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 289 records · Page 16Linked to original sources

PubMed1999

Computer-supported diagnosis of melanoma in profilometry.

Laser profilometry offers new possibilities to improve non-invasive tumor diagnostics in dermatology. In this paper, a new approach to computer-supported analysis and interpretation of high-resolution skin-surface profiles of melanomas and nevocellular nevi is presented. Image analysis methods are used to describe the profile's structures by texture parameters based on co-occurrence matrices, features extracted from the Fourier power spectrum, and fractal features. Different feature selection strategies, including genetic algorithms, are applied to determine the best possible subsets of features for the classification task. Several architectures of multilayer perceptrons with error back-propagation as learning paradigm are trained for the automatic recognition of melanomas and nevi. Furthermore, network-pruning algorithms are applied to optimize the network topology. In the study, the best neural classifier showed an error rate of 4.5% and was obtained after network pruning. The smallest error rate in all, of 2.3%, was achieved with nearest neighbor classification.

Diagnosis, Differential↗

PubMed1999

Estimation of prostate cancer probability by logistic regression: free and total prostate-specific antigen, digital rectal examination, and heredity are significant variables.

BACKGROUND: Despite low specificity, serum prostate-specific antigen (PSA) is widely used in screening for prostate cancer. Specificity can be improved by measuring free and total PSA and by combining these results with clinical findings. Methods such as neural networks and logistic regression are alternatives to multistep algorithms for clinical use of the combined findings. METHODS: We compared multilayer perceptron (MLP) and logistic regression (LR) analysis for predicting prostate cancer in a screening population of 974 men, ages 55-66 years. The study sample comprised men with PSA values >3 microg/L. Explanatory variables considered were age, free and total PSA and their ratio, digital rectal examination (DRE), transrectal ultrasonography, and a family history of prostate cancer. RESULTS: When at least 90% sensitivity in the training sets was required, the mean sensitivity and specificity obtained were 87% and 41% with LR and 85% and 26% with MLP, respectively. The cancer specificity of an LR model comprising the proportion of free to total PSA, DRE, and heredity as explanatory variables was significantly better than that of total PSA and the proportion of free to total PSA (P <0.01, McNemar test). The proportion of free to total PSA, DRE, and heredity were used to prepare cancer probability curves. CONCLUSION: The probability calculated by logistic regression provides better diagnostic accuracy for prostate cancer than the presently used multistep algorithms for estimation of the need to perform biopsy.

Blood Proteins↗

PubMed1999

Evaluating variable selection methods for diagnosis of myocardial infarction.

This paper evaluates the variable selection performed by several machine-learning techniques on a myocardial infarction data set. The focus of this work is to determine which of 43 input variables are considered relevant for prediction of myocardial infarction. The algorithms investigated were logistic regression (with stepwise, forward, and backward selection), backpropagation for multilayer perceptrons (input relevance determination), Bayesian neural networks (automatic relevance determination), and rough sets. An independent method (self-organizing maps) was then used to evaluate and visualize the different subsets of predictor variables. Results show good agreement on some predictors, but also variability among different methods; only one variable was selected by all models.

Algorithms↗

PubMed2000

Three-dimensional ultrasound analyses of the prostate.

Although conventional ultrasonography has proven to be clinically useful for depicting many types of cancerous lesions, it cannot distinguish reliably between cancerous and noncancerous tissue of the prostate. Therefore, conventional transrectal ultrasonography (TRUS) is used primarily for general evaluations of the gland and for guiding biopsies based on clearly imaged anatomic features such as the capsule, seminal vesicles, and urethra. Spectrum analysis extracts ultrasound signal parameters associated with biopsy-proven tissue types, and these parameters are then classified using neural network tools such as learning vector quantization, radial basis, and multilayer perceptron algorithms. Classification of cancerous and noncancerous prostate tissue using neural networks produces receiver operating characteristic (ROC) curves of 0.87 +/- 0.04 compared with 0.64 +/- 0.04 for conventional ultrasonography. To image the prostate using these methods, parameter values are computed at each pixel location, then translated into a score for the likelihood of cancer using a look-up table generated using the best classification algorithm. The score for cancer likelihood is expressed as a gray-scale or color value, and the resulting image may be useful to guide biopsies or therapy. Changes in parameter or score values over time potentially can be used to assess progression of disease or efficacy of therapy.

