients was compared with that of human "experts." The best neural network model was marginally better than the linear discriminant model; the variance in patient ratings was predicted by these models to a degree that the human experts were unable to predict. The authors anticipate expanding the role of these models and incorporating them into expert systems for clinical use.

Journal Article↗
PubMed2001

On the stability analysis of delayed neural networks systems.

In this paper, the problems of stability of delayed neural networks are investigated, including the stability of discrete and distributed delayed neural networks. Under the generalization of dropping the Lipschitzian hypotheses for output functions, some stability criteria are obtained by using the Liapunov functional method. We do not assume the symmetry of the connection matrix and we establish that the system admits a unique equilibrium point in which the output functions do not satisfy the Lipschitz conditions and do not require them to be differential or strictly monotonously increasing. These criteria can be used to analyze the dynamics of biological neural systems or to design globally stable artificial neural networks.

Action Potentials↗
PubMed2000

A neural network regression model for relative dose computation.

This paper describes a neural network (NN) regression model for relative dose computation. The input signal of the NN model includes depths and field sizes and the output is the relative dose, for example, percentage depth dose (%DD) or tissue-air ratio (TAR) in this paper. After a functional link has been created, the expressing ability of input patterns and the resolution of neural networks are enhanced. The trained neural network exhibits a good generalization and interpolation ability, and it can be easily extended to other relative dose regressions. We present two examples to verify the fitness of this model. The first calculates the percentage depth dose of a 14 MV x-ray beam for field sizes of 5 cm x 5 cm to 28 cm x 28 cm. The average error computed from the NN is less than 0.47% comparing with the original measured %DD. The second example calculates the TAR of a 4 MV x-ray beam for field sizes of 5 cm x 5 cm to 20 cm x 20 cm. In this example, the training data show that the average error computed from the NN is less than 0.48%, whereas that from the testing data is less than 0.37% after training. Such an NN model can be generalized to fit data for treatment planning with any linear accelerator and can also fit data stored as tissue-maximum ratio (TMR) and tissue-phantom ratio (TPR).

Linear Models↗
PubMed1998

Acute pulmonary embolism: cost-effectiveness analysis of the effect of artificial neural networks on patient care.

PURPOSE: To evaluate the cost-effectiveness of artificial neural networks for diagnosis in patients suspected of having acute pulmonary embolism who are typically referred for pulmonary angiography. MATERIALS AND METHODS: Four diagnostic strategies were explored to help define the diagnostic role of neural networks in patients suspected of having pulmonary embolism in whom nondiagnostic ventilation-perfusion lung scans were obtained. First, a network was used to determine which patients could be directly referred for treatment without angiography. Second, the network was applied to determine in which patients treatment could be withheld. Third, the network was used to distinguish patients in whom the network gave indeterminate responses and who should proceed to angiography. Each strategy was compared with use of angiography in terms of morbidity, mortality, and cost per life saved. RESULTS: The use of the neural network reduced the average cost per patient by more than one-half relative to the cost of angiography. Morbidity and mortality rates were also comparable to or lower than those associated with angiography. The results were consistent regardless of the prevalence of disease. CONCLUSION: The use of neural networks in the diagnosis of pulmonary embolism is a promising way to improve cost-effectiveness in the care of patients with nondiagnostic lung scans.

Acute Disease↗
PubMed1996

A predictor of transmembrane alpha-helix domains of proteins based on neural networks.

Back-propagation, feed-forward neural networks are used to predict alpha-helical transmembrane segments of proteins. The networks are trained on the few membrane proteins whose transmembrane alpha-helix domains are known to atomic or nearly atomic resolution. When testing is performed with a jackknife procedure on the proteins of the training set, the fraction of total correct assignments is as high as 0.87, with an average length for the transmembrane segments of 20 residues. The method correctly fails to predict any transmembrane domain for porin, whose transmembrane segments are beta-sheets. When tested on globular proteins, lower and upper limits of 1.6 and 3.5% for a total of 26826 residues are determined for the mispredicted cases, indicating that the predictor is highly specific for alpha-helical domains of membrane proteins. The predictor is also tested on 37 membrane proteins whose transmembrane topology is partially known. The overall accuracy is 0.90, two percentage points higher than that obtained with statistical methods. The reliability of the prediction is 100% for 60% of the total 18242 predicted residues of membrane proteins. Our results show that the local directional information automatically extracted by the neural networks during the training phase plays a key role in determining the accuracy of the prediction.

