Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Bayesian computational modeling”

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 1,099 records · Page 61Linked to original sources
PubMed1994

Removing the assumption of conditional independence from Bayesian decision models by using artificial neural networks: some practical techniques and a case study.

The article describes how artificial neural networks with special designs can be applied to approximate a subjective Bayesian decision model without the assumption of conditional independence. New techniques are proposed to resolve some of the practical difficulties during the processes of problem structuring, knowledge elicitation, quantitative modeling, and model interpretation. A Bayesian model considering the conditional dependencies to predict a teenager's marijuana use was constructed by experts using these techniques, and compared to another conventional Bayesian model which assumed conditional independence. The new approach without the assumption of conditional independence had predictive power (r = 0.7) in the test of linearity compared to the conventional approach (r = 0.58) on a data set (n = 129). Its receiver operating characteristic curve dominated the alternative approach within the range (true positive fraction > 0.7) that we were interested in. The interpretations of the possible conditional dependencies provided by the artificial neural network after the training process were consistent with the expert's descriptions.

Adolescent↗
PubMed2005

Does a Bayesian model of V1 contrast coding offer a neurophysiological account of human contrast discrimination?

The dipper effect for contrast discrimination provides strong evidence that the underlying neural response is accelerating at low contrasts and saturating at high contrasts. The contrast-response functions of V1 neurons do have this sigmoidal shape, but individual neurons do not generally have a dynamic range wide enough to account for the dipper effect. This paper presents a Bayesian model of neurons in monkey V1, whose contrast-response function is described by a modified Naka-Rushton with multiplicative noise. It is shown that a model of groups of twelve or more neurons gives a reasonable explanation of the psychophysical data of two observers, but there is a large systematic error which is apparently due to the shape of the distribution of the monkey's sensitivity parameter, c50. A further model provides a better fit to the data by sacrificing strict adherence to V1 neuronal parameters and, instead using an arbitrary bimodal c50 distribution, perhaps reflecting differences between M- and P-cells.

Bayes Theorem↗
PubMedDate not supplied

Bayesian learning for cardiac SPECT image interpretation.

In this paper, we describe a system for automating the diagnosis of myocardial perfusion from single-photon emission computerized tomography (SPECT) images of male and female hearts. Initially we had several thousand of SPECT images, other clinical data and physician-interpreter's descriptions of the images. The images were divided into segments based on the Yale system. Each segment was described by the physician as showing one of the following conditions: normal perfusion, reversible perfusion defect, partially reversible perfusion defect, fixed perfusion defect, defect showing reverse redistribution, equivocal defect or artifact. The physician's diagnosis of overall left ventricular (LV) perfusion, based on the above descriptions, categorizes a study as showing one or more of eight possible conditions: normal, ischemia, infarct and ischemia, infarct, reverse redistribution, equivocal, artifact or LV dysfunction. Because of the complexity of the task, we decided to use the knowledge discovery approach, consisting of these steps: problem understanding, data understanding, data preparation, data mining, evaluating the discovered knowledge and its implementation. After going through the data preparation step, in which we constructed normal gender-specific models of the LV and image registration, we ended up with 728 patients for whom we had both SPECT images and corresponding diagnoses. Another major contribution of the paper is the data mining step, in which we used several new Bayesian learning classification methods. The approach we have taken, namely the six-step knowledge discovery process has proven to be very successful in this complex data mining task and as such the process can be extended to other medical data mining projects.

Bayes Theorem↗
PubMed2006

Component optimization for image understanding: a Bayesian approach.

In this paper, the optimizations of three fundamental components of image understanding: segmentation/annotation, 3D sensing (stereo) and 3D fitting, are posed and integrated within a Bayesian framework. This approach benefits from recent advances in statistical learning which have resulted in greatly improved flexibility and robustness. The first two components produce annotation (region labeling) and depth maps for the input images, while the third module integrates and resolves the inconsistencies between region labels and depth maps to fit most likely 3D models. To illustrate the application of these ideas, we have focused on the difficult problem of fitting individual tree models to tree stands which is a major challenge for vision-based forestry inventory systems.

Algorithms↗
PubMed2004

Nonhomogeneous model of sequence evolution indicates independent origins of primary endosymbionts within the enterobacteriales (gamma-Proteobacteria).