Adenocarcinoma↗

PubMed2000

Neural network prediction of polyproline type II secondary structures.

This study considers detection of polyproline type II secondary structures from protein sequences. This difficult problem was handled with multilayer perceptron neural networks, which were found to be useful for such bioinformatics studies. Polyproline II secondary structures have not previously been tried to be predicted from sequences.

Humans↗

PubMed2001

Absolute activity quantitation in simultaneous 123I/99mTc brain SPECT.

UNLABELLED: Dual-isotope imaging can allow simultaneous assessment of brain perfusion using a 99mTc-labeled tracer and neurotransmission using an 123I-labeled tracer. However, the images are affected by scatter, cross talk, attenuation, distance-dependent collimator response (DCR), and partial-volume effect. We determined the accuracy and precision of activity quantitation in simulated normal and pathologic studies of simultaneous 123I/99mTc brain SPECT when compensating for all degrading phenomena. METHODS: Monte Carlo simulations were performed using the Zubal brain phantom. Contamination caused by high-energy 123I decay photons was incorporated. Twenty-four 99mTc and 123I activity distributions were simulated on the basis of normal and pathologic patient activity distributions. Cross talk and scatter were corrected using a new method based on a multilayer perceptron artificial neural network (ANN), as well as by the asymmetric window (AW) approach; for comparison, unscattered (U) photons of 99mTc and 123I were recorded. Nonuniform attenuation and DCR were modeled in an iterative ordered-subset expectation maximization (OSEM) algorithm. Mean percentage biases and SDs over the 12 normal and 12 pathologic simulated studies were computed for each structure with respect to the known activity distributions. RESULTS: For 123I, AW + OSEM yielded a bias of 7% in the cerebellum, 21% in the frontal cortex, and 36% in the corpus callosum in the simulated normal population. The bias was increased significantly in the striata of simulated pathologic studies (P < 0.05). The bias associated with ANN was significantly lower (<9% in these brain structures, P < 0.05). For 99mTc with AW + OSEM, the bias was 60% in the corpus callosum, 36% in the striata, and 18%-22% in the cortical lobes in the simulated normal population. This bias was <11% in all brain structures with ANN. In the simulated pathologic population, the bias associated with AW increased significantly in the cortical lobes to 55% (P < 0.05), although it did not change significantly with ANN. CONCLUSION: The accuracy and variability over simulated normal and pathologic studies of both 99mTc and 123I activity estimates were very close with ANN to those obtained with U + OSEM. ANN + OSEM is a promising approach for absolute activity quantitation in simultaneous 99mTc/123I SPECT.

Alzheimer Disease↗

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↗

PubMed2001

On classification capability of neural networks: a case study with otoneurological data.

We investigated the capability of multilayer perceptron neural networks and Kohonen neural networks to recognize difficult otoneurological diseases from each other. We found that they are efficient methods, but the distribution of a learning set should be rather uniform. Also it is important that the number of learning cases is sufficient. If the two mentioned conditions are satisfied, these neural networks are similarly efficient as some other machine learning methods. The conditions are known in the theory of neural networks [1,2], but not often taken seriously in practice. Both networks functioned as well, excluding the case with several input variables, where the Kohonen neural networks surpassed the perceptron.

Algorithms↗

PubMed2001

Neural network and wavelet recognition of facial electromyographic signals.

The present aim was to explore the possibilities of using neural networks for recognizing significant changes in electrical activity of human facial muscles. We used multilayer perceptron neural networks to recognize bursts of electromyographic signals recorded with bipolar surface electrodes from two subject's facial muscles. Wavelets were applied for the detection of high frequency components of electromyographic signals. Coefficients of wavelets were used as an input to a neural network in order to differentiate bursts from the signals. The results showed that the recognition of bursts was very successful resulting to 84-97 percent total accuracies. The results were very encouraging and suggest further that the measurement of facial muscle activity may be a potentially useful computer input signal, for example, for affective computing which can be seen as a future versatile interaction between the computer and the user.