Amino Acid Sequence↗
PubMed1995

Neural network classification of infrared spectra of control and Alzheimer's diseased tissue.

Artificial neural network classification methods were applied to infrared spectra of histopathologically confirmed Alzheimer's diseased and control brain tissue. Principal component analysis was used as a preprocessing technique for some of these artificial neural networks while others were trained using the original spectra. The leave-one-out method was used for cross-validation and linear discriminant analysis was used as a performance benchmark. In the cases where principal components were used, the artificial neural networks consistently outperformed their linear discriminant counterparts; 100% versus 98% correct classifications, respectively, for the two class problem, and 90% versus 81% for a more complex five class problem. Using the original spectra, only one of the three selected artificial neural network architectures (a variation of the back-propagation algorithm using fuzzy encoding) produced results comparable to the best corresponding principal component cases: 98% and 85% correct classifications for the two and five class problems, respectively.

Alzheimer Disease↗
PubMed1978

An inference upon the neural network finding binocular correspondence.

Previously, the authors proposed a model of neural network extracting binocular parallax (Hirai and Fukushima, 1975). It is a multilayered network whose final layers consist of neural elements corresponding to "binocular depth neurons" found in monkey's visual cortex. The binocular depth neuron is selectively sensitive to a binocular stimulus with a specific amount of binocular parallax and does not respond to a monocular one. As described in the last chapter of the previous article (Hirai and Fukushima, 1975), when a binocular pair of input patterns consist of, for example, many vertical bars placed very closely to each other, the binocular depth neurons might respond not only to correct binocular pairs, but also to incorrect ones. Our present study is concentrated upon how the visual system finds correct binocular pairs or binocular correspondence. It is assumed that some neural network is cascaded after the binocular depth neurons and finds out correct binocular correspondence by eliminating the incorrect binocular pairs. In this article a model of such neural network is proposed. The performance of the model has been simulated on a digital computer. The results of the computer simulation show that this model finds binocular correspondence satisfactorily. It has been demonstrated by the computer simulation that this model also explains the mechanism of the hysteresis in the binocular depth perception reported by Fender and Julesz (1967).

Computers↗
PubMed2000

Improving diagnostic accuracy using a hierarchical neural network to model decision subtasks.

A number of quantitative models including linear discriminant analysis, logistic regression, k nearest neighbor, kernel density, recursive partitioning, and neural networks are being used in medical diagnostic support systems to assist human decision-makers in disease diagnosis. This research investigates the decision accuracy of neural network models for the differential diagnosis of six erythematous-squamous diseases. Conditions where a hierarchical neural network model can increase diagnostic accuracy by partitioning the decision domain into subtasks that are easier to learn are specifically addressed. Self-organizing maps (SOM) are used to portray the 34 feature variables in a two dimensional plot that maintains topological ordering. The SOM identifies five inconsistent cases that are likely sources of error for the quantitative decision models; the lower bound for the diagnostic decision error based on five errors is 0.0140. The traditional application of the quantitative models cited above results in diagnostic error levels substantially greater than this target level. A two-stage hierarchical neural network is designed by combining a multilayer perceptron first stage and a mixture-of-experts second stage. The second stage mixture-of-experts neural network learns a subtask of the diagnostic decision, the discrimination between seborrheic dermatitis and pityriasis rosea. The diagnostic accuracy of the two stage neural network approaches the target performance established from the SOM with an error rate of 0.0159.

Chronic Disease↗
PubMed2001

Pseudo-outer product based fuzzy neural network fingerprint verification system.

