Registration-based data assimilation from medical images
Please login to view abstract download link
Image-based, patient-specific modeling of hemodynamics can improve diagnostic capabilities and provide complementary insights to better understand the after-effects of treatments. However, computational fluid dynamics simulations remain relatively costly in a clinical context. Moreover, projection-based reduced-order models and purely data-driven surrogate models struggle due to the high geometric variability of biomedical datasets. A possible solution is shape registration: a reference template geometry is designed from a cohort of available geometries, which can then be diffeomorphically mapped onto it. This provides a natural encoding that can be exploited by machine learning architectures and, at the same time, a reference computational domain in which efficient dimension-reduction strategies can be performed~\cite{Alfonso2014}. We compare state-of-the-art graph neural network models with recent data assimilation strategies~\cite{Galarce2022} for the prediction of physical quantities and clinically relevant biomarkers in the context of aortic coarctation.
