SEARCH · PubMed Health

Results for “Neural Networks, Computer”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original 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

PubMed1991

Application of a virtual neurode in a model thyroid diagnostic network.

UNLABELLED: Screening laboratory tests for thyroid disease often include serum levels for thyroxine (T4), thyrotropic hormone (TSH), and triiodothyronine resin binding (T3) as a measure of T4 binding to serum. A neural network using the above values as input was unable to converge during training to identify an output diagnoses of six common thyroid functional states. When binding protein (TBG) data were supplied the network readily converged. Since thyroxine binding can be roughly estimated from a relationship between T4 and T3, a virtual input node reflecting the binding was calculated from each T4/T3 input set and used as additional input. With this addition, the system trained easily and accurately diagnosed from the training set. CONCLUSION: 1) Quantitative laboratory data can be used in input neurodes in a diagnostic network 2) Training and diagnostic accuracy for the network is more efficient using the virtual TBG neurode than by either omitting TBG data or using actual TBG values.

Diagnosis, Computer-Assisted↗

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↗

PubMed2006

A functional hypothesis for adult hippocampal neurogenesis: avoidance of catastrophic interference in the dentate gyrus.

The dentate gyrus is part of the hippocampal memory system and special in that it generates new neurons throughout life. Here we discuss the question of what the functional role of these new neurons might be. Our hypothesis is that they help the dentate gyrus to avoid the problem of catastrophic interference when adapting to new environments. We assume that old neurons are rather stable and preserve an optimal encoding learned for known environments while new neurons are plastic to adapt to those features that are qualitatively new in a new environment. A simple network simulation demonstrates that adding new plastic neurons is indeed a successful strategy for adaptation without catastrophic interference.

Adaptation, Physiological↗

PubMed2004

Bayesian computation in recurrent neural circuits.

A large number of human psychophysical results have been successfully explained in recent years using Bayesian models. However, the neural implementation of such models remains largely unclear. In this article, we show that a network architecture commonly used to model the cerebral cortex can implement Bayesian inference for an arbitrary hidden Markov model. We illustrate the approach using an orientation discrimination task and a visual motion detection task. In the case of orientation discrimination, we show that the model network can infer the posterior distribution over orientations and correctly estimate stimulus orientation in the presence of significant noise. In the case of motion detection, we show that the resulting model network exhibits direction selectivity and correctly computes the posterior probabilities over motion direction and position. When used to solve the well-known random dots motion discrimination task, the model generates responses that mimic the activities of evidence-accumulating neurons in cortical areas LIP and FEF. The framework we introduce posits a new interpretation of cortical activities in terms of log posterior probabilities of stimuli occurring in the natural world.

Action Potentials↗

PubMed2006

A novel connectionist framework for computation of an approximate convex-hull of a set of planar points, circles and ellipses.

We propose a two layer neural network for computation of an approximate convex-hull of a set of points or a set of circles/ellipses of different sizes. The algorithm is based on a very elegant concept - shrinking of a rubber band surrounding the set of planar objects. Logically, a set of neurons is placed on a circle (rubber band) surrounding the objects. Each neuron has a parameter vector associated with it. This may be viewed as the current position of the neuron. The given set of points/objects exerts a force of attraction on every neuron, which determines how its current position will be updated (as if, the force determines the direction of movement of the neuron lying on the rubber band). As the network evolves, the neurons (parameter vectors) approximate the convex-hull more and more accurately. The scheme can be applied to find the convex-hull of a planar set of circles or ellipses or a mixture of the two. Some properties related to the evolution of the algorithm are also presented.

Algorithms↗

PubMed2004

Bio-inspired nano-sensor-enhanced CNN visual computer.

Nanotechnology opens new ways to utilize recent discoveries in biological image processing by translating the underlying functional concepts into the design of CNN (cellular neural/nonlinear network)-based systems incorporating nanoelectronic devices. There is a natural intersection joining studies of retinal processing, spatio-temporal nonlinear dynamics embodied in CNN, and the possibility of miniaturizing the technology through nanotechnology. This intersection serves as the springboard for our multidisciplinary project. Biological feature and motion detectors map directly into the spatio-temporal dynamics of CNN for target recognition, image stabilization, and tracking. The neural interactions underlying color processing will drive the development of nanoscale multispectral sensor arrays for image fusion. Implementing such nanoscale sensors on a CNN platform will allow the implementation of device feedback control, a hallmark of biological sensory systems. These biologically inspired CNN subroutines are incorporated into the new world of analog-and-logic algorithms and software, containing also many other active-wave computing mechanisms, including nature-inspired (physics and chemistry) as well as PDE-based sophisticated spatio-temporal algorithms. Our goal is to design and develop several miniature prototype devices for target detection, navigation, tracking, and robotics. This paper presents an example illustrating the synergies emerging from the convergence of nanotechnology, biotechnology, and information and cognitive science.

