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 1,693 records · Page 94Linked to original sources
PubMed2003

Breast lesions on sonograms: computer-aided diagnosis with nearly setting-independent features and artificial neural networks.

PURPOSE: To develop a computer-aided diagnosis (CAD) algorithm with setting-independent features and artificial neural networks to differentiate benign from malignant breast lesions. MATERIALS AND METHODS: Two sets of breast sonograms were evaluated. The first set contained 160 lesions and was stored directly on the magnetic optic disks from the ultrasonographic (US) system. Four different boundaries were delineated by four persons for each lesion in the first set. The second set comprised 111 lesions that were extracted from the hard-copy images. Seven morphologic features were used, five of which were newly developed. A multilayer feed-forward neural network was used as the classifier. Reliability, extendability, and robustness of the proposed CAD algorithm were evaluated. Results with the proposed algorithm were compared with those with two previous CAD algorithms. All performance comparisons were based on paired-samples t tests. RESULTS: The area under the receiver operating characteristic curve (A(z)) was 0.952 +/- 0.014 for the first set, 0.982 +/- 0.004 for the first set as the training set and the second set as the prediction set, 0.954 +/- 0.016 for the second set as the training set and the first set as the prediction set, and 0.950 +/- 0.005 for all 271 lesions. At the 5% significance level, the performance of the proposed CAD algorithm was shown to be extendible from one set of US images to the other set and robust for both small and large sample sizes. Moreover, the proposed CAD algorithm was shown to outperform the two previous CAD algorithms in terms of the A(z) value. CONCLUSION: The proposed CAD algorithm could effectively and reliably differentiate benign and malignant lesions. The proposed morphologic features were nearly setting independent and could tolerate reasonable variation in boundary delineation.

Adolescent↗
PubMed2006

Classification of EMG signals using wavelet neural network.

An accurate and computationally efficient means of classifying electromyographic (EMG) signal patterns has been the subject of considerable research effort in recent years. Quantitative analysis of EMG signals provides an important source of information for the diagnosis of neuromuscular disorders. Following the recent development of computer-aided EMG equipment, different methodologies in the time domain and frequency domain have been followed for quantitative analysis. In this study, feedforward error backpropagation artificial neural networks (FEBANN) and wavelet neural networks (WNN) based classifiers were developed and compared in relation to their accuracy in classification of EMG signals. In these methods, we used an autoregressive (AR) model of EMG signals as an input to classification system. A total of 1200 MUPs obtained from 7 normal subjects, 7 subjects suffering from myopathy and 13 subjects suffering from neurogenic disease were analyzed. The success rate for the WNN technique was 90.7% and for the FEBANN technique 88%. The comparisons between the developed classifiers were primarily based on a number of scalar performance measures pertaining to the classification. The WNN-based classifier outperformed the FEBANN counterpart. The proposed WNN classification may support expert decisions and add weight to EMG differential diagnosis.

Adolescent↗
PubMed2005

The use of discriminant analysis and neural networks to forecast the severity of the Poaceae pollen season in a region with a typical Mediterranean climate.

Biological particles in the air such as pollen grains can cause environmental problems in the allergic population. Medical studies report that a prior knowledge of pollen season severity can be useful in the management of pollen-related diseases. The aim of this work was to forecast the severity of the Poaceae pollen season by using weather parameters prior to the pollen season. To carry out the study a historical database of 21 years of pollen and meteorological data was used. First, the years were grouped into classes by using cluster analysis. As a result of the grouping, the 21 years were divided into 3 classes according to their potential allergenic load. Pre-season meteorological variables were used, as well as a series of characteristics related to the pollen season. When considering pre-season meteorological variables, winter variables were separated from early spring variables due to the nature of the Mediterranean climate. Second, a neural network model as well as a discriminant linear analysis were built to forecast Poaceae pollen season severity, according to the three classes previously defined. The neural network yielded better results than linear models. In conclusion, neural network models could have a high applicability in the area of prevention, as the allergenic potential of a year can be determined with a high degree of reliability, based on a series of meteorological values accumulated prior to the pollen season.

Climate↗
PubMed2002

Neural network identification of high-risk suicide patients.

