Generation of personalized synthetic 3-dimensional inlet velocity profiles for computational fluid dynamics simulations of type B aortic dissection.

Computational fluid dynamics (CFD) simulations have shown promise in assessing type B aortic dissection (TBAD) to predict disease progression, and inlet velocity profiles (IVPs) are essential for such simulations. To truly capture patient-specific hemodynamic features, 3D IVPs extracted from 4D-flow magnetic resonance imaging (4D MRI) should be used, but 4D MRI is not commonly available. A new workflow was devised to generate personalized synthetic 3D IVPs that can replace 4D MRI-derived IVPs in CFD simulations. Based on 3D IVPs extracted from 4D MRI of 33 TBAD patients, statistical shape modelling and principal component analysis were performed to generate 270 synthetic 3D IVPs accounting for specific flow features. The synthetic 3D IVPs were then scaled and fine-tuned to match patient-specific stroke volume and systole-to-diastole ratio. The performance of personalized synthetic IVPs in CFD simulations was evaluated against patient-specific IVPs and compared with parabolic and flat IVPs. Our results showed that the synthetic 3D IVP was sufficient for faithful reproduction of hemodynamics throughout the aorta. In the ascending aorta (AAo), where non-patient-specific IVPs failed to replicate in vivo flow features in previous studies, the personalized synthetic IVP was able to match not only the flow pattern but also time-averaged wall shear stress (TAWSS), with a mean TAWSS difference of 5.9 %, which was up to 36.5 % by idealized IVPs. Additionally, the predicted retrograde flow index in both the AAo (8.36 %) and descending aorta (8.17 %) matched closely the results obtained with the 4D MRI-derived IVP (7.36 % and 6.55 %). The maximum false lumen pressure difference was reduced to 11.6 % from 68.8 % by the parabolic IVP and 72.6 % by the flat IVP. This study demonstrates the superiority of personalized synthetic 3D IVPs over commonly adopted parabolic or flat IVPs and offers a viable alternative to 4D MRI-derived IVP for CFD simulations of TBAD.