I Congreso de Postgrado fcfm: ingeniería, ciencias e innovación

94 Santiago, 10 al 12 de agosto, 2022 PARAMETER ESTIMATION IN FLUID FLOW MODELS FROM ALIASED VELOCITY MEASUREMENTS Jeremías Garay¹*, David Nolte², Miriam Löcke¹, Cristóbal Bertoglio¹ 1 Bernoulli Institute, University of Groningen, The Netherlands . 2 Department for Fluid Dynamics, Technical University of Berlin, Germany. *Email: j.e.garay.labra@rug.nl ABSTRACT In blood f lows, the personalization of spatially distributed (i.e. 3D) models is a key step in performing predictive patient-specific simulations. Such a step relies on the formulation and numerical solution of inverse problems using clinical data, namely medical images for measuring both anatomy and function of the vasculature. In the context of 3D-0D coupled models, the personalization typically relies on estimating those 0D model parameters at each outlet boundary of the 3D model from velocity (and eventually pressure) data using non-linear least-squares approaches solved via variational [1] or sequential [2,3] methods. On the other hand, the gold standard for distributed blood f low velocity measurements in the clinical context is Phase-Contrast Magnetic Resonance Imaging (PC-MRI) [4,5]. However, PCMRI presents important artifacts, noise and velocity aliasing being the most important ones. When personalizing the models with such data, not taking them into account can render to important inaccuracies in the blood f low model personalization. Therefore, in this work we introduce a new but straightforward inverse problem formulation in order to effectively account for aliased velocity data. This is accomplished by a generalization of the cost function using the fact that phase-contrast problem accounts for multiple periodic solutions. This new formulation is naturally derived from the phase-contrast problem with the complex MRI signal as input. Numerical results in an aortic f low show robust parameter estimation for velocity encoding ranges until 30% of the maximal velocity of the problem, while the standard inverse problem fails already for any encoding velocity smaller than the true one. Moreover, the parameter estimation results are even improved for reduced velocity encoding ranges when using the new cost function. The presented approach allows therefore for great f lexibility in personalization of blood f lows models from MRI data commonly encountered in the clinical context. REFERENCES [1] Fevola, E. et al. P. Int. Journal for Numerical Methods in Biomedical Engineering 37.10 (2021) [2] Pant, S., Fabreges, B., Gerbeau, J-F, Vignon-Clementel, I. E. Int. Journal for Numerical Methods in Biomedical Engineering 30 (2021) [3] Arthurs, C., Xiao, N., Moireau, P., Schaeffter, T., Figueroa, C. Advanced modeling and simulation in engineering sciences. 7 (2021) [4] Markl M, Frydrychowicz A, Kozerke S, Hope M, Wieben O. Journal of Magnetic Resonance Imaging 36 (5) (2012) [5] Soulat G, McCarthy P, Markl M. Annual Review of Biomedical Engineering 22 (2020) F LU I DOD I NÁM I CA 06