Abstract Body (Do not enter title and authors here): Introduction

Hemodynamics and wall biomechanics jointly drive ascending-aortic remodeling in bicuspid aortic valve (BAV) disease, yet most in silico studies model only flow or strain. Manual segmentations yield noisy meshes, impeding strain and computational fluid dynamics (CFD) analyses. We present an end-to-end framework integrating patient-specific CFD and dynamic strain mapping via our Bayesian remeshing algorithm, preserving geometry and eliminating manual mesh edits.

Objective

Develop and validate an end-to-end approach to automate mesh generation, quantify surface strain, and enable strain mapping and patient-specific CFD in ascending aortas of normal and bicuspid aortic valves.

Methods

We retrospectively analyzed 12 ECG-gated 4D CT datasets from 11 adults in a BAV study. Two observers performed semi-automated segmentation in ITK-SNAP using manual annotation, random forest voxel classification, and contour evolution; interobserver agreement—Dice coefficient, mean surface distance (MSD)—is reported as median [IQR]. Segmentations were converted to surface meshes and remeshed via a Bayesian adaptive algorithm. A reference mesh was propagated across cardiac phases using deformable registration; registration accuracy was assessed by MSD against meshes from manual segmentations. Areal strain was computed as fractional area change from end-diastole (Figure 1). Transient-dynamic CFD was performed on three patients to simulate flow and wall shear stress (WSS); aortic growth, reported in Figure 2, was derived from follow-up CT-based maximum diameters.

Results

Table 1 summarizes patient characteristics. Manual segmentations showed strong interobserver agreement (Dice 0.96 [0.95–0.99]; MSD 0.45 mm [0.17–0.70]). Propagated meshes had a median MSD of 0.42 mm [IQR 0.34–0.74] (n = 24) against manual references. Peak areal strain ranged 7–23%, and WSS 2–6 Pa.

Figure 2 shows strain and WSS maps for two representative BAV cases. In these cases, lower peak WSS/velocity coincided with faster radial growth over subsequent scans, mirroring inverse WSS-growth trends seen in larger BAV cohorts.

Conclusions

Our 4D CT workflow produces spatially aligned WSS and areal-strain maps, without manual mesh edits. Preliminary results suggest higher peak WSS may be negatively associated with aortic growth rate, consistent with prior studies. This proof-of-concept demonstrates technical feasibility and motivates further study of WSS-strain interaction as a predictor of BAV aortic remodeling.