SEARCH · PubMed Health

Results for “radial basis function network”

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 217 records · Page 12Linked to original sources

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↗

PubMed2003

Automatic vs hand-controlled walking of paraplegics.

A rule-based control and its application in functional electrical stimulation (FES) assisted walking of subjects with paraplegia are described in this paper. The design of rules for control comprises the following two steps: (1) determination of muscle activation patterns by using a fully customized spatial (3D) model of paraplegic walking, and (2) learning of rules, that is, correlation between the muscle activation patterns and kinematics of walking by means of an artificial neural network. The adopted FES system activated eight muscle groups with surface electrodes. The only joints allowing movement in the coronal plane were the hips, and externally controlled joints in sagittal plane were ankles, knees and hips. The simulation minimized the tracking error of the joint angles and the total activation of all eight muscles being stimulated. A radial-basis function artificial neural network was applied for learning of rules. Three automatically controlled modes (slow, near-normal, and near-ballistic) and hand-controlled walking were evaluated in six subjects with a complete spinal cord lesion (T8-T10). The performance of walking was assessed by the following: (1) energy consumption based on oxygen uptake, (2) physiological cost index, (3) maximum speed of walking, and (4) a questionnaire. The results showed that all modes of walking are achievable and that automatic control leads to more efficient and faster walking. The speed of walking achieved by automatic control was almost three times bigger compared with the speed of hand-controlled walking. The energy cost and rate decreased significantly when automatic control was applied; yet, they were still much bigger than the values measured in able-bodied subjects. The objective outcome measures suggest that the near-ballistic walking was the most effective, yet a questionnaire shows that most subjects preferred slow walking. The most likely reason for the preference of lower efficiency walking over the faster end energy efficient near-ballistic walking was that paraplegic patients had difficulties in synchronizing the voluntary movement of the trunk and arms to the artificially controlled movements of legs.

Adolescent↗

PubMed2003

Neural network-based assessment of prognostic markers and outcome prediction in bilharziasis-associated bladder cancer.

In this paper the potential value of two prognostic factors, namely, bilharziasis status and tumor histological type, is investigated in relation to their abilities to predict disease progression and outcome of patients with bladder cancer, using radial basis function (RBF) neural networks. The bladder cancer data set is described by eight clinical and pathological markers. Two outcomes are of interest: either a patient is alive and free of disease or the patient is dead within five years of diagnosis. Three hundred and twenty-one (321) patients are involved in this retrospective study, 83.5% of whom had been confirmed with bilharziasis history. Selected marker subsets are examined to improve the outcome predictive accuracy and to evaluate the effects of the assessed prognostic factors on such outcome. The highest predictive accuracy for patients with bladder adenocarcinoma, as obtained from the RBF network, is found to be 85% with one subset of markers. The predictive analysis shows that bilharziasis history and patients' histology type are both important prognostic factors in prediction and, for each histology type, different marker combinations with significant characteristics have been observed.

Adult↗

PubMed1998

Improved diagnosis of breast implant rupture with sonographic findings and artificial neural networks.

RATIONALE AND OBJECTIVES: The authors evaluated the use of sonographic findings combined with artificial neural networks as an aid to the diagnosis of breast implant rupture. MATERIALS AND METHODS: From a database of 78 breast implants that were evaluated prospectively with sonography and then surgically removed, sonographic findings and surgical results were used to train and test backpropagation and radial basis function artificial neural networks by using the leave-one-out method. Receiver operating characteristic (ROC) curve analysis was used to compare the performance of the different neural networks with that of the radiologists involved. RESULTS: By using the ROC area index as a measure of performance, the artificial neural network (Az = 0.8744) outperformed the radiologists (Az = 0.8057), although not by a statistically significant difference (P = .09). The best-performing network used, in addition to the sonographic findings, the diagnosis of the radiologist as an input. This network (Az = 0.9245) outperformed both the radiologists and the "unaided" networks by a statistically significant margin (P = .02 for radiologists, P = .04 for the unaided network). The network performed remarkably well in those cases in which the radiologists classified the implant as indeterminate, predicting the correct diagnosis in 23 of 25 cases (92%). CONCLUSION: The results suggest that artificial neural networks in tandem with the unaided radiologic diagnosis can improve the accuracy rate in the detection of implant rupture based on sonographic findings. This "team" approach provided the best results.