BACKGROUND: A history of medically serious suicide attempts (MSSA) has been considered a major risk factor for suicide. Backpropagation (BP) neural networks (NN) successfully detected patient files of patients who had committed MSSA but failed to identify MSSA patients when trained with data collected during face-to-face interviews. PRIMARY OBJECTIVES: To develop an expert support system for the detection of suicide risk in patients with major psychiatric disorders using fuzzy adaptive learning control networks (FALCON) and BP enhanced with fuzzy logic neural networks. METHODS AND PROCEDURES: We used four types of input sets: (a) the most significant variables as defined by Garson's calculation of trained BP; (b) variables identified by logistic regression; (c) variables considered significant by experienced clinicians; or (d) the variables considered definitive by the computer programmer. In addition, the complete set of 59 input variables was used. The target variable was the existence of prior MSSA. EXPERIMENTAL INTERVENTIONS: Four training trials were performed with each input set (75% of the data was used for training and 25% for testing). Data was collected from 197 (50 MSSA, 147 non-MSSA) patients hospitalized at Sha'ar Menashe Mental Health Centre. MAIN OUTCOMES AND RESULTS: BP enhanced with fuzzy logic achieved the most accurate results with the three variables chosen by the programmer (mean +/- SD; sensitivity=97% +/- 3.4%, specificity=69.2% +/- 6.9%), and FALCON trained with 15 variables chosen by an expert clinician obtained the highest sensitivity and specificity (mean +/- SD; sensitivity=94% +/- 6.9%, specificity=69% +/- 6.9%). CONCLUSIONS: A combined system of BP enhanced with fuzzy logic and FALCON may improve the detection rate of MSSA patients during patient interviews. A prospective validation study testing a combined system across 1, 2 and 5-year spans may provide a reliable instrument to aid in the detection of suicide prone psychiatric patients.

Adult↗
PubMed1999

Beam-orientation customization using an artificial neural network.

A methodology for the constrained customization of coplanar beam orientations in radiotherapy treatment planning using an artificial neural network (ANN) has been developed. The geometry of the patients, with cancer of the prostate, was modelled by reducing the external contour, planning target volume (PTV) and organs at risk (OARs) to a set of cuboids. The coordinates and size of the cuboids were given to the ANN as inputs. A previously developed beam-orientation constrained-customization (BOCC) scheme employing a conventional computer algorithm was used to determine the customized beam orientations in a training set containing 45 patient datasets. Twelve patient datasets not involved in the training of the artificial neural network were used to test whether the ANN was able to map the inputs to customized beam orientations. Improvements from the customized beam orientations were compared with standard treatment plans with fixed gantry angles and plans produced from the BOCC scheme. The ANN produced customized beam orientations within 5 degrees of the BOCC scheme in 62.5% of cases. The average difference in the beam orientations produced by the ANN and the BOCC scheme was 7.7 degrees (+/-1.7, 1 SD). Compared with the standard treatment plans, the BOCC scheme produced plans with an increase in the average tumour control probability (TCP) of 5.7% (+/-1.4, 1 SD) whilst the ANN generated plans increased the average TCP by 3.9% (+/-1.3, 1 SD). Both figures refer to the TCP at a fixed rectal normal tissue complication probability (NTCP) of 1%. In conclusion, even using a very simple model for the geometry of the patient, an ANN was able to produce beam orientations that were similar to those produced by a conventional computer algorithm.

Humans↗
PubMed2002

A global optimum approach for one-layer neural networks.

The article presents a method for learning the weights in one-layer feedforward neural networks minimizing either the sum of squared errors or the maximum absolute error, measured in the input scale. This leads to the existence of a global optimum that can be easily obtained solving linear systems of equations or linear programming problems, using much less computational power than the one associated with the standard methods. Another version of the method allows computing a large set of estimates for the weights, providing robust, mean or median, estimates for them, and the associated standard errors, which give a good measure for the quality of the fit. Later, the standard one-layer neural network algorithms are improved by learning the neural functions instead of assuming them known. A set of examples of applications is used to illustrate the methods. Finally, a comparison with other high-performance learning algorithms shows that the proposed methods are at least 10 times faster than the fastest standard algorithm used in the comparison.

Journal Article↗
PubMed2006

Class identity assignment for amphetamines using neural networks and GC-FTIR data.

