Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Sensitivity Analysis of Simulated Blood Flow in Cerebral Aneurysms Øyvind Evju August 19, 2011 Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Aneurysms Blood Why simulate? Uncertainties Aim of this study What are aneurysms? An aneurysm is an abnormal bulge of a blood vessel. Most common in the Circle of Willis, part of the brains blood supply. These are called cerebral aneurysms. Large variations in size, up to above 50 mm in diameter. Several types of aneurysms: (a) Fuseform (b) Saccular/ sidewall Øyvind Evju (c) Saccular/ bifurcation SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Aneurysms Blood Why simulate? Uncertainties Aim of this study Rupture of cerebral aneurysms Common cause of subarachnoid hemorrhage. Often leads to serious brain damage or death. In a population of 100,000, about 10-11 cases of aneurysm rupture is expected. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Aneurysms Blood Why simulate? Uncertainties Aim of this study Risk factors and cause Some factors have been identified to increase proneness for aneurysm development and rupture: Environmental factors such as smoking, alcoholism and hypertension. Women are more prone to aneurysm rupture than men. People with an asymmetric or incomplete Circle of Willis more often develop aneurysms. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Aneurysms Blood Why simulate? Uncertainties Aim of this study Wall shear stress From a mechanical point of view, especially high values of wall shear stress (WSS) is indentified as a possible factor of aneurysm development. A surface force working tangential to the vessel wall. Induced by the blood flow. Calculated from the stress tensor. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Aneurysms Blood Why simulate? Uncertainties Aim of this study Blood Blood flow is complicated to simulate. Blood behaves as a non-Newtonian fluid. It is a heterogenous fluid, consisting mainly of blood cells (45%) and plasma (55%). It shows clear shear thinning properties, arising from concentration of red blood cells in the middle of the blood vessel. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Aneurysms Blood Why simulate? Uncertainties Aim of this study Blood viscosity The viscosity of is complex, and many models try to explain it. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Aneurysms Blood Why simulate? Uncertainties Aim of this study Why simulate? Better resolution than any measurement methods available. Minimal disturbance to the patient. Easily change physical parameters. Increased computational power yields greater accuracy. Assist medical personnel in making prognoses and determening treatment. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Aneurysms Blood Why simulate? Uncertainties Aim of this study Uncertainties There are many sources of errors present: Poor resolution of medical images. Little exact patient specific data available. Several simplifications and assumptions are made on the model. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Aneurysms Blood Why simulate? Uncertainties Aim of this study Aim of this study Assess qualitative and quantitative effects of several common simplifications and assumptions. Viscosity Geometry Boundary conditions Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion The Mathematical Model Implementation Verification of implementation Main assumptions Rigid walls. Body forces such as gravity are negligible. Incompressibility. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion The Mathematical Model Implementation Verification of implementation The Navier-Stokes equations ∂u 1 + u · ∇u = ∇ · 2ν(u) − ∇p ∂t ρ ∇·u=0 for x ∈ Ω u(x, 0) = 0 p(x, 0) = 0 u=0 for x ∈ Γw u = u0 for x ∈ ΓI p = p0 for x ∈ ΓO Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion The Mathematical Model Implementation Verification of implementation Problem ∂u ∂t + u · ∇u = ∇ · 2ν(u) − ρ1 ∇p ∇·u =0 ) for x ∈ Ω. Several difficulties: Nonlinear. Combination of a hyperbolic and a parabolic term. Two unknowns. Exact solutions exist only to simple problems. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion The Mathematical Model Implementation Verification of implementation Implementation The Incremental Pressure Correction Scheme was used. Implementation done using the finite element method, and the software library FEniCS. Source code modified from a previous project (nsbench). Meshes are built using tetrahedral cells. The solution is approximated by using polynomials at each cell. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion The Mathematical Model Implementation Verification of implementation An exact solution Fully developed, steady state flow in a straight channel/cylinder. Yields the exact solutions for velocity and WSS: r 2 − a2 dp 4µ dx a dp τw = 2 dx u= where dp dx is determined from the average flow velocity applied at the inlet. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion The Mathematical Model Implementation Verification of implementation Method A common set of parameters, modelling the middle cerebral artery (MCA). Tests were performed in both 2D and 3D. A range of different time steps were tested. Comparisons were made between a quadratic and a linear approximation to the velocity. The exact solution of the WSS was used as reference. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion The Mathematical Model Implementation Verification of implementation Resulting choices A timestep of 0.00125s was chosen. A linear approximation of the velocity was preferred to a quadratic approximation. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Method Simulation Results Method A single aneurysm was studied. The effects of three uncertainties were measured: Geometric effects Non-Newtonian effects Effects of different hematocrit levels Segmented out an aneurysm from CT-images using VMTK. (a) Cross section (b) Isosurface Øyvind Evju (c) Zoom-in SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Method Simulation Results Method Three different meshes were created, all with about 1,300,000 cells. A pulsatile flow profile was set at inlet, with a heart rate of 75bpm. