Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Parallel Algorithms”

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 649 records · Page 36Linked to original sources
PubMed1999

[Calculating the localization and dimension of the real pupil in keratoconus with ray tracing of corneal topography data].

BACKGROUND: It is crucial to center surgical procedures for optical indications on the pupil or the optical axis of the eye. In keratoconus the pupil appears to be dislocated due to optical aberrations of corneal topography. The purpose of this study was to evaluate the real pupil structure from the virtual image using exact raytracing techniques. PATIENTS AND METHODS: Eighty-eight patients with keratoconus (46 with mild and 42 with severe clinical signs) and a control group of 40 normal subjects were included in this study. Topographic height data were calculated from refraction data of a commercially available topographer (TMS-1) using a local approximation algorithm and a convex surface was modelled using a subdivision scheme. For the posterior corneal surface we postulated an aspherical surface with a central radius of curvature of 6.5 mm using Navarro's model eye. At the virtual pupil outline a bundle of parallel rays were intersected with the anterior and posterior corneal surface and refracted into the anterior chamber. The intersections of these rays with the pupil plane was defined as the real pupil outline. We assessed the amount and direction of pupil dislocation, the ratio between the virtual and real pupil size for each group and correlated these parameters with the central corneal power. RESULTS: The size of the virtual pupil exceeded the reference value of the real pupil in the normal group by 11%, in the group with mild keratoconus by 19% and in the group with severe keratoconus by 35%. The center of the virtual pupil was decentered 0.06 mm in the normal group, 0.49 in the group with mild keratoconus and 1.24 mm in the group with severe keratoconus. Whereas the direction of decentration was randomly in the normal group, we measured a preferred decentration to the inferior quadrants in mild keratoconus and a systematic decentration to the temporal inferior quadrant in severe keratoconus. Correlation of the optical dislocation did not correlate with central corneal power in any group. CONCLUSIONS: In keratoconic eyes the pupil outline is distorted and dislocated due to optical aberrations of the cornea. Exact raytracing technique allows the calculation of the real pupil outline from the virtual image and the topographic height of both corneal surfaces. Knowledge about the real pupil position may have an impact on adequate centration of keratorefractive surgery and penetrating keratoplasty.

Adult↗
PubMed2000

Interior-point methodology for 3-D PET reconstruction.

Interior-point methods have been successfully applied to a wide variety of linear and nonlinear programming applications. This paper presents a class of algorithms, based on path-following interior-point methodology, for performing regularized maximum-likelihood (ML) reconstructions on three-dimensional (3-D) emission tomography data. The algorithms solve a sequence of subproblems that converge to the regularized maximum likelihood solution from the interior of the feasible region (the nonnegative orthant). We propose two methods, a primal method which updates only the primal image variables and a primal-dual method which simultaneously updates the primal variables and the Lagrange multipliers. A parallel implementation permits the interior-point methods to scale to very large reconstruction problems. Termination is based on well-defined convergence measures, namely, the Karush-Kuhn-Tucker first-order necessary conditions for optimality. We demonstrate the rapid convergence of the path-following interior-point methods using both data from a small animal scanner and Monte Carlo simulated data. The proposed methods can readily be applied to solve the regularized, weighted least squares reconstruction problem.

Algorithms↗
PubMed2002

Investigating cervical spinal cord structure using axial diffusion tensor imaging.

