Showing papers by "Lucian Mihai Itu published in 2014"

Optimized three-dimensional stencil computation on Fermi and Kepler GPUs

Abstract: Stencil based algorithms are used intensively in scientific computations. Graphics Processing Units (GPU) based implementations of stencil computations speed-up the execution significantly compared to conventional CPU only systems. In this paper we focus on double precision stencil computations, which are required for meeting the high accuracy requirements, inherent for scientific computations. Starting from two baseline implementations (using two dimensional and three dimensional thread block structures respectively), we employ different optimization techniques which lead to seven kernel versions. Both Fermi and Kepler GPUs are used, to evaluate the impact of different optimization techniques for the two architectures. Overall, the GTX680 GPU card performs best for a kernel with 2D thread block structure and optimized register and shared memory usage. We show that, whereas shared memory is not essential for Fermi GPUs, it is a highly efficient optimization technique for Kepler GPUs (mainly due to the different L1 cache usage). Furthermore, we evaluate the performance of Kepler GPU cards designed for desktop PCs and notebook PCs. The results indicate that the ratio of execution time is roughly equal to the inverse of the ratio of power consumption.

Model based non-invasive estimation of PV loop from echocardiography.

Abstract: We introduce a model-based approach for the non-invasive estimation of patient specific, left ventricular PV loops. A lumped parameter circulation model is used, composed of the pulmonary venous circulation, left atrium, left ventricle and the systemic circulation. A fully automated parameter estimation framework is introduced for model personalization, composed of two sequential steps: first, a series of parameters are computed directly, and, next, a fully automatic optimization-based calibration method is employed to iteratively estimate the values of the remaining parameters. The proposed methodology is first evaluated for three healthy volunteers: a perfect agreement is obtained between the computed quantities and the clinical measurements. Additionally, for an initial validation of the methodology, we computed the PV loop for a patient with mild aortic valve regurgitation and compared the results against the invasively determined quantities: there is a close agreement between the time-varying LV and aortic pressures, time-varying LV volumes, and PV loops.

Double precision stencil computations on Kepler GPUs

Abstract: Graphics Processing Units (GPU) have been used extensively for accelerating parallelizable applications in general, and scientific computations in particular. Stencil based algorithms are used intensively in various research areas and represent good candidates for GPU based acceleration. Since scientific computations have high accuracy requirements, herein we focus on stencil based double precision computations. For a seven-point stencil we introduce two basic implementations, which use two-dimensional and three-dimensional thread organization respectively. Different optimization techniques lead then to a total of seven different implementations, which are evaluated for two NVIDIA Kepler GPUs. The best performance is obtained for the GTX680 card, for a kernel with two-dimensional thread organization and optimized shared memory and register usage.

Uncertainty quantification in medical image-based hemodynamic computations

Abstract: In this paper, we present a framework for uncertainty quantification in medical image-based patient-specific hemodynamic computations. To illustrate the overall methodology, we have used an aortic coarctation model for computing trans-stenotic pressure gradient. Variance-based Sobol sensitivity indices are used to evaluate the relative influence of the various uncertain measurements and model parameters on the global variance of the output. Next, a generalized Polynomial Chaos Expansion (PCE) method is used to quantify the uncertainties in the computed mean and peak pressure gradient in terms of a probability density functions and error bars over a full cardiac cycle.

Patient-specific automated tuning of boundary conditions for distal vessel tree

Abstract: Boundary conditions for a distal vessel tree are modeled and tuned to a specific patient. Measurements from the patient are used to find reference compliance and resistance for the root of the distal vessel tree model. The reference compliance and resistance are used to tune properties of a structured tree model, such as by iteratively solving for the properties while matching the compliance and resistance of the structured tree model to the patient-specific reference compliance and reference resistance. The tuned structured tree is then used to calculate boundary conditions for computing flow of a scanned vessel of the patient.

A method for modeling surrounding tissue support and its global effects on arterial hemodynamics

Abstract: We propose a methodology for separating the total stiffness of arteries, determined in vivo, into stiffness of the arterial wall and stiffness of the surrounding tissue. An effective perivascular pressure is considered which introduces a radial constraint. Next, based on vivo data, acquired at diastolic pressure, the cross-sectional area at zero pressure is estimated. Finally, the stiffness of the arterial wall and of the surrounding tissue are determined based on a model with two parallel springs. By employing a reduced-order multiscale model, the methodology is used for studying the global effects of surrounding tissue support on arterial hemodynamics. The main effects are: higher wave speed, earlier arriving backward travelling pressure and flow rate waves, lower total compliance, higher pressure pulse, and reduced arterial cross-sectional areas. In a Big Data perspective, by combining this methodology with arterial wall growth models and by comparing simulation results and patient evolution over different time ranges, such an approach is useful for predicting patient-specific disease evolution and outcome.

Transit time estimations from coronary angiograms

Abstract: This X-ray angiography is the current gold standard for morphology based diagnosis of coronary arteries due to their spatial and temporal resolution. Transit time methods are applied on coronary angiograms to estimate the interval of time required for the contrast material to traverse the distance between two regions of interest. They represent an important prerequisite for estimating blood flow parameters like velocity or flow rate which characterize the hemodynamic patterns of blood vessels. In this article we perform an analysis of ten different transit time estimation methods on routine clinical angiographic data acquired at the Clinical Emergency Hospital of Bucharest by using the time density curves. Different coronary arteries are considered and the analysis is performed on raw data, filtered data and fitted data. The most robust methods are the mean transit time, the mean arrival time and the cross correlation method. Furthermore, the mean transit time and the cross correlation method tend to underestimate, while the rise time method tends to overestimate the transit time. Automated filtering and fitting methods improve the results significantly and should always be used in conjunction with transit time estimation methods.

An external tissue support model for the arterial wall based on in vivo data

Abstract: We introduce a method for modeling external tissue support of human arterial hemodynamics. An effective perivascular pressure is considered and the external tissue support model is bas ed on the separati on of total stiffness into arterial wall stiffness and external tissue stiffness. To perform this separation, first the cross-sectional area values at the hypothetical zero pressure state are computed. Finally, a model with two parallel springs is used to determine the material properties of each component. The parameter values are estimated from in vivo data acquired at end diastole. By employing a reduced-order multiscale blood flow model the method is used to study the global effects of external tissue support on the human arterial circulation. The main conclusions are: pressure pulse increases (especially in the proximal aorta), wave speed increases, backward travelling pressure and flow rate waves arrive earlier, the total arterial compliance decreases, cross-sectional area values decrease and oscillations of flow rate and pressure profiles at distal locations are dampened. The computed hemodynamic quantities of interest can be combined with a growth model to predict patient-specific arterial wall remodeling.