An exploratory analysis was performed in order to evaluate the feasibility of building of neural network (NN) systems automating the identification of amphetamines necessary in the investigation of drugs of abuse for epidemiological, clinical and forensic purposes. A first neural network system was built to distinguish between amphetamines and nonamphetamines. A second, more refined system, aimed to the recognition of amphetamines according to their toxicological activity (stimulant amphetamines, hallucinogenic amphetamines, nonamphetamines). Both systems proved that discrimination between amphetamines and nonamphetamines, as well as between stimulants, hallucinogens and nonamphetamines is possible (83.44% and 85.71% correct classification rate, respectively). The spectroscopic interpretation of the 40 most important input variables (GC-FTIR absorption intensities) shows that the modeling power of an input variable seems to be correlated with the stability and not with the intensity of the spectral interaction. Thus, discarding variables only because they correspond to spectral windows with weak absorptions does not seem be not advisable.

Amphetamines↗
PubMed2003

Artificial neural network learns to differentiate normal TMJs and nonreducing displaced disks after training on incisor-point chewing movements.

Previous authors have described four frontal gum-chewing patterns associated with normal and abnormal TMJ disk-condyle relationships. The objective of this study was to create an automatic detection capability (expert system) by training an artificial neural network to recognize nonreducing displaced disks from frontal chewing data. Sixty-eight (68) subjects, 29 with normal joints, 18 with unilateral nonreducing displaced disks and 21 with bilateral nonreducing displaced disks were selected from a continuous series of patients seeking treatment for TMD. Right-sided gum chewing was recorded from all patients. Left-sided chewing was also recorded from the right unilateral patients. 50% of the vertical, lateral and timing values at 10%, 65% and 100% of opening and at 30%, 70% and 90% of closing were used to train an artificial neural network. The remaining 50% were used for testing. All normal subjects were detected as normal (specificity = 100%). Two bilateral and two unilateral patients were not detected (sensitivity = 91.8%). Four (4) patients received the wrong classification (unilateral vs. bilateral) and one patient received both (undecided) for an overall accuracy = 86.8%. The artificial neural network detected, at an acceptable level of error, the presence and type of nonreducing disk displacement from frontal plane jaw recordings of gum chewing in a group of real patients seeking treatment for TMD. Since it is very inexpensive to conduct, mastication analysis appears to have the potential of an excellent cost/benefit ratio.

Diagnosis, Computer-Assisted↗
PubMedDate not supplied

Detection and deletion of motion artifacts in electrogastrogram using feature analysis and neural networks.

Electrogastrogram is a surface measurement of gastric myoelectrical activity, and electrogastrography has been an attractive method for physiological and pathophysiological studies of the stomach due to its noninvasive nature. Motion artifacts, however, ruin the electrogastrogram (EGG), and make the analysis very difficult and sometimes even impossible. They must be eliminated from EGG signals before analysis. Up to now, this can only be done by visual inspection, which is not only time-consuming but also subjective. In this study, a method using feature analysis and neural networks has been developed to realize automatic detection and elimination of the motion artifacts in EGG recordings by computer. Experiments were conducted to investigate the characteristics of different motion artifacts. Useful features were extracted, and different combinations of the features used as the input of the neural network were compared to obtain the optimal performance for the detection of motion artifacts using the artificial neural network.

Biomedical Engineering↗
PubMed2000

Artificial neural networks as a method to improve the precision of subcutaneous adipose tissue thickness measurements by means of the optical device LIPOMETER.

The LIPOMETER is an optical device for measuring the thickness of a subcutaneous adipose tissue layer. It illuminates the interesting layer, measures the backscattered light signals and from these, it computes absolute values of subcutaneous adipose tissue layer thickness (in mm). Previously, these light pattern values were fitted by nonlinear regression analysis to absolute values provided by computed tomography. Nonlinear regression analysis might provide slight limitations for our problem: a selected curve type cannot be changed afterwards during the application of the measurement device. Artificial neural networks yield a more flexible approach to this fitting problem and might be able to refine the fitting results. In the present paper we compare nonlinear regression analysis with the behaviour of different architectures of multilayer feed forward neural networks trained by error back propagation. Specifically, we are interested whether neural networks are able to yield a better fit of the LIPOMETER light patterns to absolute subcutaneous adipose tissue layer thicknesses than the nonlinear regression techniques. Different architectures of these networks are able to surpass the best result of regression analysis in training and test, providing higher correlation coefficients, regression lines with absolute values obtained from computed tomography closer to the line of identity, decreased sums of absolute and squared deviations, and higher measurement agreement.