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Method Simulation Results Simulation Simulation of blood flow through aneurysm at 75 bpm. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Method Simulation Results Simulation Simulation of blood flow through aneurysm, at 1/5th of the speed. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Method Simulation Results Results None of the three changes studied seem to have any significant effect on the overall flow pattern within the aneurysm. Locally, the effects on the velocity could be quite large. Areas of high flow velocity were less affected by the changes than areas of low velocity. Areas of high WSS were less affected by the changes than areas of low WSS. The effects were most prominent at systole. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Non-Newtonian effects Effects of increased hematocrit Effects of increased inflow Effects of a different outlet boundary condition The aneurysms (a) M1 (b) M2 (c) M3 (d) M5 (e) M8 (f) M9 (g) M11 (h) M12 (i) M15 (j) M16 (k) M18 (l) M20 Large variation in sizes and types. Used in a previous study at Simula. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Non-Newtonian effects Effects of increased hematocrit Effects of increased inflow Effects of a different outlet boundary condition Effects studied The four different effects which sensitivity analysis has been performed on are Neglecting the shear thinning (non-Newtonian) behaviour of blood. An increase in the hematocrit level from 38% to 40%. Increasing the inlet flux by 33%. Appliying a different set of outlet boundary conditions for the pressure. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Non-Newtonian effects Effects of increased hematocrit Effects of increased inflow Effects of a different outlet boundary condition Non-Newtonian effects Method: Comparing a Casson viscosity model to a reference Newtonian viscosity model with the same asymptotic viscosity at the limit of infinite shear rate. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Non-Newtonian effects Effects of increased hematocrit Effects of increased inflow Effects of a different outlet boundary condition Non-Newtonian effects - Results Large differences between the aneurysms: Aneurysms with a high average shear rate shows little non-Newtonian effects. Differences seems to be largest at diastole and early systole. Largest predicted change in average WSS: 2.24%. Largest predicted change in maximum WSS: -7.14%. Including non-Newtonian effects predicts a significantly lower maximum WSS (mean=-2.31%, P=0.0091). Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Non-Newtonian effects Effects of increased hematocrit Effects of increased inflow Effects of a different outlet boundary condition Effects of increased hematocrit Physiological motivation: The increase of two percentage points is an average increase seen in women going through menopause. The average age of menopause for women is 51.7 years, and the average age of aneurysm rupture is 52 years. This triggers a hypothesis of a correlation. Method: Using the Casson viscosity model (which incorporates the hematocrit level) the hematocrit is increased from 38% to 40%. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Non-Newtonian effects Effects of increased hematocrit Effects of increased inflow Effects of a different outlet boundary condition Effects of increased hematocrit - Results Large differences between the aneurysms. Predicted change in average WSS range from -3.2% to 5.2%. Significantly higher average WSS is predicted (mean=1.56, P=0.026). Predicted change in maximum WSS range from -12.7% to 5.7%. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Non-Newtonian effects Effects of increased hematocrit Effects of increased inflow Effects of a different outlet boundary condition Effects of increased inflow Method: Adjusting the inlet spatial peak velocity from an average of 535mm/s to 695mm/s. This corresponds to an increase in inlet flux of 33%. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Non-Newtonian effects Effects of increased hematocrit Effects of increased inflow Effects of a different outlet boundary condition Effects of increased inflow - Results The changes in WSS deviate highly from an expected linear relation between inlet flux and WSS. Predicted average WSS and maximum WSS increased by an average of 72.8% and 73.6% respectively. The changes in WSS seems greater within the aneurysm than in the surrounding arteries. All changes are highly significant. All aneurysms showed the same tendencies. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Non-Newtonian effects Effects of increased hematocrit Effects of increased inflow Effects of a different outlet boundary condition Effects of a different outlet boundary condition Method: Changing the outlet boundary conditions for pressure from a resistance boundary condition to a zero-pressure boundary condition. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Non-Newtonian effects Effects of increased hematocrit Effects of increased inflow Effects of a different outlet boundary condition Effects of a different outlet boundary condition - Results Differences in outlet flux of up to 231.6%. Significant changes in flow pattern in and after the bifurcation. Largest predicted change in average WSS: 36.15%. Largest predicted change in maximum WSS: 56.92%. Very large variation in the prediction of WSS. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Uncertainties regarding boundary conditions are far more important than uncertainties in flow parameters. Patient specific boundary data is absolutely necessary to accurately simulate cerebral blood flow accurately. Simulations might still be relevant even without patient specific boundary data. Non-Newtonian effects may safely be neglected if the boundary data is unknown ur uncertain. The WSS is unexpectedly sensitive to changes in inlet flux. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms Background And Motivation Numerical Methods Qualitative Analysis Quantitative Analysis Conclusion Thank you. Øyvind Evju SA of Simulated Blood Flow in Cerebral Aneurysms