Electromyography↗

PubMed2002

Comparing machine learning classifiers for diagnosing glaucoma from standard automated perimetry.

PURPOSE: To determine which machine learning classifier learns best to interpret standard automated perimetry (SAP) and to compare the best of the machine classifiers with the global indices of STATPAC 2 and with experts in glaucoma. METHODS: Multilayer perceptrons (MLP), support vector machines (SVM), mixture of Gaussian (MoG), and mixture of generalized Gaussian (MGG) classifiers were trained and tested by cross validation on the numerical plot of absolute sensitivity plus age of 189 normal eyes and 156 glaucomatous eyes, designated as such by the appearance of the optic nerve. The authors compared performance of these classifiers with the global indices of STATPAC, using the area under the ROC curve. Two human experts were judged against the machine classifiers and the global indices by plotting their sensitivity-specificity pairs. RESULTS: MoG had the greatest area under the ROC curve of the machine classifiers. Pattern SD (PSD) and corrected PSD (CPSD) had the largest areas under the curve of the global indices. MoG had significantly greater ROC area than PSD and CPSD. Human experts were not better at classifying visual fields than the machine classifiers or the global indices. CONCLUSIONS: MoG, using the entire visual field and age for input, interpreted SAP better than the global indices of STATPAC. Machine classifiers may augment the global indices of STATPAC.

Diagnosis, Computer-Assisted↗

PubMed2002

Development and validation of a method for using breast core needle biopsies for gene expression microarray analyses.

PURPOSE: Gene expression microarray technologies have the potential to define molecular profiles that may identify specific phenotypes(diagnosis), establish a patient's expected clinical outcome (prognosis), and indicate the likelihood of a beneficial effect of a specific therapy (prediction). We wished to develop optimal tissue acquisition, processing, and analysis procedures for exploring the gene expression profiles of breast core needle biopsies representing cancer and noncancer tissues. EXPERIMENTAL DESIGN: Human breast cancer xenografts were used to evaluate several processing methods for prospectively collecting adequate amounts of high-quality RNA for gene expression microarray studies. Samples were assessed for the preservation of tissue architecture and the quality and quantity of RNA recovered. An optimized protocol was applied to a small study of core needle breast biopsies from patients, in which we compared the molecular profiles from cancer with those from noncancer biopsies. Gene expression data were obtained using Research Genetics, Inc. Named Genes cDNA microarrays. Data were visualized using simple hierarchical clustering and a novel principal component analysis-based multidimensional scaling. Data dimensionality was reduced by simple statistical approaches. Predictive neural networks were built using a multilayer perceptron and evaluated in an independent data set from snap-frozen mastectomy specimens. RESULTS: Processing tissue through RNALater preserves tissue architecture when biopsies are washed for 5 min on ice with ice-cold PBS before histopathological analysis. Cell margins are clear, tissue folding and fragmentation are not observed, and integrity of the cores is maintained, allowing optimal pathological interpretation and preservation of important diagnostic information. Adequate concentrations of high-quality RNA are recovered; 51 of 55 biopsies produced a median of 1.34 microg of total RNA (range, 100 ng to 12.60 microg). Snap-freezing or the use of RNALater does not affect RNA recovery or the molecular profiles obtained from biopsies. The neural network predictors accurately discriminate between predominantly cancer and noncancer breast biopsies. CONCLUSIONS: The approaches generated in these studies provide a simple, safe, and effective method for prospectively acquiring and processing breast core needle biopsies for gene expression studies. Gene expression data from these studies can be used to build accurate predictive models that separate different molecular profiles. The data establish the use and effectiveness of these approaches for future prospective studies.

Animals↗

PubMed2002

Comparing neural networks and linear discriminant functions for glaucoma detection using confocal scanning laser ophthalmoscopy of the optic disc.