Breast Implants↗

PubMed2003

Prediction of corneal permeability using artificial neural networks.

The purpose of this study was to develop a simple model for prediction of corneal permeability of structurally different drugs as a function of calculated molecular descriptors using artificial neural networks. A set of 45 compounds with experimentally derived values of corneal permeability (log C) was used to develop, test and validate a predictive model. Each compound was encoded with 1194 calculated molecular structure descriptors. A genetic algorithm was used to select a subset of descriptors that best describe corneal permeability coefficient log C and a supervised network with radial basis transfer function (RBF) was used to correlate calculated molecular descriptors with experimentally derived measures of corneal permeability. The best model, with 4 input descriptors and 12 hidden neurones was chosen, and the significance of the selected descriptors to corneal permeability was examined. Strong correlation of predicted with experimentally derived log C values (correlation coefficient greater than 0.87 and 0.83 respectively) was obtained for the training and testing data sets. The developed model could be useful for the rapid prediction of the corneal permeability of candidate drugs based on molecular structure alone as it does not require experimentally derived data.

Chemical Phenomena↗

PubMed2001

Theoretically-derived molecular descriptors important in human intestinal absorption.

A quantitative structure-human intestinal absorption relationship was developed using artificial neural network (ANN) modeling. A set of 86 drug compounds and their experimentally-derived intestinal absorption values used in this study was gathered from the literature and a total of 57 global molecular descriptors, including constitutional, topological, chemical, geometrical and quantum chemical descriptors, calculated for each compound. A supervised network with radial basis transfer function was used to correlate calculated molecular descriptors with experimentally-derived measures of human intestinal absorption. A genetic algorithm was then used to select important molecular descriptors. Intestinal absorption values (IA%) were used as the ANN's output and calculated molecular descriptors as the inputs. The best genetic neural network (GNN) model with 15 input descriptors was chosen, and the significance of the selected descriptors for intestinal absorption examined. Results obtained with the model that was developed indicate that lipophilicity, conformational stability and inter-molecular interactions (polarity, and hydrogen bonding) have the largest impact on intestinal absorption.

Artificial Intelligence↗

PubMed2004

Study of CNG/diesel dual fuel engine's emissions by means of RBF neural network.

Great efforts have been made to resolve the serious environmental pollution and inevitable declining of energy resources. A review of Chinese fuel reserves and engine technology showed that compressed natural gas (CNG)/diesel dual fuel engine (DFE) was one of the best solutions for the above problems at present. In order to study and improve the emission performance of CNG/diesel DFE, an emission model for DFE based on radial basis function (RBF) neural network was developed which was a black-box input-output training data model not require priori knowledge. The RBF centers and the connected weights could be selected automatically according to the distribution of the training data in input-output space and the given approximating error. Studies showed that the predicted results accorded well with the experimental data over a large range of operating conditions from low load to high load. The developed emissions model based on the RBF neural network could be used to successfully predict and optimize the emissions performance of DFE. And the effect of the DFEmain performance parameters, such as rotation speed, load, pilot quantity and injection timing, were also predicted by means of this model. In resumé, an emission prediction model for CNG/diesel DFE based on RBF neural network was built for analyzing the effect of the main performance parameters on the CO, NOx, emissions of DFE. The predicted results agreed quite well with the traditional emissions model, which indicated that the model had certain application value, although it still has some limitations, because of its high dependence on the quantity of the experimental sample data.

Energy-Generating Resources↗

PubMed2002

Classifying coronary dysfunction using neural networks through cardiovascular auscultation.

The paper applies artificial neural networks (ANNs) to the analysis of heart sound abnormalities through auscultation. Audio auscultation samples of 16 different coronary abnormalities were collected. Data pre-processing included down-sampling of the auscultated data and use of the fast Fourier transform (FFT) and the Levinson-Durbin autoregression algorithms for feature extraction and efficient data encoding. These data were used in the training of a multi-layer perceptron (MLP) and radial basis function (RBF) neural network to develop a classification mechanism capable of distinguishing between different heart sound abnormalities. The MLP and RBF networks attained classification accuracies of 84% and 88%, respectively. The application of ANNs to the analysis of respiratory auscultation and consequently the development of a combined cardio-respiratory analysis system using auscultated data could lead to faster and more efficient treatment.