Standard methods of phylogenetic reconstruction are based on models that assume homogeneity of nucleotide composition among taxa. However, this assumption is often violated in biological data sets. In this study, we examine possible effects of nucleotide heterogeneity among lineages on the phylogenetic reconstruction of a bacterial group that spans a wide range of genomic nucleotide contents: obligately endosymbiotic bacteria and free-living or commensal species in the gamma-Proteobacteria. We focus on AT-rich primary endosymbionts to better understand the origins of obligately intracellular lifestyles. Previous phylogenetic analyses of this bacterial group point to the importance of accounting for base compositional variation in estimating relationships, particularly between endosymbiotic and free-living taxa. Here, we develop an approach to compare susceptibility of various phylogenetic reconstruction methods to the effects of nucleotide heterogeneity. First, we identify candidate trees of gamma-Proteobacteria groEL and 16S rRNA using approaches that assume homogeneous and stationary base composition, including Bayesian, maximum likelihood, parsimony, and distance methods. We then create permutations of the resulting candidate trees by varying the placement of the AT-rich endosymbiont Buchnera. These permutations are evaluated under the nonhomogeneous and nonstationary maximum likelihood model of Galtier and Gouy, which allows equilibrium base content to vary among examined lineages. Our results show that commonly used phylogenetic methods produce incongruent trees of the Enterobacteriales, and that the placement of Buchnera is especially unstable. However, under a nonhomogeneous model, various groEL and 16S rRNA phylogenies that separate Buchnera from other AT-rich endosymbionts (Blochmannia and Wigglesworthia) have consistently and significantly higher likelihood scores. Blochmannia and Wigglesworthia appear to have evolved from secondary endosymbionts, and represent an origin of primary endosymbiosis that is independent from Buchnera. This application of a nonhomogeneous model offers a computationally feasible way to test specific phylogenetic hypotheses for taxa with heterogeneous and nonstationary base composition.

Animals↗
PubMed2005

Control of neuromuscular blockade in the presence of sensor faults.

The problem of embedding sensor fault tolerance in feedback control of neuromuscular blockade is considered. For tackling interruptions of feedback measurements, a structure based upon Bayesian inference as well as a predictive filter is proposed. This algorithm is general and can be applied to different situations. Here, it is incorporated in an adaptive automatic system for feedback control of neuromuscular blockade using continuous infusion of muscle relaxants. A significant contribution consists in the experimental clinical testing of the algorithm in patients undergoing surgery.

Algorithms↗
PubMed2001

Face verification through tracking facial features.

We propose an algorithm for face verification through tracking facial features by using sequential importance sampling. Specifically, we first formulate tracking as a Bayesian inference problem and propose to use Markov chain Monte Carlo techniques for obtaining an empirical solution. A reparameterization is introduced under parametric motion assumption, which facilitates the empirical estimation and also allows verification to be addressed along with tracking. The facial features to be tracked are defined on a grid with Gabor attributes (jets). The motion of facial feature points is modeled as a global two-dimensional (2-D) affine transformation (accounting for head motion) plus a local deformation (accounting for residual motion that is due to inaccuracies in 2-D affine modeling and other factors such as facial expression). Motion of both types is processed simultaneously by the tracker: The global motion is estimated by importance sampling, and the residual motion is handled by incorporating local deformation into the measurement likelihood in computing the weight of a sample. Experiments with a real database of face image sequences are presented.

Journal Article↗
PubMed2003

Comparison of neural network, Bayesian, and multiple stepwise regression-based limited sampling models to estimate area under the curve.

This study compared limited sampling methods (LSM) of estimating area under the plasma concentration versus time curve (AUC) based on a Bayesian regularized neural network, the Bayesian approach, and multiple forward stepwise regression models from selected concentration-time points. Plasma concentration versus time data sets with a linear two-compartmental pharmacokinetic model were simulated. A limited sampling method based on the forward stepwise regression model was developed and validated. Plasma concentration-time points selected by the stepwise regression model were used for neural network and Bayesian evaluation. In addition, 55 plasma concentration-time profiles from two clinical studies were used to develop and compare the predicted AUC(last) for the three approaches. From simulated data sets, mean prediction errors for AUC(last) estimation were 0.00, -5.32, and -6.06 for the neural network, Bayesian approach, and forward stepwise regression LSM, respectively. Mean square errors were 581, 588, and 618, respectively. For clinical data set, model mean prediction errors were 0.00, 3.51, and 3.87, respectively. Model mean square errors were 30.6, 109, and 76, respectively. For both simulated and clinical data sets, the neural network approach to estimate AUC(last) from selected time points was numerically more precise and significantly less biased than the other two methods.

Antiviral Agents↗
PubMed2001

Genetic evaluation of dairy cattle using test-day models.

