Does clinical data quality affect fluid-structure interaction simulations of patient-specific stenotic aortic valve models?

Numerical models are increasingly used in the cardiovascular field to reproduce, study and improve devices and clinical treatments. The recent literature involves a number of patient-specific models replicating the transcatheter aortic valve implantation procedure, a minimally invasive treatment for high-risk patients with aortic diseases. The representation of the actual patient's condition with truthful anatomy, materials and working conditions is the first step toward the simulation of the clinical procedure. The aim of this work is to quantify how the quality of routine clinical data, from which the patient-specific models are built, affects the outputs of the numerical models representing the pathological condition of stenotic aortic valve. Seven fluid-structure interaction (FSI) simulations were performed, completed with a sensitivity analysis on patient-specific reconstructed geometries and boundary conditions. The structural parts of the models consisted of the aortic root, native tri-leaflets valve and calcifications. Ventricular and aortic pressure curves were applied to the fluid domain. The differences between clinical data and numerical results for the aortic valve area were less than 2% but reached 12% when boundary conditions and geometries were changed. The difference in the aortic stenosis jet velocity between measured and simulated values was less than 11% reaching 27% when the geometry was changed. The CT slice thickness was found to be the most sensitive parameter on the presented FSI numerical model. In conclusion, the results showed that the segmentation and reconstruction phases need to be carefully performed to obtain a truthful patient-specific domain to be used in FSI analyses.