Sethuraman Sankaran

A Stochastic Collocation Method for Uncertainty Quantification and Propagation in Cardiovascular Simulations

Simulations of blood flow in both healthy and diseased vascular models can be used to compute a range of hemodynamic parameters including velocities, time varying wall shear stress, pressure drops, and energy losses. The confidence in the data output from cardiovascular simulations depends directly on our level of certainty in simulation input parameters. In this work, we develop a general set of tools to evaluate the sensitivity of output parameters to input uncertainties in cardiovascular simulations. Uncertainties can arise from boundary conditions, geometrical parameters, or clinical data. These uncertainties result in a range of possible outputs which are quantified using probability density functions (PDFs). The objective is to systemically model the input uncertainties and quantify the confidence in the output of hemodynamic simulations. Input uncertainties are quantified and mapped to the stochastic space using the stochastic collocation technique. We develop an adaptive collocation algorithm for Gauss-Lobatto-Chebyshev grid points that significantly reduces computational cost. This analysis is performed on two idealized problems--an abdominal aortic aneurysm and a carotid artery bifurcation, and one patient specific problem--a Fontan procedure for congenital heart defects. In each case, relevant hemodynamic features are extracted and their uncertainty is quantified. Uncertainty quantification of the hemodynamic simulations is done using (a) stochastic space representations, (b) PDFs, and (c) the confidence intervals for a specified level of confidence in each problem.

Patient-Specific Multiscale Modeling of Blood Flow for Coronary Artery Bypass Graft Surgery

We present a computational framework for multiscale modeling and simulation of blood flow in coronary artery bypass graft (CABG) patients. Using this framework, only CT and non-invasive clinical measurements are required without the need to assume pressure and/or flow waveforms in the coronaries and we can capture global circulatory dynamics. We demonstrate this methodology in a case study of a patient with multiple CABGs. A patient-specific model of the blood vessels is constructed from CT image data to include the aorta, aortic branch vessels (brachiocephalic artery and carotids), the coronary arteries and multiple bypass grafts. The rest of the circulatory system is modeled using a lumped parameter network (LPN) 0 dimensional (0D) system comprised of resistances, capacitors (compliance), inductors (inertance), elastance and diodes (valves) that are tuned to match patient-specific clinical data. A finite element solver is used to compute blood flow and pressure in the 3D (3 dimensional) model, and this solver is implicitly coupled to the 0D LPN code at all inlets and outlets. By systematically parameterizing the graft geometry, we evaluate the influence of graft shape on the local hemodynamics, and global circulatory dynamics. Virtual manipulation of graft geometry is automated using Bezier splines and control points along the pathlines. Using this framework, we quantify wall shear stress, wall shear stress gradients and oscillatory shear index for different surgical geometries. We also compare pressures, flow rates and ventricular pressure–volume loops pre- and post-bypass graft surgery. We observe that PV loops do not change significantly after CABG but that both coronary perfusion and local hemodynamic parameters near the anastomosis region change substantially. Implications for future patient-specific optimization of CABG are discussed.

Quantitative assessment of collagen fibre orientations from two-dimensional images of soft biological tissues

In this work, we outline an automated method for the extraction and quantification of material parameters characterizing collagen fibre orientations from two-dimensional images. Morphological collagen data among different length scales were obtained by combining the established methods of Fourier power spectrum analysis, wedge filtering and progressive regions of interest splitting. Our proposed method yields data from which we can determine parameters for computational modelling of soft biological tissues using fibre-reinforced constitutive models and gauge the length scales most appropriate for obtaining a physically meaningful measure of fibre orientations, which is representative of the true tissue morphology of the two-dimensional image. Specifically, we focus on three parameters quantifying different aspects of the collagen morphology: first, using maximum-likelihood estimation, we extract location parameters that accurately determine the angle of the principal directions of the fibre reinforcement (i.e. the preferred fibre directions); second, using a dispersion model, we obtain dispersion parameters quantifying the collagen fibre dispersion about these principal directions; third, we calculate the weighted error entropy as a measure of changes in the entire fibre distributions at different length scales, as opposed to their average behaviour. With fully automated imaging techniques (such as multiphoton microscopy) becoming increasingly popular (which often yield large numbers of images to analyse), our method provides an ideal tool for quickly extracting mechanically relevant tissue parameters which have implications for computational modelling (e.g. on the mesh density) and can also be used for the inhomogeneous modelling of tissues.