This study describes a new technique for Diffusion Tensor Imaging (DTI) that acquires axial (transverse) images of the cervical spinal cord. The DTI images depict axonal fiber orientation, enable quantification of diffusion characteristics along the spinal cord, and have the potential to demonstrate the connectivity of cord white matter tracts. Because of the high sensitivity to motion of diffusion-weighted magnetic resonance imaging and the small size of the spinal cord, a fast imaging method with high in-plane resolution was developed. Images were acquired with a single-shot EPI technique, named ZOOM-EPI (zonally magnified oblique multislice echo planar imaging), which selects localized areas and reduces artefacts caused by susceptibility changes between soft tissue and the adjacent vertebrae. Cardiac gating was used to reduce pulsatile flow artefacts from the surrounding cerebrospinal fluid. Voxel resolution was 1.25 x 1.25 mm(2) in-plane with 5-mm slice thickness. Both the mean diffusivity (MD) and the fractional anisotropy (FA) indices of the cervical spinal cord were measured. The FA index demonstrated high anisotropy of the spinal cord with an average value of 0.61 +/- 0.05 (highest value of 0.66 +/- 0.03 at C3), comparable to white matter tracts in the brain. The diffusivity components parallel and orthogonal to the longitudinal axes of the cord were lambda( parallel) = (1648 +/- 123) x 10(-6) mm(2)s(-1) and lambda( perpendicular) = (570 +/- 47) x 10(-6) mm(2) s(-1), respectively. The high axial resolution allowed preliminary evaluation of fiber connectivity using the fast-marching tractography algorithm, which generated traces of fiber paths consistent with the well-known cord anatomy.

Algorithms↗
PubMed1983

Using SEEK for multichannel pattern recognition.

The author's work on computerized analysis of the 2-channel, 24-hr electrocardiogram has previously resulted in the development of multichannel signal processing systems that learn by observation. A new tool for implementing such algorithms is described: the pattern recognition language SEEK. Programs written in SEEK build a knowledge base containing treelike data structures, each of which stores acquired information about a particular multichannel waveform. Input data are interpreted by performing an efficient parallel evaluation of the structures in the knowledge base. The work is applicable to a wide variety of pattern recognition problems that arise in medical signal processing. The approach is illustrated with examples drawn from ECG analysis.

Computers↗
PubMed2002

Protein therapeutics: promises and challenges for the 21st century.

Recent advances in massively parallel experimental and computational technologies are leading to radically new approaches to the early phases of the drug production pipeline. The revolution in DNA microarray technologies and the imminent emergence of its analogue for proteins, along with machine learning algorithms, promise rapid acceleration in the identification of potential drug targets, and in high-throughput screens for subpopulation-specific toxicity. Similarly, advances in structural genomics in conjunction with in vitro and in silico evolutionary methods will rapidly accelerate the number of lead drug candidates and substantially augment their target specificity. Taken collectively, these advances will usher in an era of predictive medicine, which will move medical practice from reactive therapy after disease onset, to proactive prevention.

Animals↗
PubMed1981

Time-constant histograms from the forced expired volume signal.

The forced expired volume signal was analyzed using a parallel compartment model in which each compartment emptied exponentially. With this model the forced expired volume signal was represented by a histogram showing the fraction of the vital capacity as a functional of compartmental time constants. We developed an algorithm to compute this histogram from the volume signal. The algorithm used the least-squares criterion function with both smoothness and nonnegativity constraints. In a stimulation study reasonable histograms were obtained even in the presence of realistic random error. Three dependent forced expired volume signals from 16 subjects were analyzed, and the histograms were reproducible. Most histograms were bimodal with fast time constants of 0.12-0.55 s and slow time constants of 1.3-2.7 s. In all normal subjects and patients with restrictive disease more than 75% of the vital capacity was in the fast time-constant mode. Subjects with obstructive disease had more than 40% of the vital capacity in the slow time-constant mode.

Adult↗
PubMed1999

Discontinuities in recurrent neural networks.

This article studies the computational power of various discontinuous real computational models that are based on the classical analog recurrent neural network (ARNN). This ARNN consists of finite number of neurons; each neuron computes a polynomial net function and a sigmoid-like continuous activation function. We introduce arithmetic networks as ARNN augmented with a few simple discontinuous (e.g., threshold or zero test) neurons. We argue that even with weights restricted to polynomial time computable reals, arithmetic networks are able to compute arbitrarily complex recursive functions. We identify many types of neural networks that are at least as powerful as arithmetic nets, some of which are not in fact discontinuous, but they boost other arithmetic operations in the net function (e.g., neurons that can use divisions and polynomial net functions inside sigmoid-like continuous activation functions). These arithmetic networks are equivalent to the Blum-Shub-Smale model, when the latter is restricted to a bounded number of registers. With respect to implementation on digital computers, we show that arithmetic networks with rational weights can be simulated with exponential precision, but even with polynomial-time computable real weights, arithmetic networks are not subject to any fixed precision bounds. This is in contrast with the ARNN that are known to demand precision that is linear in the computation time. When nontrivial periodic functions (e.g., fractional part, sine, tangent) are added to arithmetic networks, the resulting networks are computationally equivalent to a massively parallel machine. Thus, these highly discontinuous networks can solve the presumably intractable class of PSPACE-complete problems in polynomial time.