PURPOSE: To determine whether neural network techniques can improve differentiation between glaucomatous and nonglaucomatous eyes, using the optic disc topography parameters of the Heidelberg Retina Tomograph (HRT; Heidelberg Engineering, Heidelberg, Germany). METHODS: With the HRT, one eye was imaged from each of 108 patients with glaucoma (defined as having repeatable visual field defects with standard automated perimetry) and 189 subjects without glaucoma (no visual field defects with healthy-appearing optic disc and retinal nerve fiber layer on clinical examination) and the optic nerve topography was defined by 17 global and 66 regional HRT parameters. With all the HRT parameters used as input, receiver operating characteristic (ROC) curves were generated for the classification of eyes, by three neural network techniques: linear and Gaussian support vector machines (SVM linear and SVM Gaussian, respectively) and a multilayer perceptron (MLP), as well as four previously proposed linear discriminant functions (LDFs) and one LDF developed on the current data with all HRT parameters used as input. RESULTS: The areas under the ROC curves for SVM linear and SVM Gaussian were 0.938 and 0.945, respectively; for MLP, 0.941; for the current LDF, 0.906; and for the best previously proposed LDF, 0.890. With the use of forward selection and backward elimination optimization techniques, the areas under the ROC curves for SVM Gaussian and the current LDF were increased to approximately 0.96. CONCLUSIONS: Trained neural networks, with global and regional HRT parameters used as input, improve on previously proposed HRT parameter-based LDFs for discriminating between glaucomatous and nonglaucomatous eyes. The performance of both neural networks and LDFs can be improved with optimization of the features in the input. Neural network analyses show promise for increasing diagnostic accuracy of tests for glaucoma.

Discriminant Analysis↗

PubMed1992

A processing strategy for automated Papanicolaou smear screening.

A multilayer processing strategy was developed for the automatic screening of conventionally prepared Papanicolaou smears. The processing stages include image segmentation, feature extraction, object classification and slide classification. Mathematical morphology functions were implemented in hardware with custom-built gate array processors for image segmentation. There were 68 features used for classifier training. In object classification we combined the evidential supports of a binary decision tree classifier and a multilayer perceptron classifier to achieve an integrated decision. In this feasibility study, 449 conventionally prepared cervical Papanicolaou smears were tested in a prototype research system between January and May 1991. The 95% confidence interval for the slide false-negative rate was 1-9%, and the 95% confidence interval for the slide sort rate was 45-55%. The estimated sort rate for clearly normal slides is within the range required for a cost-efficient screening system, and the estimated false-negative rate for premalignant and malignant smears is an improvement over published false-negative rates for human performance. Several performance improvement efforts are still under way. We expect that they will result in a vastly reduced slide false-negative rate.

Automation↗

PubMed2003

The effect of sample size and MLP architecture on Bayesian learning for cancer prognosis--a case study.

In this paper we investigate the independent effects of training sample size and multilayer perceptron (MLP) architecture on Bayesian learning to build prognostic models for metastatic breast cancer. We trained two types of Bayesian neural networks on a data set of 1477 metastatic breast cancer patients followed at the Institut Curie using disjoint training sets of sizes k = 50, 100, 200, 300, and 450. The learning performance as measured by an expected loss appeared independent of the two architectures modelling the log hazard function under either proportional or non proportional hazard assumptions, thus indicating that no other sources of nonlinearity besides interactions are present. We found a performance breakdown at k = 50, and no sample size effect for k > or = 100.

Bayes Theorem↗

PubMed2005

Prediction methods for babies' birth weight using linear and nonlinear regression analysis.

The aim of this study is to determine more accurate prediction methods between linear and non-linear methods for prediction of babies' birth weight among maternal demographic characteristics. Three hundred pregnant women were included in the study. Blood glucose level before and after ingestion of glucose load, age, body mass index, % of change in weight during pregnancy, height, gestational age, parity, and fetal sex were collected as independent variables and baby birth weight as dependent variable. In linear regression, least squares estimation method was used to estimate parameters. Non-linear regression method was performed using neural network model with multilayer perceptrons, back propagation method was preferred as learning algorithm. Coefficient of determination, R2, of the linear regression equation was found 59.8% and the standard error of the estimate was calculated as 325.69 gr. In non-linear regression method R2 value was also found 59.8% and standard error of estimate was calculated as 320.30 gr. According to the results of the present study, one method is not significantly better than the other. When "accuracy in prediction" is aimed, it is better to use the two methods and compare the results, and then decide on the selection of the favourable method.