A method for stochastic constrained optimization using derivative-free surrogate pattern search and collocation

Recent advances in coupling novel optimization methods to large-scale computing problems have opened the door to tackling a diverse set of physically realistic engineering design problems. A large computational overhead is associated with computing the cost function for most practical problems involving complex physical phenomena. Such problems are also plagued with uncertainties in a diverse set of parameters. We present a novel stochastic derivative-free optimization approach for tackling such problems. Our method extends the previously developed surrogate management framework (SMF) to allow for uncertainties in both simulation parameters and design variables. The stochastic collocation scheme is employed for stochastic variables whereas Kriging based surrogate functions are employed for the cost function. This approach is tested on four numerical optimization problems and is shown to have significant improvement in efficiency over traditional Monte-Carlo schemes. Problems with multiple probabilistic constraints are also discussed.

The impact of uncertainty on shape optimization of idealized bypass graft models in unsteady flow

It is well known that the fluid mechanics of bypass grafts impacts biomechanical responses and is linked to intimal thickening and plaque deposition on the vessel wall. In spite of this, quantitative information about the fluid mechanics is not currently incorporated into surgical planning and bypass graft design. In this work, we use a derivative-free optimization technique for performing systematic design of bypass grafts. The optimization method is coupled to a three-dimensional pulsatile Navier–Stokes solver. We systematically account for inevitable uncertainties that arise in cardiovascular simulations, owing to noise in medical image data, variable physiologic conditions, and surgical implementation. Uncertainties in the simulation input parameters as well as shape design variables are accounted for using the adaptive stochastic collocation technique. The derivative-free optimization framework is coupled with a stochastic response surface technique to make the problem computationally tractable. Two ...

Uncertainty quantification in coronary blood flow simulations: Impact of geometry, boundary conditions and blood viscosity.

Computational fluid dynamic methods are currently being used clinically to simulate blood flow and pressure and predict the functional significance of atherosclerotic lesions in patient-specific models of the coronary arteries extracted from noninvasive coronary computed tomography angiography (cCTA) data. One such technology, FFRCT, or noninvasive fractional flow reserve derived from CT data, has demonstrated high diagnostic accuracy as compared to invasively measured fractional flow reserve (FFR) obtained with a pressure wire inserted in the coronary arteries during diagnostic cardiac catheterization. However, uncertainties in modeling as well as measurement results in differences between these predicted and measured hemodynamic indices. Uncertainty in modeling can manifest in two forms - anatomic uncertainty resulting in error of the reconstructed 3D model and physiologic uncertainty resulting in errors in boundary conditions or blood viscosity. We present a data-driven framework for modeling these uncertainties and study their impact on blood flow simulations. The incompressible Navier-Stokes equations are used to model blood flow and an adaptive stochastic collocation method is used to model uncertainty propagation in the Navier-Stokes equations. We perform uncertainty quantification in two geometries, an idealized stenosis model and a patient specific model. We show that uncertainty in minimum lumen diameter (MLD) has the largest impact on hemodynamic simulations, followed by boundary resistance, viscosity and lesion length. We show that near the diagnostic cutoff (FFRCT=0.8), the uncertainty due to the latter three variables are lower than measurement uncertainty, while the uncertainty due to MLD is only slightly higher than measurement uncertainty. We also show that uncertainties are not additive but only slightly higher than the highest single parameter uncertainty. The method presented here can be used to output interval estimates of hemodynamic indices and visualize patient-specific maps of sensitivities.

Computational Simulation of the Adaptive Capacity of Vein Grafts in Response to Increased Pressure

Vein maladaptation, leading to poor long-term patency, is a serious clinical problem in patients receiving coronary artery bypass grafts (CABGs) or undergoing related clinical procedures that subject veins to elevated blood flow and pressure. We propose a computational model of venous adaptation to altered pressure based on a constrained mixture theory of growth and remodeling (G&R). We identify constitutive parameters that optimally match biaxial data from a mouse vena cava, then numerically subject the vein to altered pressure conditions and quantify the extent of adaptation for a biologically reasonable set of bounds for G&R parameters. We identify conditions under which a vein graft can adapt optimally and explore physiological constraints that lead to maladaptation. Finally, we test the hypothesis that a gradual, rather than a step, change in pressure will reduce maladaptation. Optimization is used to accelerate parameter identification and numerically evaluate hypotheses of vein remodeling.

