Computational Fluid Dynamics in Tetralogy of Fallot: A Method to Generate Pseudo Patient-Specific Geometries
Louvelle, Leslie 1 ; Doyle, Matthew 2, 3 ; Van Arsdell, Glen 4, 5 ; Amon, Cristina 1, 2
1. Institute of Biomaterials and Biomedical Engineering, University of Toronto; 2. Department of Mechanical and Industrial Engineering, University of Toronto; 3. University Health Network, Toronto; 4. The Hospital for Sick Children, Toronto; 5. Division of Cardiac Surgery, Department of Surgery, University of Toronto
Tetralogy of Fallot (TOF) is the most common cyanotic congenital heart defect, accounting for 7-10% of all congenital heart disease. These patients require surgical repair within the first year of life. Although certain surgical techniques have shown improved long-term outcomes, the suitability of these techniques for individual patients is unknown preoperatively. Virtual surgery techniques, involving the modification of preoperative patient-specific models (PSMs) to reflect the hypothetical postoperative anatomy, could predict an optimal repair strategy for each patient. Our goal is to facilitate the development of TOF virtual surgery techniques via simplified yet representative versions of patient geometry, called pseudo patient-specific models (PPSMs). Here, we present a method of generating PPSMs from patient imaging data.
5 TOF patient geometries (13.2±1.9 years after surgical repair) were used in this study. For each patient, the postoperative cardiac MRI was uploaded into Simvascular. Contours of the right ventricular outflow tract extending to the left and right pulmonary arteries (LPA, RPA) were traced and subsequently lofted to generate 3D models. Each of these PSMs was smoothed in MeshLab and the inlets and outlets extended using SolidWorks. Planes normal to the vessel centreline were constructed in SolidWorks at key segments in each PSM, including the infundibulum, pulmonary valve, main pulmonary artery, LPA and RPA. The lengths of these segments, as well as the hydraulic diameters of the vessels at each plane, were measured. Sets of circular profiles were created using these lengths and diameters. The profile sets were then lofted to generate the 5 PPSMs.
For each patient, simulations were completed in ANSYS Fluent for both the PSM and PPSM, under assumptions of laminar flow and Newtonian behaviour. The inlet mass flow boundary condition was derived from patient-specific heart rate and stroke volume; the outlet flow split boundary condition was derived from patient-specific MRI data.
Across all patients, the overlap between the PSM and the corresponding PPSM was 87±3%. The mean Hausdorff distance, a measure of the average separation of the vertex points of the two models, was 1.1±0.6%. For accurate comparison across a range in patient vessel size, this value is expressed as the deviation relative to the length of the bounding box diagonal. Qualitatively, we saw good agreement in the velocity values and pressure gradients across all models. The energy loss was also calculated for each simulated geometry; the mean percentage error in the energy loss values between each PSM and PPSM pairing was 7.4±5.6%.
We have developed and validated a novel method of generating pseudo patient-specific models of the TOF geometry, for incorporation into virtual TOF surgery. Future work will refine this method by increasing the sample size and including elliptical profiles to more accurately represent non-circular anatomical regions.