Algorithms↗

PubMed2006

Applying an artificial neural network to predict osteoporosis in the elderly.

Osteoporosis is an essential index of health and economics in every country. Recognizing asymptomatic elderly population with high risks of osteoporosis remains a difficult challenge. For this purpose, we developed and validated an artificial neural network (ANN) to identify the osteoporotic subjects in the elderly. The study population consisted of 1403 elderly adults (mean age 63.50 +/- 0.24 years ranged from 50 to 91 years old, 157 male and 1246 female) randomly selected into 3 sets, 703 participants in training set, 350 participants in selection set, and the remaining 350 participants in test set. The input variables included demographic characteristics, anthropometric measurements, and clinical data. The outcome variable was dichotomous, either non-osteoporotic (T-score of greater than -2.5) or osteoporotic (T-score of -2.5 or less) groups classified by the measurement from dual energy X-ray absorptiometry. ANN was constructed with data from training and selection sets and validated in test set whose outcome variable was unknown to the network. The performance of ANN was evaluated by discrimination and calibration simultaneously. After training processes, the final best ANN was a multilayer perceptron network which determined seven input variables (gender, age, weight, height, body mass index, postmenopausal status, and coffee consumption) as significant features. The discriminatory power of ANN for test set was excellent (area under receiver operating characteristics curve = 0.82 +/- 0.03). ANN also had statistically good fit represented by statistically insignificant Hosmer-Lemeshow statistic (p = 0.24). These results suggested that our final ANN concurrently had good discriminatory power and good-fit calibration. ANN can be used as a promising tool for the elderly to stratify high risk subjects into osteoporotic group.

Aged↗

PubMed1995

Artificial neural network-based channel selection and loudness mapping.

We present in this paper artificial neural network techniques for implementing loudness mapping and "smart" channel selection for cochlear implant systems. For loudness mapping, a multilayer perceptron (MLP) is trained to perform the mapping for each channel according to threshold and comfort levels. It is shown that good accuracy mapping can be performed by a very simple MLP architecture. For channel selection, we propose a neural network-based method that can make "smart" selection. We describe and report results for the case in which 6 channels are to be selected from 18. The neural network-based selection system is trained on a multispeaker labeled speech database and tested on a database of different speakers and spoken sentences. Compared with methods used by leading cochlear implant systems, our approach produces significantly better results, and it is easy to implement in the speech processor of the cochlear implant system.

Cochlear Implants↗

PubMed1993

Artificial neural networks in diagnosis of thyroid function from in vitro laboratory tests.

We studied the potential benefit of using artificial neural networks (ANNs) for the diagnosis of thyroid function. We examined two types of ANN architecture and assessed their robustness in the face of diagnostic noise. The thyroid function data we used had previously been studied by multivariate statistical methods and a variety of pattern-recognition techniques. The total data set comprised 392 cases that had been classified according to both thyroid function and 19 clinical categories. All cases had a complete set of results of six laboratory tests (total thyroxine, free thyroxine, triiodothyronine, triiodothyronine uptake test, thyrotropin, and thyroxine-binding globulin). This data set was divided into subsets used for training the networks and for testing their performance; the test subsets contained various proportions of cases with diagnostic noise to mimic real-life diagnostic situations. The networks studied were a multilayer perceptron trained by back-propagation, and a learning vector quantization network. The training data subsets were selected according to two strategies: either training data based on cases with extreme values for the laboratory tests with randomly selected nonextreme cases added, or training cases from very pure functional groups. Both network architectures were efficient irrespective of the type of training data. The correct allocation of cases in test data subsets was 96.4-99.7% when extreme values were used for training and 92.7-98.8% when only pure cases were used.

Humans↗