Animals↗

PubMed2004

Why voxel-based morphometric analysis should be used with great caution when characterizing group differences.

A variety of voxel-based morphometric analysis methods have been adopted by the neuroimaging community in the recent years. In this commentary we describe why voxel-based statistics, which are commonly used to construct statistical parametric maps, are very limited in characterizing morphological differences between groups, and why the effectiveness of voxel-based statistics is significantly biased toward group differences that are highly localized in space and of linear nature, whereas it is significantly reduced in cases with group differences of similar or even higher magnitude, when these differences are spatially complex and subtle. The complex and often subtle and nonlinear ways in which various factors, such as age, sex, genotype and disease, can affect brain morphology, suggest that alternative, unbiased methods based on statistical learning theory might be able to better quantify brain changes that are due to a variety of factors, especially when relationships between brain networks, rather than individual structures, and disease are examined.

Bias↗

PubMedDate not supplied

A more biologically plausible learning rule than backpropagation applied to a network model of cortical area 7a.

Area 7a of the posterior parietal cortex of the primate brain is concerned with representing head-centered space by combining information about the retinal location of a visual stimulus and the position of the eyes in the orbits. An artificial neural network was previously trained to perform this coordinate transformation task using the backpropagation learning procedure, and units in its middle layer (the hidden units) developed properties very similar to those of area 7a neurons presumed to code for spatial location (Andersen and Zipser, 1988; Zipser and Andersen, 1988). We developed two neural networks with architecture similar to Zipser and Andersen's model and trained them to perform the same task using a more biologically plausible learning procedure than backpropagation. This procedure is a modification of the Associative Reward-Penalty (AR-P) algorithm (Barto and Anandan, 1985), which adjusts connection strengths using a global reinforcement signal and local synaptic information. Our networks learn to perform the task successfully to any degree of accuracy and almost as quickly as with backpropagation, and the hidden units develop response properties very similar to those of area 7a neurons. In particular, the probability of firing of the hidden units in our networks varies with eye position in a roughly planar fashion, and their visual receptive fields are large and have complex surfaces. The synaptic strengths computed by the AR-P algorithm are equivalent to and interchangeable with those computed by backpropagation. Our networks also perform the correct transformation on pairs of eye and retinal positions never encountered before. All of these findings are unaffected by the interposition of an extra layer of units between the hidden and output layers. These results show that the response properties of the hidden units of a layered network trained to perform coordinate transformations, and their similarity with those of area 7a neurons, are not a specific result of backpropagation training. The fact that they can be obtained by a more biologically plausible learning rule corroborates the validity of this neural network's computational algorithm as a plausible model of how area 7a may perform coordinate transformations.

Algorithms↗

PubMed2000

Disparity tuning as simulated by a neural net.

Previous research has suggested that the processing of binocular disparity in complex cells may be described with an energy formalism. The energy formalism allows for a representation of disparity by differences in the position or in the phase of monocular receptive subfields of binocular cells, or by combination of these two types. We studied the coding of disparities with an approach complementary to previous algorithmic investigations. Since realization of these representations is probably not genetically determined but learned during ontogeny, we used backpropagation networks to study which of these three possibilities were realized within neural nets. Three types of networks were trained with noise patterns in analogy to the three types of energy models. The networks learned the task and generalized to untrained correlated noise pattern input. Outputs were broadly tuned to spatial frequency and did not respond to anti-correlated noise patterns. Although the energy model was not explicitly implemented, we could analyze the outputs of the networks using predictions of the energy formalism. After learning was completed, the model neurons preferred position shifts over phase shifts in representing disparity. We discuss the general meaning of these findings and the correspondences and deviations between the energy model, V1 neurons, and our networks.

Computer Simulation↗

PubMed2004

Automated EEG preprocessing during anaesthesia: new aspects using artificial neural networks.