Adult↗
PubMed2002

Neural network simulations of the primate oculomotor system IV. A distributed bilateral stochastic model of the neural integrator of the vertical saccadic system.

The present report examines the performance of a distributed bi-directional neural network that simulates the vertical velocity to position integrator of the primate brain. Consistent with anatomy and physiology, its units receive stochastically weighted input from vertical medium-lead burst neurons. Also consistent with anatomy, units belonging to integrators with opposite on-directions (up or down) are interconnected via the posterior commissure (again in a stochastically weighted manner) and they can be excitatory or inhibitory. To demonstrate that integration can be a one-step process, the output of model units was routed directly to vertical motoneurons. Model units replicate the wide range of saccade-related discharge patterns encountered in the portion of the primate brain that is thought to house the vertical neural integrator (the interstitial nucleus of Cajal) while "lesions" of model units and/or their interconnections replicate the symptoms which follow insults to this brain area.

Animals↗
PubMed1996

Interactions between the neural networks for escape and swimming in goldfish.

Interactions between neural networks for different motor behaviors occur frequently in nature; however, there are few vertebrate models for studying these interactions. One potentially useful model involves the interactions between escape and swimming behaviors in fish. Fish can produce escape bends while swimming, using some of the same axial muscles for both behaviors. Here we study the interactions between escape and swimming in a paralyzed goldfish preparation in which we can activate the networks for both behaviors. Fictive swimming was elicited by electrical stimulation in the midbrain locomotor region. During the swimming, we fired a single action potential in the reticulospinal Mauthner (M) cell, which initiates the escape behavior (Zottoli, 1977). Firing the M cell overrode the swimming motor output to produce an output appropriate for escape regardless of the phase of swimming at which it was fired. The M cell also could reset the swimming rhythm dramatically in a way that led to a smooth transition from an escape bend to one side into subsequent swimming. Both the override and reset supported predictions based on previous studies of the organization of the M-cell network. They apparently allow for a well coordinated motor output when a fish must produce an escape while swimming. The potent effects of one action potential in a single, identifiable reticulospinal neuron make this an attractive model system for future studies of the cellular basis of interactions between descending pathways and spinal rhythm-generating networks.

Animals↗
PubMedDate not supplied

Artificial neural network modification of simulation-based fitting: application to a protein-lipid system.

Simulation-based fitting has been applied to data analysis and parameter determination of complex experimental systems in many areas of chemistry and biophysics. However, this method is limited because of the time costs of the calculations. In this paper it is proposed to approximate and substitute a simulation model by an artificial neural network during the fitting procedure. Such a substitution significantly speeds up the parameter determination. This approach is tested on a model of fluorescence resonance energy transfer (FRET) within a system of site-directed fluorescence labeled M13 major coat protein mutants incorporated into a lipid bilayer. It is demonstrated that in our case the application of a trained artificial neural network for the substitution of the simulation model results in a significant gain in computing time by a factor of 5 x 10(4). Moreover, an artificial neural network produces a smooth approximation of the noisy results of a stochastic simulation.

Algorithms↗
PubMed2005

Prediction of siRNA knockdown efficiency using artificial neural network models.

Selective knockdown of gene expression by short interference RNAs (siRNAs) has allowed rapid validation of gene functions and made possible a high throughput, genome scale approach to interrogate gene function. However, randomly designed siRNAs display different knockdown efficiencies of target genes. Hence, various prediction algorithms based on siRNA functionality have recently been constructed to increase the likelihood of selecting effective siRNAs, thereby reducing the experimental cost. Toward this end, we have trained three Back-propagation and Bayesian neural network models, previously not used in this context, to predict the knockdown efficiencies of 180 experimentally verified siRNAs on their corresponding target genes. Using our input coding based primarily on RNA structure thermodynamic parameters and cross-validation method, we showed that our neural network models outperformed most other methods and are comparable to the best predicting algorithm thus far published. Furthermore, our neural network models correctly classified 74% of all siRNAs into different efficiency categories; with a correlation coefficient of 0.43 and receiver operating characteristic curve score of 0.78, thus highlighting the potential utility of this method to complement other existing siRNA classification and prediction schemes.