Fingerprint identification has been used in law enforcement applications over the last century, and has become the de facto international standard for positive identification. With the emergence of automated fingerprint identification technologies, it has assisted in making the once labour-intensive process of classifying, searching and matching a thing of the past. As a biometrics proof of identification, not many have ventured into the world of fingerprint identification using fuzzy neural networks. In this paper, a database of fingerprint images is constructed and a fuzzy neural network called the pseudo outer product fuzzy neural network (POPFNN) [Zhou, R.W. & Quek, C. (1996). A pseudo outer-product based fuzzy neural network. Neural Networks, 9(9), 1569-1581] is trained to detect similarity between two fingerprints and decide whether they belong to the same person. The fundamental idea is that, given a person's fingerprints taken under different conditions, the POPFNN based fingerprint verification system should be sufficiently robust to distinguish the difference. The people providing the fingerprint samples are subjected to different 'adverse' conditions; from wetness to chemical treatments. Fingerprint images are taken after conditions such as: after a shower, holding pineapples (mild acid from fruit), after washing one's hands, etc. The characteristics of POPFNN, such as the learning, generalisation, and high computational abilities, make fingerprint verification particularly powerful when verifying authentic fingerprints subjected to external conditions and recognising spurious ones. In order to demonstrate the efficacy of POPFNN and its application in the fingerprint verification system (FVS), several types of experiments have been designed and implemented in this work. The experimental results and analysis are presented at the end of the paper for discussion.

Algorithms↗
PubMedDate not supplied

The use of artificial neural networks in biomedical technologies: an introduction.

Artificial neural networks (NN) are systems than can learn. In the most common situation, an operator trains the system on a set of input and output data belonging to a particular category. If new data of the same category, but not in the training set, are presented to the system, the NN can use the learned data to predict outcomes without any specific programming relating to the category of events involved. The fields of application of NN have increased dramatically in the past few years. Originally, the NN technique was mainly in the hands of computer programming specialists and the applications concentrated on tasks such as decision systems and signal processing. However, this picture has changed due to the emergence of user-friendly NN software for personal computers. A large variety of possible NN applications now exist for non-computer specialists. Thus, with only a very modest knowledge of the theory behind neural networks, it is possible to attack complicated problems in a researcher's own area of specialty with the NN technique. This is especially true in the field of medical technology, the topic of this review. The review is divided into three sections: 1) an elementary introduction to useful NN methods; 2) a review of the most important applications of the NN technique to this point in time; 3) a summary of available computer details that would be needed for a beginner in this field.

Humans↗
PubMed1996

Pharmaceutical granulation and tablet formulation using neural networks.

Current-day pharmaceutical formulation may be trial and error in nature due to the absence of a clear relationship between the formulation characteristics (output variables) and the material and process variables (input variables). Neural networks are networks of adaptable nodes, which through a process of learning from task examples, store experiential knowledge and make it available for prediction. Prediction of a model granulation and tablet system characteristics from the knowledge of material and process variables utilizing neural networks is the basis of this presentation. The formulation design contained the following variables: granulation equipment, diluent, method of binder addition, and the binder concentration. The material, process, granulation evaluation, and tablet evaluation data of the formulations were used as the data set for training and testing of the neural network models. A comparison of the neural network prediction performance with that of regression models was also done. Both the granulation model and the tablet model converged fairly rapidly in the training step. In the testing step, the predictions for all granulation model variables (geometric mean particle size, flow value, bulk density, and tap density) were satisfactory. In the tablet model, the predictions for disintegration and thickness were also satisfactory. The predictions for hardness and friability were less than satisfactory. Two situations where the neural network may not perform adequately are discussed. The neural network prediction is better or comparable for all the predicted variables in this study compared to regression methods. The results clearly show the applicability of neural networks to formulation modeling.

Caffeine↗
PubMed2006

Stability analysis of Cohen-Grossberg neural networks.