The computer-aided detection of artefacts became an essential task with increasing automation of quantitative electroencephalogram (EEG) analysis during anaesthesiological applications. The different algorithms published so far required individual manual adjustment or have been based on limited decision criteria. In this study, we developed an artificial neural networks-(ANN-)aided method for automated detection of artefacts and EEG suppression periods. 72 hr EEG recorded before, during and after anaesthesia with propofol have been evaluated. Selected parameterized patterns of 0.25 s length were used to train the ANN (22 input, 8 hidden and 4 output neurons) with error back propagation. The detection performance of the ANN-aided method was tested with processing epochs between 1 to10 s. Related to examiner EEG evaluation, the average detection performance of the method was 72% sensitivity and 80% specificity for artefacts and 90% sensitivity and 92% specificity for EEG suppression. The improvement in signal-to-noise ratio with automated artefact processing was 1.39 times for the spectral edge frequency 95 (SEF95) and 1.89 times for the approximate entropy (ApEn). We conclude that ANN-aided preprocessing provide an useful tool for automated EEG evaluation in anaesthesiological applications.

Algorithms↗

PubMed1993

Analog circuits for relaxation networks.

Selected examples are presented of recent advances, primarily from the U.S. and Canada, in analog circuits for relaxation networks. Relaxation networks having feedback connections exhibit potentially greater computational power per neuron than feedforward networks. They are also more poorly understood especially with respect to learning algorithms. Examples are described of analog circuits for (i) supervised learning in deterministic Boltzmann machines, (ii) unsupervised competitive learning and feature maps and (iii) networks with resistive grids for vision and audition tasks. We also discuss recent progress on in-circuit learning and synaptic weight storage mechanisms.

Artificial Intelligence↗

PubMed1997

A neural network based artificial vision system for licence plate recognition.

This paper presents a neural network based artificial vision system able to analyze the image of a car given by a camera, locate the registration plate and recognize the registration number of the car. The paper describes in detail various practical problems encountered in implementing this particular application and the solutions used to solve them. The main features of the system presented are: controlled stability-plasticity behavior, controlled reliability threshold, both off-line and on-line learning, self assessment of the output reliability and high reliability based on high level multiple feedback. The system has been designed using a modular approach. Sub-modules can be upgraded and/or substituted independently, thus making the system potentially suitable in a large variety of vision applications. The OCR engine was designed as an interchangeable plug-in module. This allows the user to choose an OCR engine which is suited to the particular application and to upgrade it easily in the future. At present, there are several versions of this OCR engine. One of them is based on a fully connected feedforward artificial neural network with sigmoidal activation functions. This network can be trained with various training algorithms such as error backpropagation. An alternative OCR engine is based on the constraint based decomposition (CBD) training architecture. The system has showed the following performances (on average) on real-world data: successful plate location and segmentation about 99%, successful character recognition about 98% and successful recognition of complete registration plates about 80%.

Algorithms↗

PubMedDate not supplied

[Replication of a computer model of auditory hallucinations in schizophrenia].

INTRODUCTION: Abnormal synaptic pruning has been proposed by some authors as a theory to explain schizophrenia. It has been studied in different ways, some of which are computerized models. METHODS: A PC computer with MATLAB version 5.3 and Neural Networks Toolbox programs were used. To reproduce the model we used Elman's network for neuronal unit and McGlashan and Hoffman's network. RESULTS: The corrected percentage of detection improved as synapses were prunned. It increased from 68% to 91 %, reaching the highest detection level when 60% of the synapses were eliminated. Detection capacity was reduced when synaptic elimination continued and the program started detecting words in the absence of input. This was considered as a hallucination. When pruning reached from 80% to 95%, hallucinations occurred more frequently. CONCLUSIONS: The computer simulation model provides a symptom formation model, and also a way of understanding pruning's adaptative utility. When the neuronal units were eliminated (instead of connections), there was only progressive worsening in word perception. This confirms the hypothesis proposed that schizophrenia pathophysiology underlies neuronal connections, not the neurons.

Hallucinations↗

PubMed1998

Analysis of DNA curvature using artificial neural networks.

MOTIVATION: Our aim is to utilize an artificial neural network (ANN) for the prediction of DNA curvature in terms of retardation anomaly. RESULTS: An ANN capturing the role of phasing, increased helix flexibility, run of poly(A) tracts and flanking base pair effects in determining the extent of DNA curvature has been developed. The network predictions validate the known experimental results and also explain how the base pairs other than ApA affect the curvature. The results suggest that ANN can be used as a model-free tool for studying DNA curvature. AVAILABILITY: The optimal weights and the procedure to compute the retardation anomaly value are available on request from the authors. CONTACT: bdk@ems. ncl.res.in

Base Sequence↗

PubMed2001

Automatic identification of patients eligible for a pneumonia guideline: comparing the diagnostic accuracy of two decision support models.