Recently there has been considerable interest in modeling individual test-day records (TDR) for genetic evaluation of dairy cattle as a replacement for the traditional use of estimated accumulated 305-d yields. Some advantages of test-day models (TDM) include the ability to account for environmental effects of each test day, the ability to model the trajectory of the lactation for individual genotypes or groups of animals, and the possibility of genetic evaluations for persistency of production. Also, the use of test-day models avoids the necessity of extending short lactations on culled animals and animals with records in progress. The disadvantages of TDM include computational difficulties associated with analyzing much larger datasets and the need to estimate many more parameters than in a traditional 305-d lactation model. Several different models have been proposed to model the trajectory of the lactation, including so-called "biological functions," various polynomials and character process models. At present, there is not universal agreement on which models to use in routine prediction of breeding values and better methods to compare models are desirable. Obtaining accurate estimates of the dispersion parameters to use in TDM remains a challenge. Methods used include a two-step procedure in which the dispersion parameters are estimated in a series of multivariate models followed by a reduction in order of fit using covariance functions, and a one-step procedure in which the parameters of TDM are estimated using restricted maximum likelihood or Bayesian methods in a random regression model. Further research should focus on including multiple lactation data and accounting for heterogeneity variance.

Algorithms↗
PubMed2002

A Bayesian approach to discriminate between alternative DNA sequence segmentations.

MOTIVATION: As a result of recombination or rate variation, a DNA sequence alignment may have a mosaic structure, where different segments correspond to different evolutionary histories. While several methods have been developed to predict DNA mosaic structures, they do not properly address the question of whether the predicted segmentation itself is statistically significant, or whether it is significantly better than an alternative mosaic structure predicted with another method. The objective of the present article is to devise an approximate Bayesian hypothesis test to discriminate between alternative candidate mosaic structures. RESULTS: We have applied the proposed discrimination scheme to various synthetic and real-world DNA sequence alignments. On the synthetic data, the algorithm identified the true mosaic structure in nine out of ten cases. On the real-world sequence alignments, it selected the same mosaic structures as predicted in the literature.

Algorithms↗
PubMed1999

Statistical methods for mapping quantitative trait loci from a dense set of markers.

Lander and Botstein introduced statistical methods for searching an entire genome for quantitative trait loci (QTL) in experimental organisms, with emphasis on a backcross design and QTL having only additive effects. We extend their results to intercross and other designs, and we compare the power of the resulting test as a function of the magnitude of the additive and dominance effects, the sample size and intermarker distances. We also compare three methods for constructing confidence regions for a QTL: likelihood regions, Bayesian credible sets, and support regions. We show that with an appropriate evaluation of the coverage probability a support region is approximately a confidence region, and we provide a theroretical explanation of the empirical observation that the size of the support region is proportional to the sample size, not the square root of the sample size, as one might expect from standard statistical theory.

Chromosome Mapping↗
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

Identification of transcription factor binding sites with variable-order Bayesian networks.

MOTIVATION: We propose a new class of variable-order Bayesian network (VOBN) models for the identification of transcription factor binding sites (TFBSs). The proposed models generalize the widely used position weight matrix (PWM) models, Markov models and Bayesian network models. In contrast to these models, where for each position a fixed subset of the remaining positions is used to model dependencies, in VOBN models, these subsets may vary based on the specific nucleotides observed, which are called the context. This flexibility turns out to be of advantage for the classification and analysis of TFBSs, as statistical dependencies between nucleotides in different TFBS positions (not necessarily adjacent) may be taken into account efficiently--in a position-specific and context-specific manner. RESULTS: We apply the VOBN model to a set of 238 experimentally verified sigma-70 binding sites in Escherichia coli. We find that the VOBN model can distinguish these 238 sites from a set of 472 intergenic 'non-promoter' sequences with a higher accuracy than fixed-order Markov models or Bayesian trees. We use a replicated stratified-holdout experiment having a fixed true-negative rate of 99.9%. We find that for a foreground inhomogeneous VOBN model of order 1 and a background homogeneous variable-order Markov (VOM) model of order 5, the obtained mean true-positive (TP) rate is 47.56%. In comparison, the best TP rate for the conventional models is 44.39%, obtained from a foreground PWM model and a background 2nd-order Markov model. As the standard deviation of the estimated TP rate is approximately 0.01%, this improvement is highly significant.

Algorithms↗
PubMedDate not supplied

Evaluation of a computerized Bayesian model for diagnosis of renal cyst vs. tumor vs. normal variant from urogram information.