Fast Computation of Hemodynamic Sensitivity to Lumen Segmentation Uncertainty

Patient-specific blood flow modeling combining imaging data and computational fluid dynamics can aid in the assessment of coronary artery disease. Accurate coronary segmentation and realistic physiologic modeling of boundary conditions are important steps to ensure a high diagnostic performance. Segmentation of the coronary arteries can be constructed by a combination of automated algorithms with human review and editing. However, blood pressure and flow are not impacted equally by different local sections of the coronary artery tree. Focusing human review and editing towards regions that will most affect the subsequent simulations can significantly accelerate the review process. We define geometric sensitivity as the standard deviation in hemodynamics-derived metrics due to uncertainty in lumen segmentation. We develop a machine learning framework for estimating the geometric sensitivity in real time. Features used include geometric and clinical variables, and reduced-order models. We develop an anisotropic kernel regression method for assessment of lumen narrowing score, which is used as a feature in the machine learning algorithm. A multi-resolution sensitivity algorithm is introduced to hierarchically refine regions of high sensitivity so that we can quantify sensitivities to a desired spatial resolution. We show that the mean absolute error of the machine learning algorithm compared to 3D simulations is less than 0.01. We further demonstrate that sensitivity is not predicted simply by anatomic reduction but also encodes information about hemodynamics which in turn depends on downstream boundary conditions. This sensitivity approach can be extended to other systems such as cerebral flow, electro-mechanical simulations, etc.

Real-Time Sensitivity Analysis of Blood Flow Simulations to Lumen Segmentation Uncertainty

Patient-specific modeling of blood flow combining CT image data and computational fluid dynamics has significant potential for assessing the functional significance of coronary artery disease. An accurate segmentation of the coronary arteries, an essential ingredient for blood flow modeling methods, is currently attained by a combination of automated algorithms with human review and editing. However, not all portions of the coronary artery tree affect blood flow and pressure equally, and it is of significant importance to direct human review and editing towards regions that will most affect the subsequent simulations. We present a data-driven approach for real-time estimation of sensitivity of blood-flow simulations to uncertainty in lumen segmentation. A machine learning method is used to map patient-specific features to a sensitivity value, using a large database of patients with precomputed sensitivities. We validate the results of the machine learning algorithm using direct 3D blood flow simulations and demonstrate that the algorithm can predict sensitivities in real time with only a small reduction in accuracy as compared to the 3D solutions. This approach can also be applied to other medical applications where physiologic simulations are performed using patient-specific models created from image data.

Automated estimation of image quality for coronary computed tomographic angiography using machine learning

ObjectivesOur goal was to evaluate the efficacy of a fully automated method for assessing the image quality (IQ) of coronary computed tomography angiography (CCTA).MethodsThe machine learning method was trained using 75 CCTA studies by mapping features (noise, contrast, misregistration scores, and un-interpretability index) to an IQ score based on manual ground truth data. The automated method was validated on a set of 50 CCTA studies and subsequently tested on a new set of 172 CCTA studies against visual IQ scores on a 5-point Likert scale.ResultsThe area under the curve in the validation set was 0.96. In the 172 CCTA studies, our method yielded a Cohen’s kappa statistic for the agreement between automated and visual IQ assessment of 0.67 (p < 0.01). In the group where good to excellent (n = 163), fair (n = 6), and poor visual IQ scores (n = 3) were graded, 155, 5, and 2 of the patients received an automated IQ score > 50 %, respectively.ConclusionFully automated assessment of the IQ of CCTA data sets by machine learning was reproducible and provided similar results compared with visual analysis within the limits of inter-operator variability.Key points• The proposed method enables automated and reproducible image quality assessment.• Machine learning and visual assessments yielded comparable estimates of image quality.• Automated assessment potentially allows for more standardised image quality.• Image quality assessment enables standardization of clinical trial results across different datasets.