Algorithms↗
PubMed1994

Genetic map construction with constraints.

A pilot program, CME, is described for generating a physical genetic map from hybridization fingerprinting data. CME is implemented in the parallel constraint logic programming language ElipSys. The features of constraint logic programming are used to enable the integration of pre-existing mapping information (partial probe orders from cytogenetic maps and local physical maps) into the global map generation process, while parallelism enables the search space to be traversed more efficiently. CME was tested using data from chromosome 2 of Schizosaccharomyces pombe and was found able to generate maps as well as (and sometimes better than) a more traditional method. This paper illustrates the practical benefits of using a symbolic logic programming language and shows that the features of constraint handling and parallel execution bring the development of practical systems based on AI programming technologies nearer to being a reality.

Algorithms↗
PubMed1987

Accuracy in clinical electron beam dose planning using pencil beam algorithms.

The accuracy of two different pencil beam models for electron beam dose planning is shown by comparisons with measured dose distributions. The investigation is restricted to two-dimensional geometries. In one model the multiple scattering of the electrons is considered by the small angle Gaussian approximation of Fermi and Eyges. In the other, the generalized Gaussian model, the large angle single scattering events are also considered. The parallel slab approximation is used in both models. The comparisons have been made for two different anatomical phantoms. One phantom was constructed to simulate a transversal cross-section through the chest wall and lung in radiotherapy of breast cancer. The other phantom was made to simulate the head at the level of the nose. The measurements in these phantoms were made with thermoluminescence dosimetry, LiF rods. The accuracy of the oblique incidence case was investigated in a homogenous water phantom. The results show that for oblique incidence and geometries where the semi-infinite slab approximation is reasonably good, the generalized Gaussian model is more accurate. Within and behind low density cavities, in the phantom, the Gaussian model often gives a better agreement to experimental data. This is shown to be due to mutually balancing errors from the semi-infinite slab approximation and the Gaussian approximation.

Algorithms↗
PubMed2002

Probabilistic sampling of protein conformations: new hope for brute force?

Protein structure prediction from sequence alone by "brute force" random methods is a computationally expensive problem. Estimates have suggested that it could take all the computers in the world longer than the age of the universe to compute the structure of a single 200-residue protein. Here we investigate the use of a faster version of our FOLDTRAJ probabilistic all-atom protein-structure-sampling algorithm. We have improved the method so that it is now over twenty times faster than originally reported, and capable of rapidly sampling conformational space without lattices. It uses geometrical constraints and a Leonard-Jones type potential for self-avoidance. We have also implemented a novel method to add secondary structure-prediction information to make protein-like amounts of secondary structure in sampled structures. In a set of 100,000 probabilistic conformers of 1VII, 1ENH, and 1PMC generated, the structures with smallest Calpha RMSD from native are 3.95, 5.12, and 5.95A, respectively. Expanding this test to a set of 17 distinct protein folds, we find that all-helical structures are "hit" by brute force more frequently than beta or mixed structures. For small helical proteins or very small non-helical ones, this approach should have a "hit" close enough to detect with a good scoring function in a pool of several million conformers. By fitting the distribution of RMSDs from the native state of each of the 17 sets of conformers to the extreme value distribution, we are able to estimate the size of conformational space for each. With a 0.5A RMSD cutoff, the number of conformers is roughly 2N where N is the number of residues in the protein. This is smaller than previous estimates, indicating an average of only two possible conformations per residue when sterics are accounted for. Our method reduces the effective number of conformations available at each residue by probabilistic bias, without requiring any particular discretization of residue conformational space, and is the fastest method of its kind. With computer speeds doubling every 18 months and parallel and distributed computing becoming more practical, the brute force approach to protein structure prediction may yet have some hope in the near future.