The diagnostic problem of cyst/tumor/normal variant raised on an excretory urogram leads to a decision to do needle aspiration or renal arteriography. This decision depends critically upon the probability distribution for the three diagnoses. A computerized Bayesian model of a uroradiologist's diagnostic process in solving the problem was developed. The model was based on subjective probabilities supplied by an experienced uroradiologist. The model was evaluated in terms of its ability to decrease the cost of further diagnosis regarding aspiration versus arteriography. The model's output was compared with decisions made by unaided radiologists viewing the same panel of 50 urogram test cases. Results indicate that the model does not improve upon the decisions made by a radiologist highly experienced with this diagnostic problem. However, the decisions made by unaided, less experienced radiologists result in greater cost than those of the model.

Angiography↗
PubMed2001

Population pharmacokinetic analysis resulting in a tool for dose individualization of busulphan in bone marrow transplantation recipients.

The aims of the present study were (1) to investigate and quantify the pharmacokinetics, including inter-occasion variability and covariate relationships, of busulphan in BMT patients and (2) to develop a user-friendly initial dosing and therapeutic drug monitoring (TDM) strategy for the treatment of those patients with busulphan. The pharmacokinetics of busulphan was studied in 64 adults and 12 children who received busulphan (1 mg/kg) four times daily for 4 days. A one-compartment model with first order absorption and a lag time was sufficient in describing the concentration-time profile. Oral clearance (CL/F) was found to be correlated to weight (+1.2%/kg), ALT (-13%/microcat/l) and concomitant phenytoin treatment (+21%). CL/F and the volume of distribution (V/F) were estimated to 9.23 l/h and 39.3 l, respectively, in a typical individual. Inter-occasion variability (9.4%) in CL/F was estimated to be less than inter-individual variability (28%), a prerequisite for the value of TDM. Bayesian CL/F estimates based on three samples were in good accordance with those based on all samples. The final population model was implemented into the program Excel. The resulting flexible and easy to use dosing program might be used for both initial and, requiring only three plasma samples, maintenance dose individualization of busulphan therapy.

Administration, Oral↗
PubMed2006

Fuzzy Markov random fields versus chains for multispectral image segmentation.

This paper deals with a comparison of recent statistical models based on fuzzy Markov random fields and chains for multispectral image segmentation. The fuzzy scheme takes into account discrete and continuous classes which model the imprecision of the hidden data. In this framework, we assume the dependence between bands and we express the general model for the covariance matrix. A fuzzy Markov chain model is developed in an unsupervised way. This method is compared with the fuzzy Markovian field model previously proposed by one of the authors. The segmentation task is processed with Bayesian tools, such as the well-known MPM (Mode of Posterior Marginals) criterion. Our goal is to compare the robustness and rapidity for both methods (fuzzy Markov fields versus fuzzy Markov chains). Indeed, such fuzzy-based procedures seem to be a good answer, e.g., for astronomical observations when the patterns present diffuse structures. Moreover, these approaches allow us to process missing data in one or several spectral bands which correspond to specific situations in astronomy. To validate both models, we perform and compare the segmentation on synthetic images and raw multispectral astronomical data.

Algorithms↗
PubMed1997

Selecting optimal experiments for multiple output multilayer perceptrons.

Where should a researcher conduct experiments to provide training data for a multilayer perceptron? This question is investigated, and a statistical method for selecting optimal experimental design points for multiple output multilayer perceptrons is introduced. Multiple class discrimination problems are examined using a framework in which the multilayer perceptron is viewed as a multivariate nonlinear regression model. Following a Bayesian formulation for the case where the variance-covariance matrix of the responses is unknown, a selection criterion is developed. This criterion is based on the volume of the joint confidence ellipsoid for the weights in a multilayer perceptron. An example is used to demonstrate the superiority of optimally selected design points over randomly chosen points, as well as points chosen in a grid pattern. Simplification of the basic criterion is offered through the use of Hadamard matrices to produce uncorrelated outputs.

Algorithms↗
PubMed2005

Probabilistic prediction of protein-protein interactions from the protein sequences.

Prediction of protein-protein interactions is very important for several bioinformatics tasks though it is not a straightforward problem. In this paper, employing only protein sequence information, a framework is presented to predict protein-protein interactions using a probabilistic-based tree augmented nai ve (TAN) Bayesian network. Our framework also provides a confidence level for every predicted interaction, which is useful for further analysis by the biologists. The framework is applied to the yeast interaction datasets for predicting interactions and it is shown that our framework gives better performance than support vector machine (SVM). The framework is implemented as a webserver and is available for prediction.

Artificial Intelligence↗