SEARCH · PubMed Health

Results for “Multilayer Perceptrons”

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 91 records · Page 5Linked to original sources

PubMed1999

Higher-order probabilistic perceptrons as Bayesian inference engines.

An explicit structural connection is established between the Bayes optimal classifier operating on K binary input variables and a corresponding two-layer perceptron having normalized output activities and couplings from input to output units of all orders up to K. With suitable modification of connection weights and biases, such a higher-order probabilistic perceptron should in principle be able to learn the statistics of the classification problem and match the a posteriori probabilities given by Bayes optimal inference. Specific training algorithms are developed that allow this goal to be approximated in a controlled variational sense. An application to the task of discriminating between stable and unstable nuclides in nuclear physics yields network models with predictive performance comparable to the best that has been achieved with conventional multilayer perceptrons containing only pairwise connections.

Journal Article↗

PubMed2005

Automatic detection of epileptiform events in EEG by a three-stage procedure based on artificial neural networks.

This paper introduces a three-stage procedure based on artificial neural networks for the automatic detection of epileptiform events (EVs) in a multichannel electroencephalogram (EEG) signal. In the first stage, two discrete perceptrons fed by six features are used to classify EEG peaks into three subgroups: 1) definite epileptiform transients (ETs); 2) definite non-ETs; and 3) possible ETs and possible non-ETs. The pre-classification done in the first stage not only reduces the computation time but also increases the overall detection performance of the procedure. In the second stage, the peaks falling into the third group are aimed to be separated from each other by a nonlinear artificial neural network that would function as a postclassifier whose input is a vector of 41 consecutive sample values obtained from each peak. Different networks, i.e., a backpropagation multilayer perceptron and two radial basis function networks trained by a hybrid method and a support vector method, respectively, are constructed as the postclassifier and then compared in terms of their classification performances. In the third stage, multichannel information is integrated into the system for contributing to the process of identifying an EV by the electroencephalographers (EEGers). After the integration of multichannel information, the overall performance of the system is determined with respect to EVs. Visual evaluation, by two EEGers, of 19 channel EEG records of 10 epileptic patients showed that the best performance is obtained with a radial basis support vector machine providing an average sensitivity of 89.1%, an average selectivity of 85.9%, and a false detection rate (per hour) of 7.5.

Adolescent↗

PubMed2001

Artificial neural networks: opening the black box.

Artificial neural networks now are used in many fields. They have become well established as viable, multipurpose, robust computational methodologies with solid theoretic support and with strong potential to be effective in any discipline, especially medicine. For example, neural networks can extract new medical information from raw data, build computer models that are useful for medical decision-making, and aid in the distribution of medical expertise. Because many important neural network applications currently are emerging, the authors have prepared this article to bring a clearer understanding of these biologically inspired computing paradigms to anyone interested in exploring their use in medicine. They discuss the historical development of neural networks and provide the basic operational mathematics for the popular multilayered perceptron. The authors also describe good training, validation, and testing techniques, and discuss measurements of performance and reliability, including the use of bootstrap methods to obtain confidence intervals. Because it is possible to predict outcomes for individual patients with a neural network, the authors discuss the paradigm shift that is taking place from previous "bin-model" approaches, in which patient outcome and management is assumed from the statistical groups in which the patient fits. The authors explain that with neural networks it is possible to mediate predictions for individual patients with prevalence and misclassification cost considerations using receiver operating characteristic methodology. The authors illustrate their findings with examples that include prostate carcinoma detection, coronary heart disease risk prediction, and medication dosing. The authors identify and discuss obstacles to success, including the need for expanded databases and the need to establish multidisciplinary teams. The authors believe that these obstacles can be overcome and that neural networks have a very important role in future medical decision support and the patient management systems employed in routine medical practice.

Decision Theory↗

PubMed2001

Impact of different variables on the outcome of patients with clinically confined prostate carcinoma: prediction of pathologic stage and biochemical failure using an artificial neural network.