Algorithms↗
PubMed2005

A neural network method for automatic and incremental learning applied to patient-dependent seizure detection.

OBJECTIVE: Describe and evaluate a neural network method for automatic and incremental learning applied to patient-dependent seizure detection. Compare the classification ability of various time-frequency methods including FFT spectrogram, spectral edge frequency and bicoherence. METHODS: 57 seizures from 10 epilepsy patients are used. A probabilistic neural network (PNN) is trained and incrementally updated in a novel fashion. The speed and accuracy of the method is evaluated with different training parameters and time-frequency methods. RESULTS: Training the PNN on a single seizure from each record offers better performance (sensitivity = 0.89 and false-positive-rate = 0.56/h) than 3 patient-independent seizure detection algorithms. The method is virtually unaffected by the settings of various training parameters. Training is very fast (0.9 s), and the accuracy improves as more examples are added incrementally (without retraining). The overall best time-frequency method was the FFT spectrogram. The bicoherence plus the FFT spectrogram was the best method on 4 records, improving the correlation from 0.111 to 0.940 on one and from 0.288 to 0.612 on another. CONCLUSIONS: The proposed method offers accurate, robust and virtually instantaneous training and incremental learning when applied to patient-dependent seizure detection. SIGNIFICANCE: Accurate seizure detection can improve patient care in both the epilepsy monitoring unit and the intensive care unit. Future applications include patient-independent algorithms that continue to learn as new examples are encountered.

Automation↗
PubMed2005

On using an adaptive neural network to predict lung tumor motion during respiration for radiotherapy applications.

In this study we address the problem of predicting the position of a moving lung tumor during respiration on the basis of external breathing signals--a technique used for beam gating, tracking, and other dynamic motion management techniques in radiation therapy. We demonstrate the use of neural network filters to correlate tumor position with external surrogate markers while simultaneously predicting the motion ahead in time, for situations in which neither the breathing pattern nor the correlation between moving anatomical elements is constant in time. One pancreatic cancer patient and two lung cancer patients with mid/upper lobe tumors were fluoroscopically imaged to observe tumor motion synchronously with the movement of external chest markers during free breathing. The external marker position was provided as input to a feed-forward neural network that correlated the marker and tumor movement to predict the tumor position up to 800 ms in advance. The predicted tumor position was compared to its observed position to establish the accuracy with which the filter could dynamically track tumor motion under nonstationary conditions. These results were compared to simplified linear versions of the filter. The two lung cancer patients exhibited complex respiratory behavior in which the correlation between surrogate marker and tumor position changed with each cycle of breathing. By automatically and continuously adjusting its parameters to the observations, the neural network achieved better tracking accuracy than the fixed and adaptive linear filters. Variability and instability in human respiration complicate the task of predicting tumor position from surrogate breathing signals. Our results show that adaptive signal-processing filters can provide more accurate tumor position estimates than simpler stationary filters when presented with nonstationary breathing motion.

Biophysical Phenomena↗
PubMed2003

The functional localization of neural networks using genetic algorithms.

We presented an algorithm for extracting Boolean functions (propositions, rules) from the units in trained neural networks. The extracted Boolean functions make the hidden units understandable. However, in some cases, the extracted Boolean functions are complicated, and so are not understandable, which means that the hidden units are not functionally localized. This paper presents an algorithm for the functional localization of (the hidden units of) neural networks. When a hidden unit is well approximated to a low-order Boolean function, the unit can be regarded as functionally localized. The functional localization of a hidden unit is evaluated by the error between the hidden unit and the low-order Boolean function extracted from the hidden unit. The optimization is executed by genetic algorithms. We applied it to vote data, mushroom data and chess data. Experimental results show that the algorithm works well.

Agaricales↗
PubMed2002

Fractional Fourier transform pre-processing for neural networks and its application to object recognition.

This study investigates fractional Fourier transform pre-processing of input signals to neural networks. The fractional Fourier transform is a generalization of the ordinary Fourier transform with an order parameter a. Judicious choice of this parameter can lead to overall improvement of the neural network performance. As an illustrative example, we consider recognition and position estimation of different types of objects based on their sonar returns. Raw amplitude and time-of-flight patterns acquired from a real sonar system are processed, demonstrating reduced error in both recognition and position estimation of objects.

Fourier Analysis↗