BACKGROUND: In busy clinical settings, physicians often do not have enough time to identify patients for specific therapeutic guidelines. As a solution, decision support systems could automatically identify eligible patients and trigger computerized guidelines for specific diseases. Applying this idea to community-acquired pneumonia (CAP), we developed a Bayesian network (BN) and an artificial neural network (ANN) for identifying patients who have CAP and are eligible for a pneumonia guideline. OBJECTIVE: The aim of this study was to determine whether the diagnostic accuracy of these two decision support models differs in terms of identifying CAP patients. METHODS: We trained and tested the networks with a data set of 32,662 adult patients. For each network, we (1) calculated the specificity, the positive predictive value (PPV), and the negative predictive value (NPV) at a sensitivity of 95%, and (2) determined the area under the receiver operating characteristic curve (AUC) as a measure of overall accuracy. We tested for statistical difference between the AUCs using the correlated area z statistic. RESULTS: At a sensitivity of 95%, the respective values for specificity, PPV, and NPV were: 92.3%, 15.1%, and 99.9% for the BN, and 94.0%, 18.6%, and 99.9% for the ANN. The BN had an AUC of 0.9795 (95% CI: 0.9736, 0.9843), and the ANN had an AUC of 0.9855 (95% CI: 0.9805, 0.9894). The difference between the AUCs was statistically significant (p=0.0044). CONCLUSIONS: The networks achieved high overall accuracies on the testing data set. Because the difference in accuracies is statistically significant but not clinically significant, both networks are equally suited to drive a guideline.

Area Under Curve↗

PubMed2003

Controlled analysis of neurocontrollers with informational lesioning.

How does one aim to understand neural information processing? One of the difficult first challenges is to identify the roles of the network's elements. To this end a functional contribution analysis (FCA) method has been developed and applied for studying the neurocontrollers of evolutionary autonomous agents (EAAs). The FCA processes data composed of multiple lesion experiments and the corresponding performance levels that the agent obtains under these lesions. It calculates the contribution values (CVs) of the network's elements such that the ability to predict the agent's performance under new, unseen lesions is maximized. Previous analysis has found a strong dependence of the CVs and the prediction error on the specific type of lesioning method used, i.e. on the way in which the activity of lesioned neurons is disrupted. We present a new, informational lesioning method (ILM), which views a lesion as a noisy channel and applies a controlled lesion to the network by varying the lesioning level from large to arbitrarily small magnitudes. Studying the ILM within the FCA framework, our main results are threefold: first, that lower lesioning levels permit more accurate FCA predictions; second, that the usage of minute ILM lesioning levels can uncover the long-term effects of elements on the network's functioning; and third, that as the lesioning level decreases, the CVs tend to approach limit values, reflecting the importance of these elements in the intact, normal-functioning neurocontroller.

Adaptation, Physiological↗

PubMed2005

A remark on "scalar equations for synchronous Boolean networks with biological applications" by C. Farrow, J. Heidel, J. Maloney, and J. Rogers.

The problem of finding all cycles in the exponentially growing state space of synchronous Boolean networks was studied in the paper by C. Farrow, J. Heidel, J. Maloney, and J. R. Scalar, "Equations for synchronous Boolean networks with biological applications," IEEE trans. Neural Networks, vol. 15, no. 2, pp. 348-354 Mar. 2004. No efficient algorithm was given to solve the problem. We show that even the determination of the number of fixed points (cycles of length 1) for monotone Boolean networks and the determination of the existence of fixed points for general Boolean networks are both strong NP-complete.

Algorithms↗

PubMed1996

Sensory gating in a computer model of the CA3 neural network of the hippocampus.

We have developed a unique computer model of the CA3 region of the hippocampus that simulates the P50 auditory evoked potential response to repeated stimuli in order to study the neuronal circuits involved in a sensory processing deficit associated with schizophrenia. Our computer model of the CA3 hippocampal network includes recurrent activation from within the CA3 region as well as input from the entorhinal cortex and the medial septal nucleus. We used the model to help us determine if the cortical and septal inputs to the CA3 hippocampus alone are responsible for the gating of auditory evoked activity, or if the strong recurrent activity within the CA3 region contributes to this phenomenon. The model suggests that the medial septal input is critical for normal gating; however, to a large extent the activity of the medial septal input can be replaced by simulated stimulation of the hippocampal neurons by a nicotinic agonist. The model is thus consistent with experimental data that show that nicotine restores gating of the N40 evoked potential in fimbria-fornix lesioned rats and of the P50 evoked potential in schizophrenic patients.

Animals↗