BACKGROUND: The advent of advanced computing techniques has provided the opportunity to analyze clinical data using artificial intelligence techniques. This study was designed to determine whether a neural network could be developed using preoperative prognostic indicators to predict the pathologic stage and time of biochemical failure for patients who undergo radical prostatectomy. METHODS: The preoperative information included TNM stage, prostate size, prostate specific antigen (PSA) level, biopsy results (Gleason score and percentage of positive biopsy), as well as patient age. All 309 patients underwent radical prostatectomy at the University of Colorado Health Sciences Center. The data from all patients were used to train a multilayer perceptron artificial neural network. The failure rate was defined as a rise in the PSA level > 0.2 ng/mL. The biochemical failure rate in the data base used was 14.2%. Univariate and multivariate analyses were performed to validate the results. RESULTS: The neural network statistics for the validation set showed a sensitivity and specificity of 79% and 81%, respectively, for the prediction of pathologic stage with an overall accuracy of 80% compared with an overall accuracy of 67% using the multivariate regression analysis. The sensitivity and specificity for the prediction of failure were 67% and 85%, respectively, demonstrating a high confidence in predicting failure. The overall accuracy rates for the artificial neural network and the multivariate analysis were similar. CONCLUSIONS: Neural networks can offer a convenient vehicle for clinicians to assess the preoperative risk of disease progression for patients who are about to undergo radical prostatectomy. Continued investigation of this approach with larger data sets seems warranted.

Adult↗

PubMed2002

Hybrid model building methodology using unsupervised fuzzy clustering and supervised neural networks.

This paper suggests a model building methodology for dealing with new processes. The methodology, called Hybrid Fuzzy Neural Networks (HFNN), combines unsupervised fuzzy clustering and supervised neural networks in order to create simple and flexible models. Fuzzy clustering was used to define relevant domains on the input space. Then, sets of multilayer perceptrons (MLP) were trained (one for each domain) to map input-output relations, creating, in the process, a set of specified sub-models. The estimated output of the model was obtained by fusing the different sub-model outputs weighted by their predicted possibilities. On-line reinforcement learning enabled improvement of the model. The determination of the optimal number of clusters is fundamental to the success of the HFNN approach. The effectiveness of several validity measures was compared to the generalization capability of the model and information criteria. The validity measures were tested with fermentation simulations and real fermentations of a yeast-like fungus, Aureobasidium pullulans. The results outline the criteria limitations. The learning capability of the HFNN was tested with the fermentation data. The results underline the advantages of HFNN over a single neural network.

Biotechnology↗

PubMed1993

Application of artificial neural networks to chromosome classification.

This work presents an approach to the automatic classification of metaphase chromosomes using a multilayer perceptron neural network. Representation of the banding patterns by intuitively defined features is avoided. The inputs to the network are the chromosome size and centromeric index and a coarsely quantized representation of the chromosome banding profile. We demonstrate that following a fairly mechanical training procedure, the classification performance of the network compares favourably with a well-developed parametric classifier. The sensitivity of the network performance to variation in network parameters is investigated, and we show that a gain in efficiency is obtainable by an appropriate decomposition of the network. We discuss the flexibility of the classifier developed, its potential for enhancement, and how it may be adapted to suit the needs of current trends in karyotyping.

Algorithms↗

PubMed2005

Fast robust subject-independent magnetoencephalographic source localization using an artificial neural network.

We describe a system that localizes a single dipole to reasonable accuracy from noisy magnetoencephalographic (MEG) measurements in real time. At its core is a multilayer perceptron (MLP) trained to map sensor signals and head position to dipole location. Including head position overcomes the previous need to retrain the MLP for each subject and session. The training dataset was generated by mapping randomly chosen dipoles and head positions through an analytic model and adding noise from real MEG recordings. After training, a localization took 0.7 ms with an average error of 0.90 cm. A few iterations of a Levenberg-Marquardt routine using the MLP output as its initial guess took 15 ms and improved accuracy to 0.53 cm, which approaches the natural limit on accuracy imposed by noise. We applied these methods to localize single dipole sources from MEG components isolated by blind source separation and compared the estimated locations to those generated by standard manually assisted commercial software.

Artifacts↗

PubMed2004

Algorithms based on prostate-specific antigen (PSA), free PSA, digital rectal examination and prostate volume reduce false-positive PSA results in prostate cancer screening.