Adult↗

PubMed2005

Prediction of seizure onset in an in-vitro hippocampal slice model of epilepsy using Gaussian-based and wavelet-based artificial neural networks.

We propose that artificial neural networks (ANNs) can be used to predict seizure onsets in an in-vitro hippocampal slice model capable of generating spontaneous seizure-like events (SLEs) in their extracellular field recordings. This paper assesses the effectiveness of two ANN prediction schemes: Gaussian-based artificial neural network (GANN) and wavelet-based artificial neural network (WANN). The GANN prediction system consists of a recurrent network having Gaussian radial basis function (RBF) nonlinearities capable of extracting the estimated manifold of the system. It is able to classify the underlying dynamics of spontaneous in-vitro activities into interictal, preictal and ictal modes. It is also able to successfully predict the onsets of SLEs as early as 60 s before. Improvements can be made to the overall seizure predictor design by incorporating time-varying frequency information. Consequently, the idea of WANN is considered. The WANN design entails the assumption that frequency variations in the extracellular field recordings can be used to compute the times at which onsets of SLEs are most likely to occur in the future. Progressions of different frequency components can be captured by the ANN using appropriate frequency band adjustments via pruning, after the initial wavelet transforms. In the off-line processing comprised of 102 spontaneous SLEs generated from 14 in-vitro rat hippocampal slices, with half of them used for training and the other half for testing, the WANN is able to predict the forecoming ictal onsets as early as 2 min prior to SLEs with over 75% accuracy within a 30 s precision window.

Animals↗

PubMed2007

A two-stage method for MUAP classification based on EMG decomposition.

A method for the extraction and classification of individual motor unit action potentials (MUAPs) from needle electromyographic signals is presented. The proposed method automatically decomposes MUAPs and classifies them into normal, neuropathic or myopathic using a two-stage feature-based classifier. The method consists of four steps: (i) preprocessing of EMG recordings, (ii) MUAP clustering and detection of superimposed MUAPs, (iii) feature extraction and (iv) MUAP classification using a two-stage classifier. The proposed method employs Radial Basis Function Artificial Neural Networks and decision trees. It requires minimal use of tuned parameters and is able to provide interpretation for the classification decisions. The approach has been validated on real EMG recordings and an annotated collection of MUAPs. The success rate for MUAP clustering is 96%, while the accuracy for MUAP classification is about 89%.

Action Potentials↗

PubMed2001

Life-long learning cell structures--continuously learning without catastrophic interference.

As an extension of on-line learning, life-long learning challenges a system which is exposed to patterns from a changing environment during its entire lifespan. An autonomous system should not only integrate new knowledge on-line into its memory, but also preserve the knowledge learned by previous interactions. Thus, life-long learning implies the fundamental Stability-Plasticity Dilemma, which addresses the problem of learning new patterns without forgetting old prototype patterns. We propose an extension to the known Cell Structures, growing Radial Basis Function-like networks, that enables them to learn their number of nodes needed to solve a current task and to dynamically adapt the learning rate of each node separately. As shown in several simulations, the resulting Life-long Learning Cell Structures posses the major characteristics needed to cope with the Stability-Plasticity Dilemma.

Algorithms↗

PubMed1999

Nonlinear analysis of the performance and reliability of wavelet singularity detection based denoising for Doppler ultrasound fetal heart rate signals.