Algorithms↗
PubMed2000

Characterization of airborne transducers by optical tomography

This paper describes the application of an acousto-optic method to the measurement of airborne ultrasound. The method consists of a heterodyne interferometric probing of the pressure emitted by the transducer combined with a tomographic algorithm. The heterodyne interferometer measures the optical phase shift of the probe laser beam, proportional to the acoustic pressure integrated along the light path. A number of projections of the sound field, e.g. a set of ray integrals obtained along parallel paths, are made in moving the transducer to be tested. The main advantage of the method is its very high sensitivity in air (2 x 10(-4) Pa Hz-1/2), combined with a large bandwidth. Using the same principle as X-ray tomography the ultrasonic pressure in a plane perpendicular to the transducer axis can be reconstructed. Several ultrasonic fields emitted by wide-band home made electrostatic transducers, with operating frequencies between 200 and 700 kHz, have been measured. The sensitivities compared favorably with those of commercial airborne transducers.

Journal Article↗
PubMed1999

Implementation of spectral width Doppler in pulsatile flow measurements.

In this paper, we present an automatic beam-vector (Doppler) angle and flow velocity measurement method and implement it in pulsatile flow measurements using a clinical Doppler ultrasound system. In current clinical Doppler ultrasound flow velocity measurements, the axis of the blood vessel needs to be set manually on the B-scan image to enable the estimation of the beam-vector angle and the beam-vector angle corrected flow velocity (the actual flow velocity). In this study, an annular array transducer was used to generate a conical-shaped and symmetrically focused ultrasound beam to measure the flow velocity vectors parallel and perpendicular to the ultrasound beam axis. The beam-vector angle and flow velocity is calculated from the mode frequency (f(d)) and the maximum Doppler frequency (f(max)) of the Doppler spectrum. We develop a spectrum normalization algorithm to enable the Doppler spectrum averaging using the spectra obtained within a single cardiac cycle. The Doppler spectrum averaging process reduces the noise level in the Doppler spectrum and also enables the calculation of the beam-vector angle and flow velocity for pulsatile flows to be measured. We have verified the measurement method in vivo over a wide range of angles, from 52 degrees to 80 degrees, and the standard deviations of the measured beam-vector angles and flow velocities in the carotid artery are lower than 2.2 degrees and 12 cm/s (about 13.3%), respectively.

Blood Flow Velocity↗
PubMed2002

Deep insight from simple models of evolution.

On one hand, people admire the often strikingly efficient results of organic evolution. On the other hand, however, they decry mutation and selection to be a rather prodigal, inefficient trial-and-error strategy. Taking into account the parallel information processing in a heterogeneous population and sexual propagation with recombination, as well as the genetic control of the reproduction accuracy, computer simulated evolution reveals a couple of interesting, sometimes surprising, properties of nature's learning-by-doing algorithm. Survival of the fittest, Spencer's résumé of Darwin's view, turns out to be a bad advice if taken literally. Individual death, forgetting, and even regression show up to be necessary ingredients of the life game. Whether the process should be named gradualistic or punctualistic, is a matter of the observer's point of view.

Algorithms↗
PubMed2002

Functional MRI experiments: acquisition, analysis and interpretation of data.

Functional MRI is widely used to address basic and clinical neuroscience questions. In the key domains of fMRI experiments, i.e. acquisition, processing and analysis, and interpretation of data, developments are ongoing. The main issues are sensitivity for changes in fMRI signal that are associated with brain function, and the design of tasks with which brain functions are invoked. In this paper we address these issues, in terms of strengths, weaknesses and future developments. Acquisition of data is commonly achieved with techniques that measure blood oxygen level-dependent (BOLD) signal changes. Although the mechanisms that affect BOLD signal are complex and not well understood, fMRI yields results that agree with known functional topography. Sensitivity for task-related brain activity is expected to benefit from technological advances in acquisition, i.e. SENSE or parallel imaging, and higher field scanners (3 T). Data analysis is geared towards modelling sources of signal variation, i.e. reducing noise in the data time-series, and the cerebrovascular response to task-related changes in neuronal activity. Analytical algorithms such as connectivity and component analysis contribute to the extraction of meaningful information from fMRI datasets. The choice of tasks, and consequently of the statistical evaluation procedures, is best guided by the specific questions that are formulated a priori. The future is expected to bring more sophisticated questions, and tasks that allow for accurate modelling of involved brain functions. An example of a hypothesis-driven experiment is presented, where we investigated whether practise of a working memory task caused a shift in the neuronal representation of working memory or not.