Our objective was to determine whether multivariate algorithms based on serum total PSA, the free proportion of PSA, age, digital rectal examination and prostate volume can reduce the rate of false-positive PSA results in prostate cancer screening more effectively than the proportion of free PSA alone at 95% sensitivity. A total of 1,775 consecutive 55- to 67-year-old men with a serum PSA of 4-10 microg/l in the European Randomized Study of Screening for Prostate Cancer were included. To predict the presence of cancer, multivariate algorithms were constructed using logistic regression (LR) and a multilayer perceptron neural network with Bayesian regularization (BR-MLP). A prospective setting was simulated by dividing the data set chronologically into one set for training and validation (67%, n = 1,183) and one test set (33%, n = 592). The diagnostic models were calibrated using the training set to obtain 95% sensitivity. When applied to the test set, the LR model, the BR-MLP model and the proportion of free PSA reached 92%, 87% and 94% sensitivity and reduced 29%, 36% and 22% of the false-positive PSA results, respectively. At a fixed sensitivity of 95% in the test set, the LR model eliminated more false-positive PSA results (22%) than the proportion of free PSA alone (17%) (p < 0.001), whereas the BR-MLP model did not (19%) (p = 0.178). The area under the ROC curve was larger for the LR model (0.764, p = 0.030) and the BR-MLP model (0.760, p = 0.049) than for the proportion of free PSA (0.718). A multivariate algorithm can be used to reduce unnecessary prostate biopsies in screening more effectively than the proportion of free PSA alone, but the algorithms will require updating when clinical practice develops with time.

Aged↗

PubMed2005

Application of artificial neural networks for gradient elution retention modelling in ion chromatography.

Gradient elution in ion chromatography (IC) offers several advantages: total analysis time can be significantly reduced, overall resolution of a mixture can be increased, peak shape can be improved (less tailing) and effective sensitivity can be increased (because there is little variation in peak shape). More importantly, it provides the maximum resolution per time unit. The aim of this work was the development of a suitable artificial neural network (ANN) gradient elution retention model that can be used in a variety of applications for method development and retention modelling of inorganic anions in IC. Multilayer perceptron ANNs were used to model the retention behaviour of fluoride, chloride, nitrite, sulphate, bromide, nitrate and phosphate in relation to the starting time of gradient elution and the slope of the linear gradient elution curve. The advantage of the developed model is the application of an optimized two-phase training algorithm that enables the researcher to make use of the advantages of first- and second-order training algorithms in one training procedure. This results in better predictive ability, with less time required for the calculations. The number of hidden layer neurons and experimental data points used for the training set were optimized in terms of obtaining a precise and accurate retention model with respect to minimization of unnecessary experimentation and time needed for the calculation procedures. This study shows that developed, ANNs are the method of first choice for retention modelling of inorganic anions in IC.

Journal Article↗

PubMed2003

Machine learning approaches to lung cancer prediction from mass spectra.

We addressed the problem of discriminating between 24 diseased and 17 healthy specimens on the basis of protein mass spectra. To prepare the data, we performed mass to charge ratio (m/z) normalization, baseline elimination, and conversion of absolute peak height measures to height ratios. After preprocessing, the major difficulty encountered was the extremely large number of variables (1676 m/z values) versus the number of examples (41). Dimensionality reduction was treated as an integral part of the classification process; variable selection was coupled with model construction in a single ten-fold cross-validation loop. We explored different experimental setups involving two peak height representations, two variable selection methods, and six induction algorithms, all on both the original 1676-mass data set and on a prescreened 124-mass data set. Highest predictive accuracies (1-2 off-sample misclassifications) were achieved by a multilayer perceptron and Naïve Bayes, with the latter displaying more consistent performance (hence greater reliability) over varying experimental conditions. We attempted to identify the most discriminant peaks (proteins) on the basis of scores assigned by the two variable selection methods and by neural network based sensitivity analysis. These three scoring schemes consistently ranked four peaks as the most relevant discriminators: 11683, 1403, 17350 and 66107.

Algorithms↗

PubMed1998

Artificial neural networks as models of stimulus control.

We evaluate the ability of artificial neural network models (multilayer perceptrons) to predict stimulus-response relationships. A variety of empirical results are considered, such as generalization, peak shift (supernormality) and stimulus intensity effects. The networks were trained on the same tasks as the animals in the experiments considered. The subsequent generalization tests on the networks showed that the model replicates correctly the empirical results. We conclude that these models are valuable tools in the study of animal behaviour. (c) 1998 The Association for the Study of Animal Behaviour.

Journal Article↗

PubMed1991

Saccade control in a simulated robot camera-head system: neural net architectures for efficient learning of inverse kinematics.