Many studies on the physiology of the cardiovascular system revealed that nonlinear chaotic dynamics govern the generation of the heart rate signal. This is also valid for the fetal heart rate (FHR) variability, where however the variability is affected by many more factors and is significantly more complicated than for the adult case. Recently an adaptive wavelet denoising method for the Doppler ultrasound FHR recordings has been introduced. In this paper the performance and reliability of that method is confirmed by the observation that for the wavelet denoised FHR signal, a deterministic nonlinear structure, which was concealed by the noise, becomes apparent. It provides strong evidence that the denoising process removes actual noise components and can therefore be utilized for the improvement of the signal quality. Hence by observing after denoising a significant improvement of the 'chaoticity' of the FHR signal we obtain strong evidence for the reliability and efficiency of the wavelet based denoising method. The estimation of the chaoticity of the FHR signal before and after the denoising is approached with three nonlinear analysis methods. First, the rescaled scale analysis (RSA) technique reveals that the denoising process increases the Hurst exponent parameter as happens when additive noise is removed from a chaotic signal. Second, the nonlinear prediction error evaluated with radial basis function (RBF) prediction networks is significantly lower at the denoised signal. The significant gain in predictability can be attributed to the drastic reduction of the additive noise from the signal by the denoising algorithm. Moreover, the evaluation of the correlation coefficient between actual and neural network predicted values as a function of the prediction time displays characteristics of chaos only for the denoised signal. Third, a chaotic attractor, reconstructed with the embedding dimension technique, becomes evident for the denoised signal, while it is completely obscured for the original signals. The correlation dimension of the reconstructed attractor for the denoised signal tends to reach a value independent of the embedding dimension, a sign of deterministic chaotic signal. In contrast for the original signal the correlation dimension increases steadily with the embedding dimension, a fact that indicates strong contribution of noise.

Adult↗

PubMed2004

Bioavailability prediction based on molecular structure for a diverse series of drugs.

PURPOSE: Radial basis function artificial neural networks and theoretical descriptors were used to develop a quantitative structure-pharmacokinetic relationship for structurally diverse drug compounds. METHODS: Human bioavailability values were taken from the literature and descriptors were generated from the drug structures. All models were trained with 137 compounds and tested with a further 15, after which they were evaluated for predictive ability with an additional 15 compounds. RESULTS: The final model possessed a 10-31-1 topology and training and testing correlation coefficients were 0.736 and 0.897, respectively. Predictions for independent compounds agreed well with experimental literature values, especially for compounds that were well absorbed and/or had high observed bioavailability. Important theoretical descriptors included solubility parameters, electronic descriptors, and topological indices. CONCLUSIONS: Useful information regarding drug bioavailability was gained from drug structure alone, reducing the need for experimental methods in drug development.

Analysis of Variance↗

PubMed2003

Total luminescence spectroscopy with pattern recognition for classification of edible oils.

Total luminescence spectroscopy combined with pattern recognition has been used to discriminate between four different types of edible oils, extra virgin olive (EVO), non-virgin olive (NVO), sunflower (SF) and rapeseed (RS) oils. Simplified fuzzy adaptive resonance theory mapping (SFAM), traditional back propagation (BP) and radial basis function (RBF) neural networks provided 100% classification for 120 samples, SFAM was found to be the most efficient. The investigation was extended to the adulteration of percentage v/v SF or RS in EVO at levels from 5% to 90% creating a total of 480 samples. SFAM was found to be more accurate than RBF and BP for classification of adulterant level. All misclassifications for SFAM occurred at the 5% v/v level resulting in a total of 99.375% correctly classified oil samples. The percentage of adulteration may be described by either RBF network (2.435% RMSE) or a simple Euclidean distance relationship of the principal component analysis (PCA) scores (2.977% RMSE) for v/v RS in EVO adulteration.

Computational Biology↗

PubMed2006

Universal approximation using incremental constructive feedforward networks with random hidden nodes.

According to conventional neural network theories, single-hidden-layer feedforward networks (SLFNs) with additive or radial basis function (RBF) hidden nodes are universal approximators when all the parameters of the networks are allowed adjustable. However, as observed in most neural network implementations, tuning all the parameters of the networks may cause learning complicated and inefficient, and it may be difficult to train networks with nondifferential activation functions such as threshold networks. Unlike conventional neural network theories, this paper proves in an incremental constructive method that in order to let SLFNs work as universal approximators, one may simply randomly choose hidden nodes and then only need to adjust the output weights linking the hidden layer and the output layer. In such SLFNs implementations, the activation functions for additive nodes can be any bounded nonconstant piecewise continuous functions g : R --> R and the activation functions for RBF nodes can be any integrable piecewise continuous functions g : R --> R and integral of R g(x)dx not equal to 0. The proposed incremental method is efficient not only for SFLNs with continuous (including nondifferentiable) activation functions but also for SLFNs with piecewise continuous (such as threshold) activation functions. Compared to other popular methods such a new network is fully automatic and users need not intervene the learning process by manually tuning control parameters.