Brain↗
PubMed1979

Rapid analysis of hematology image data: the ADC-500 preprocessor.

A sequential, pipeline processor (that we have named the ADC-500 preprocessor) has been developed which scene segments the three color image data from the ADC-500 optics one image element at a time, groups together image elements from each object in the scene and extracts features from each object. The processing occurs at television frame rates, requiring 16.7 msec to process the entire image. This speed was instrumental in allowing the ADC-500 automated differential analyzer to perform routine 500-cell differentials. The preprocessor also contains hardware which simplifies compilation of the three color histograms. The segmentation algorithms implemented in the preprocessor are multicolor extensions of the classical monochrome density histogram threshold method. For most cell image analysis tasks, a sequential pipeline processor of this type should be more economical and as fast or faster than a parallel processor.

Blood Cells↗
PubMed1993

Hidden Markov models and iterative aligners: study of their equivalence and possibilities.

There are many shared attributes between existing iterative aligners and Hidden Markov Model (HMM). A learning algorithm of HMM called Viterbi is the same as the iteration of DP-matching of iterative aligners. HMM aligners can use the result of an iterative aligner initially, incorporate the similarity score of amino acids, and apply the detailed gap cost systems to improve the matching accuracy. On the other hand, the iterative aligner can inherit the modeling capability of HMM, and provide the better representation of the proteins than motifs. In this paper, we present an overview of several iterative aligners which include the parallel iterative aligner of ICOT and the HMM aligner of Haussler's group. We compare the merits and shortcomings of these aligners. This comparison enables us to formulate a better, more advanced aligner through proper integration of the iterative technique and HMM technique.

Algorithms↗
PubMed1998

Analysis of 3D transient blood flow passing through an artificial aortic valve by Lattice-Boltzmann methods.

The development of flow instabilities due to high Reynolds number flow in artificial heart valve geometries inducing high strain rates and stresses often leads to hemolysis and related highly undesired effects. Geometric and functional optimization of artificial heart valves is therefore mandatory. In addition to experimental work in this field it is meanwhile possible to obtain increasing insight into flow dynamics by computer simulation of refined model problems. After giving an introductory overview we report the results of the simulation of three-dimensional transient physiological flows in fixed geometries similar to a CarboMedics bileaflet heart valve at different opening angles. The visualization of emerging complicated flow patterns gives detailed information about the transient history of the systems dynamical stability. Stress analysis indicates temporal shear stress peaks even far away from walls. The mathematical approach used is the Lattice Boltzmann method. We obtained reasonable results for velocity and shear stress fields. The code is implemented on parallel hardware in order to decrease computation time. Finally, we discuss problems, shortcomings and possible extensions of our approach.

Algorithms↗
PubMed2002

A note on sample size calculation for mean comparisons based on noncentral t-statistics.

One-sample and two-sample t-tests are commonly used in analyzing data from clinical trials in comparing mean responses from two drug products. During the planning stage of a clinical study, a crucial step is the sample size calculation, i.e., the determination of the number of subjects (patients) needed to achieve a desired power (e.g., 80%) for detecting a clinically meaningful difference in the mean drug responses. Based on noncentral t-distributions, we derive some sample size calculation formulas for testing equality, testing therapeutic noninferiority/superiority, and testing therapeutic equivalence, under the popular one-sample design, two-sample parallel design, and two-sample crossover design. Useful tables are constructed and some examples are given for illustration.

Algorithms↗