The high speed of saccades means that they cannot be guided by visual feedback, so that any saccadic control system must know in advance the correct output signals to fixate a particular retinal position. To investigate neural-net architectures for learning this inverse-kinematics problem we simulated a 4 deg-of-freedom robot camera-head system, in which the head could pan and tilt and the cameras pan and verge. The main findings were: (1) Linear nets, multilayer perceptrons (MLPs) trained by backpropagation, and cerebellar model arithmetic computers (CMACs) all learnt rapidly to 5-10% accuracy when given perfect error feedback. (2) For additional accuracy (down to 2%) two-layer nets learnt much faster than a single MLP or CMAC: the best combination tried was to have a CMAC learn the errors of a trained linear net. (3) Imperfect error signals were provided by a crude controller whose output was simply proportional to retinal input in the relevant axis, thereby providing a mechanism for (a) controlling the camera-head system when the feedforward neural net controller was wrong or inoperative, and (b) converting sensory error signals into motor error signals as required in supervised learning. It proved possible to train neural-net controllers using these imperfect error signals over a range of learning rates and crude-controller gains. These results suggest that appropriate neural-net architectures can provide practical, accurate and robust adaptive control for saccadic movements. In addition, the arrangement of a crude controller teaching a sophisticated one may be similar to that used by the primate saccadic system, with brainstem circuitry teaching the cerebellum.

Brain Stem↗

PubMed2004

Prediction of the axillary lymph node status in mammary cancer on the basis of clinicopathological data and flow cytometry.

Axillary lymph node status is a major prognostic factor in mammary carcinoma. It is clinically desirable to predict the axillary lymph node status from data from the mammary cancer specimen. In the study, the axillary lymph node status, routine histological parameters and flow-cytometric data were retrospectively obtained from 1139 specimens of invasive mammary cancer. The ten variables: age, tumour type, tumour grade, tumour size, skin infiltration, lymphangiosis carcinomatosa, pT4 category, percentage of tumour cells in G2/M- and S-phases of the cell cycle, and ploidy index were considered as predictor variables, and the single variable lymph node metastasis pN (0 for pN0, or 1 for pN1 or pN2) was used as an output variable. A stepwise logistic regression analysis, with the axillary lymph node as a dependent variable, was used for feature selection. Only lymphangiosis carcinomatosa and tumour size proved to be significant as independent predictor variables; the other variables were non-contributory. Three paradigms with supervised learning rules (multilayer perceptron, learning vector quantisation and support vector machines) were used for the purpose of prediction. If any of these paradigms was used with the information from all ten input variables, 73% of cases could be correctly predicted, with specificity ranging from 82 to 84% and sensitivity ranging from 60 to 63%. If only the two significant input variables were used, lymphangiosis carcinomatosa and tumour diameter, the prediction accuracy was no worse. Nearly identical results were obtained by two different techniques of cross-validation (leave-one-out against ten-fold cross validation). It was concluded that: artificial neural networks can be used for risk stratification on the basis of routine data in individual cases of mammary cancer; and lymphangiosis carcinomatosa and tumour size are independent predictors of axillary lymph node metastasis in mammary cancer.

Algorithms↗

PubMed1995

Catheter-manometer system damped blood pressures detected by neural nets.

Degraded catheter-manometer systems cause distortion of blood pressure waveforms, often leading to erroneously resonant or damped waveforms, requiring waveforms quality control. We have tried multilayer perceptron back-propagation trained neural nets of varying architecture to detect damping on sets of normal and artificially damped brachial arterial pressure waves. A second-order digital simulation of a catheter-manometer system is used to cause waveform distortion. Each beat in the waveforms is represented by an 11 parameter input vector. From a group of normotensive or (borderline) hypertensive subjects, pressure waves are used to statistically test and train the neural nets. For each patient and category 5-10 waves are available. The best neural nets correctly classify about 75-85% of the individual beats as either adequate or damped. Using a single majority vote classification per subject per damped or adequate situation, the best neural nets correctly classify at least 16 of the 18 situations in nine test subjects (binomial P = 0.001). More importantly, these neural nets can always detect damping before clinically relevant parameters such as systolic pressure and computed stroke volume are reduced by more than 2%. Neural nets seem remarkably well adapted to solving such subtle problems as detecting a slight damping of arterial pressure waves before it affects waveforms to a clinically relevant degree.

Adult↗

PubMed2000

Prediction of bone strength from cancellous structure of the distal radius: can we improve on DXA?

