论文和技术资料

In the medical field the risk assessment of strokes caused by pathological irregularities in the carotid artery bifurcation area plays a great role for the best patient’s treatment decision.The stroke risk evaluation by physicians traditionally incorporates evaluation of the stenosis degree in the respected area, flow velocity measurements during systole and diastole and objective charactericstics such as age, gender and many more.This decision-making process can be significantly improved with accurate flow field data obtained from simulations in patient-specific geometries. We developed a semi-automatic workflow for numerical data generation utilizing the COMSOL Multiphysics® software for a large set of 120 individual geometries. The exported datasets can either be used to later train a deep neural network or serve as a hemodynamic database for visual atlases and comparative analysis. Our semi-automatic workflow consists of 1. computational domain preprocessing and simulation parameter setup, 2. simulation and 3. postprocessing and data export. In the beginning, we start with COMSOL® preparations for simulation- and export-parameters. In this step, we use the centerline of each individual’s vessel tree to determine parameters for cutting the in- and outlets, analytically define areas around the bifurcation and the internal carotid area and also set up slicing planes in those areas to evaluate flow field variables. In addition, we calculate a three-dimensional vector field on the vessel surface, which is integrated in COMSOL® to evaluate a directional wall shear stress parameter (WSS) in the postprocessing step. The carotid artery model is then imported with the CAD Import Module and is geometrically adapted in the geometry node according to our predefined parameters from step 1. The blood flow is simulated by solving the incompressible Navier-Stokes equations for laminar flow. At the inflow boundary we choose a pulsatile flow velocity condition which mimics the systolic and diastolic flow volumes during a cardiac cycle. Of particular importance for the reliability of the simulation is the implementation of resistance boundary conditions for each of the outflow surfaces. These boundary conditions are used to ensure the realistic flow volume distribution between the internal and external subbranches of the carotid vessel tree. We simulate the blood flow in a time-dependent study over two complete cardiac cycles with the built-in PARDISO solver and the adaptive time-stepping BDF algorithm using P1+P1 elements. In step 3 of our workflow we evaluate and export computed and derived data. Beside the flow variables velocity and pressure we are particularly interested in the wall shear stresses (WSS) at the vessel surface.These data sets have been already used to successfully train a neural network which can categorize different stenosis degrees depending on the WSS. Furthermore, these simulations were the groundwork for an explorative visualization tool of the carotid morphology specifically designed for practiced physicians