Algorithms↗

PubMed1998

Artificial neural networks for drug vulnerability recognition and dynamic scenarios simulation.

Semeion researchers have developed and used different kinds of Artificial Neural Networks (ANN) in order to process selected, "standard" data coming from drug users and from people who never used drugs before. In the first step a collection of 112 general variables, not traditionally connected to drug user's behavior, were collected from a sample of 545 people (223 heroin addicted and 322 non-users). Different types of ANNs were used to test the capability of the system to classify the drug users and the non-drug users correctly. A special ANN tool, created by Semeion, was also used to prune the number of the independent variables. The ANN selected for this first experiment was a Supervised Feed Forward Network, whose equations were enhanced by Semeion researchers. For the validation of the capability of generalization of the ANN, the Training-Testing protocol was used. This ANN was able, in the Testing phase, to classify approximately 95% of the sample with accuracy. A special sensitivity tool selected only 47 among the 112 independent variables as necessary to train the ANN. In the second step, different types of ANN were tested on the new 47 variables to decide which kind of ANN was better able to classify the sample. This benchmark included the following ANNs: a) Back Propagation with Soft Max; b) Learning Vector Quantization; c) Logicon Projection; d) Radial Basis Function; e) Squash (Semeion Network); f) Fuzzy Art Map; g) Modular Neural Network. In the third step a Constraint Satisfaction Network, specifically created by Semeion, was used to simulate a dynamic fuzzy map of the drug user's world; that is, which fuzzy, or approximate, variables are critical to decide the fuzzy membership of a subject from the fuzzy membership of the drug users to the fuzzy membership of non-users and vice versa.

Artificial Intelligence↗

PubMed1997

Classification of low back pain from dynamic motion characteristics using an artificial neural network.

STUDY DESIGN: Data were collected from 183 subjects who were randomly assigned to the training and test groups. During testing of the classification system, knowledge of the low back pain condition or motion characteristics of the patients in the test group was not made available to the system. OBJECTIVES: To determine specific characteristics of trunk motion associated with different categories of spinal disorders and to determine whether a neural network analysis system can be effective in distinguishing patterns. SUMMARY OF BACKGROUND DATA: Numerous studies have established the difficulty of evaluating lower back pain. Imaging techniques are expensive and ineffective in many cases. A technique for evaluation of lower back pain was developed on the basis of analysis of such dynamic motion features as shape, velocity, and symmetry of movements, using a neural network classification system. METHODS: Dynamic motion data were collected from 183 subjects using a triaxial goniometer. Features of the movement were extracted and provided as input to a two-stage neural network classifier governed by a radial basis function architecture. After training, the output of the classifier was compared with Québec Task Force pain classifications obtained for the patients. Linear and nonlinear classification techniques were compared. RESULTS: The system could determine low back pain classification from motion characteristics. The neural network classifier produced the best results with up to 85% accuracy on novel "validation" data. CONCLUSIONS: A neural network based on kinematic data is an excellent predictive model for classification of lower back pain. Such a system could markedly improve the management of lower back pain in the individual patient.

Adult↗

PubMed1997

Training Dynamics and Neural Network Performance.

We use an analysis of a simple model of recurrent network dynamics to gain qualitative insights into the training dynamics of feedforward multilayer perceptrons (MLPs) used for classification. These insights suggest changes to the training methods used for MLPs that improve network performance significantly. In previous work, the probabilistic neural network (PNN) was shown to provide better zero-reject error performance on character and fingerprint classification problems than radial basis function and MLP-based neural network methods. We will show that performance equal to or better than PNN can be achieved with a single three-layer MLP by making fundamental changes in the network optimization strategy. These changes are: 1) use of neuron activation functions, which reduce the probability of singular Jacobians; 2) use of successive regularization to constrain the volume of the minimized weight space; 3) use of Boltzmann pruning to constrain the dimension of the weight space; 4) use of Prior class probabilities to normalize all error calculations, so that statistically significant samples of rare but important classes can be included without distorting the error surface. All four of these changes are made in the inner loop of a conjugate gradient optimization iteration and are intended to simplify the training dynamics of the optimization. On handprinted digits and fingerprint classification problems these modifications improve error-reject performance by factors between 2 and 4, and reduce network size by 40-60%. Copyright 1997 Elsevier Science Ltd.

Journal Article↗