Recent studies show that structural parameters of bone, obtained from computerized image analysis of radiographs, can improve the noninvasive determination of bone strength when used in conjunction with bone density measurements. The present study was designed to assess the ability of image features alone to predict the mechanical characteristics of bones. A multifactorial model was used to incorporate simultaneously a number of characteristics of the image, including periodicity and spatial orientation of the trabeculae. Fifteen pairs (29 specimens) of unembalmed human distal radii were used. The cancellous bone structure was determined using computerized spectral analysis of their radiographic images and the bones were tested to failure under compression. Multilayered perceptron neural networks were used to integrate the various image parameters reflecting the periodicity and the spatial distribution of the trabeculae and to predict the mechanical strength of the specimens. The correlation between each of the isolated image parameters and bone strength was generally significant, but weak. The values of mechanical parameters predicted by the neural networks, however, had a very high correlation with those observed, namely 0.91 for the load at fracture and 0.93 for the ultimate stress. Both these correlations were superior to those obtained with dual-energy X-ray absorptiometry and with the cross-sectional area from CT scans: 0.87 and 0.49 respectively. Our observation suggests that image parameters can provide a powerful noninvasive predictor of bone strength. The simultaneous use of various parameters substantially improved the performance of the system. The multifactorial architecture applied is nonlinear and possibly more effective than traditional multicorrelation methods. Further, this system has the potential to incorporate other non-image parameters, such as age and bone density itself, with a view to improving the assessment of the risk of fracture for individual patients.

Absorptiometry, Photon↗

PubMed2005

The analysis of seasonal air pollution pattern with application of neural networks.

Air pollution monitoring includes measuring the concentrations of air contaminants such as nitrogen dioxide, sulfur dioxide, some polycyclic aromatic hydrocarbons (PAHs), suspended particulate matter (PM) and tar substances. The purpose of this study was to determine the possibility of using artificial neural networks for identification of any patterns occurring during heating and nonheating seasons. The samples included in the study were collected over a period of 5 years (1997-2001) in the area of the city of Gdansk and the levels of pollutants measured in the samples collected were used as inputs to two different types of neural networks: multilayer perceptron (MLP) and self-organizing map (SOM). The MLP was used as a tool to predict in what heating season a certain sample was collected, and the SOM was applied for mapping all samples to recognize any similarities between them. This study also presents the comparison between two projection methods--linear (principal component analysis, PCA) and nonlinear (SOM)--in extracting valuable information from multidimensional environmental data. In the research the MLP model with 13-12-1 topology was developed and successfully trained for classification of air samples from different seasons. The sensitivity analysis on the inputs to the MLP indicated benz[alpha]anthracene, benzo[alpha]pyrene, PM1, SO2, tar substances and PM10 as the most distinctive variables, while PCA pointed to PAHs and PM1.

Journal Article↗

PubMed2005

Learning machines applied to potential forest distribution.

The clearing of forests to obtain land for pasture and agriculture and the replacement of autochthonous species by other faster-growing varieties of trees for timber have both led to the loss of vast areas of forest worldwide. At present, many developed countries are attempting to reverse these effects, establishing policies for the restoration of older woodland systems. Reforestation is a complex matter, planned and carried out by experts who need objective information regarding the type of forest that can be sustained in each area. This information is obtained by drawing up feasibility models constructed using statistical methods that make use of the information provided by morphological and environmental variables (height, gradient, rainfall, etc.) that partially condition the presence or absence of a specific kind of forestation in an area. The aim of this work is to construct a set of feasibility models for woodland located in the basin of the River Liébana (NW Spain), to serve as a support tool for the experts entrusted with carrying out the reforestation project. The techniques used are multilayer perceptron neural networks and support vector machines. Their results will be compared to the results obtained by traditional techniques (such as discriminant analysis and logistic regression) by measuring the degree of fit between each model and the existing distribution of woodlands. The interpretation and problems of the feasibility models are commented on in the Discussion section.

Artificial Intelligence↗

PubMed2006

Neural networks as a tool for control and management of a biological reactor for treating hydrogen sulphide.

Based on an experimental database consisting of 194 daily cases, artificial neural networks were used to model the removal efficiency of a biofilter for treating hydrogen sulphide (H2S). In this work, the removal efficiency of the reactor was considered as a function of the changes in the air flow and concentration of H2S entering the biofilter. In order to obtain true representative values, the removal efficiencies (outputs) were measured 24 h after each input was changed. A MLP (multilayer perceptron 2-2-1) model with two input variables (unit flow and concentration of the contaminant fed into the biofilter) rendered good prediction values with a determination coefficient of 0.92 for the removal efficiency within the range studied. This means that the MLP model can explain 92% of the overall variability detected in the biofilter corresponding to a wide range of operating conditions.

Air Pollutants↗