Without assuming boundedness and differentiability of the activation functions and any symmetry of interconnections, we employ Lyapunov functions to establish some sufficient conditions ensuring existence, uniqueness, global asymptotic stability, and even global exponential stability of equilibria for the Cohen-Grossberg neural networks with and without delays. Our results are not only presented in terms of system parameters and can be easily verified and also less restrictive than previously known criteria and can be applied to neural networks, including Hopfield neural networks, bidirectional association memory neural networks, and cellular neural networks.

Journal Article↗
PubMed2006

Neural network explanation using inversion.

An important drawback of many artificial neural networks (ANN) is their lack of explanation capability [Andrews, R., Diederich, J., & Tickle, A. B. (1996). A survey and critique of techniques for extracting rules from trained artificial neural networks. Knowledge-Based Systems, 8, 373-389]. This paper starts with a survey of algorithms which attempt to explain the ANN output. We then present HYPINV, a new explanation algorithm which relies on network inversion; i.e. calculating the ANN input which produces a desired output. HYPINV is a pedagogical algorithm, that extracts rules, in the form of hyperplanes. It is able to generate rules with arbitrarily desired fidelity, maintaining a fidelity-complexity tradeoff. To our knowledge, HYPINV is the only pedagogical rule extraction method, which extracts hyperplane rules from continuous or binary attribute neural networks. Different network inversion techniques, involving gradient descent as well as an evolutionary algorithm, are presented. An information theoretic treatment of rule extraction is presented. HYPINV is applied to example synthetic problems, to a real aerospace problem, and compared with similar algorithms using benchmark problems.

Algorithms↗
PubMed1997

Application of a multi-structure neural network (MSNN) to sorting pistachio nuts.

A multi-structure neural network (MSNN) classifier consisting of four discriminators followed by a maximum selector was designed and applied to classification of four grades of pistachio nuts. Each discriminator was a multi-layer feed-forward neural network with two hidden layers and a single-neuron output layer. Fourier descriptors of the nuts' boundaries and their area were used as the recognition features. The individual discriminators were trained using a biased technique and a back-propagation algorithm. The MSNN classifier gave an average classification performance of 95.0%. This was an increase of 14.8% over the performance of a multi-layer neural network (MLNN) with similar complexity for classifying the same set of patterns.

Discrimination Learning↗
PubMed1999

A neural network to predict symptomatic lung injury.

A nonlinear neural network that simultaneously uses pre-radiotherapy (RT) biological and physical data was developed to predict symptomatic lung injury. The input data were pre-RT pulmonary function, three-dimensional treatment plan doses and demographics. The output was a single value between 0 (asymptomatic) and 1 (symptomatic) to predict the likelihood that a particular patient would become symptomatic. The network was trained on data from 97 patients for 400 iterations with the goal to minimize the mean-squared error. Statistical analysis was performed on the resulting network to determine the model's accuracy. Results from the neural network were compared with those given by traditional linear discriminate analysis and the dose-volume histogram reduction (DVHR) scheme of Kutcher. Receiver-operator characteristic (ROC) analysis was performed on the resulting network which had Az = 0.833 +/- 0.04. (Az is the area under the ROC curve.) Linear discriminate multivariate analysis yielded an Az = 0.813 +/- 0.06. The DVHR method had Az = 0.521 +/- 0.08. The network was also used to rank the significance of the input variables. Future studies will be conducted to improve network accuracy and to include functional imaging data.

Adult↗
PubMed1993

Simulation studies of data classification by artificial neural networks: potential applications in medical imaging and decision making.

Artificial neural networks are being investigated in the field of medical imaging as a means to facilitate pattern recognition and patient classification. In the work reported here, the effects of internal structure and the nature of input data on the performance of neural networks were investigated systematically using computer-simulated data. Network performance was evaluated quantitatively by means of receiver operating characteristic analysis and compared with the performance of an ideal statistical decision maker. We found that the relatively simple neural networks investigated in this study can perform at the level of an ideal decision maker. These simple networks were also found to learn accurately even when the training data are extremely unbalanced with respect to the prevalence of actually positive cases and to differentiate input data patterns by recognizing their unique characteristics.

Computer Simulation↗