Patient-specific reconstruction of coronary artery wall motion: development of a computational framework for CFD applications

Cardiovascular diseases (CVD) are one of the leading causes of death in western countries. Myocardial Bridging (MB) is a condition in which a segment of a coronary artery dives into the myocardium and it has been linked to an increased risk of atherosclerosis (AS), coronary spasms and endothelial disfunction. Computational fluid dynamics (CFD) can provide access to hemodynamic parameters based on wall shear stress (WSS) metrics, which are difficult to measure in vivo but linked to AS progression. Coronary arteries have complex geometry and dynamic motion that include torsion, bending, and radial deformation. These shape changes significantly affect blood flow and disease development. Incorporating arterial wall motion into CFD models can improve accuracy, offering more realistic assessments of hemodynamics and better predictions of AS evolution. This work develops a computational framework to reconstruct patient-specific vessel wall motion and use it as a boundary condition in hemodynamic simulations of MB coronary models. Three coherent point drift (CPD) algorithm based methods were tested: centerline-CPD (cl-CPD), which tracks centerline displacement but ignores radial deformation; sample-CPD (s-CPD), which uses surface samples for higher accuracy; and sample local-CPD (sl-CPD), a hybrid approach applying s-CPD in the MB region and cl-CPD elsewhere, balancing accuracy and efficiency. The three reconstruction methods were used as boundary conditions for the moving-wall CFD simulations, using an interface-tracking Arbitrary Lagrangian-Eulerian approach. Hemodynamic factors, including WSS-based and bulk flow descriptors—time-averaged WSS (TAWSS), oscillatory shear index (OSI), relative residence time (RRT), and local normalized helicity (LNH) —were then computed to assess the impact of different wall displacements on coronary fluid dynamics. Comparative geometrical analysis of the reconstructed geometries based on Hausdorff distance (HD), Chamfer Distance (CD) and Point to Surface Distance (P2T) was performed. The cl-CPD introduced errors of over 2 mm. The s-CPD method proved to be a good approximation, with errors never exceeding 0.5 mm. The sl-CPD method introduced large deviations when considering the whole model (>1 mm); however, when considering only the MB segment the magnitude of errors between sl-CPD and s-CPD matched (<0.4 mm), proving that in the region of interest the geometry is tracked with accuracy. CFD simulations highlighted the influence of wall displacement on blood flow, using s-CPD as reference given its anatomical accuracy. The cl-CPD method introduced larger differences in both WSS magnitude (mean: 18.7%) and bulk flow descriptors distributions. The sl-CPD method closely reproduced the s-CPD results, with differences below 6%, showing a similar distribution of WSS-based descriptors as well as comparable patterns in bulk flow descriptors. Overall, the evidences indicate that s-CPD is the most accurate method to describe vessel motion and geometry, sl-CPD is a good compromise between geometrical accuracy and computational costs, while cl-CPD is not adequate to describe complex deformations, although representing the fastest method. These findings highlight the importance of incorporating patient-specific wall motion into CFD models of MB. Future developments should focus on further advancing and enhancing the presented methods, ultimately enabling accurate motion-informed simulations for clinical applications.