CFD validation using in-vitro MRI velocity data – Methods for data matching and CFD error quantification

• Published in: Computers in biology and medicine Vol. 131; p. 104230
• Main Authors: Wüstenhagen, Carolin, John, Kristine, Langner, Sönke, Brede, Martin, Grundmann, Sven, Bruschewski, Martin
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01.04.2021; Elsevier Limited
• Subjects: Accuracy; Algorithms; Blood flow; Boundary conditions; Computational fluid dynamics; Computational fluid mechanics; Computer applications; Datasets; Error analysis; Flow geometry; Flow velocity; Geometry; Internal Medicine; Iterative algorithms; Magnetic resonance imaging; Magnetic resonance velocimetry; Mathematical models; Noise; Numerical models; Other; Patients; Reynolds similarity; Simulation; Simulation error; Three-dimensional geometry matching; Velocity; Three-dimensional geometry matching; Simulation error; Reynolds similarity; Computational fluid mechanics; Magnetic resonance velocimetry; three-dimensional geometry matching; Keyterms: Computational Fluid Mechanics; simulation error
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2021.104230

Predicting blood flow velocities in patient-specific geometries with Computational Fluid Dynamics (CFD) can provide additional data for diagnosis and treatment planning but the solution can be inaccurate. Therefore, it is crucial to understand the simulation errors and calibrate the numerical model. In-vitro velocity-encoded MRI is a versatile tool to validate CFD. The comparison between CFD and in-vitro MRI velocity data, and the analysis of the simulation error are the objectives of this study. A three-step routine is presented to validate medical CFD data. First, a properly scaled model of the patient-specific geometry is fabricated to achieve high relative resolution in the MRI experiment. Second, the measured flow geometry is matched with the CFD data using one of two algorithms, Coherent Point Drift and Iterative Closest Point. The aligned data sets are then interpolated onto a common grid to enable a point-to-point comparison. Third, the global and local deviations between CFD and MRI velocity data are calculated using different algorithms to reliably estimate the simulation error. The routine is successfully tested with a patient-specific model of a cerebral aneurysm. In conclusion, the methods presented here provide a framework for CFD validation using in-vitro MRI velocity data. [Display omitted] •Framework for CFD validation with in-vitro 3D MRI velocity data.•Automatic alignment of numerical and experimental data via data matching.•Quantification and visualization of the CFD simulation error.•Minimal user input enables automated evaluation.•Demonstration of the process with a patient-specific cerebral aneurysm model.