Medical Image-Trained Patient Specific Reduced Order Modeling for Cardiovascular Hemodynamics Analytics

Cardiovascular diseases are affecting more people because of several factors including stress, lifestyle, and genetic linkages and are among the leading causes of death worldwide. With current clinical imaging techniques, it is difficult to derive a comprehensive assessment of pressure, flow rate, wall shear stress, and vascular resistance due to constraints of acquisition and analysis time. In this regard, Computational Fluid Dynamics (CFD) simulations can prove useful, although often bottlenecked by the lack of patient-specific insight. Here, we develop a reduced order modeling paradigm of hemodynamic transport considering deformable flow boundaries by extracting patient-specific features from MR Angiogram images, 4D flow MRI and other relevant meta-data. In doing so, first, image segmentation is performed to reconstruct the geometry from MR Angiogram images, and, subsequently, the relevant boundary conditions as well as wall mechanical properties are derived by utilizing 4D Flow MRI data and measured blood pressure from the brachial artery. This enables a patient-data-trained classification of the resulting flow features, providing clinicians with insight into patient-specific hemodynamics. The advantage of using reduced order modeling as opposed to more expensive 3D simulation is that it requires significantly less computational expense without sacrificing the essential physics of interest. The outcomes from simulation studies are shown to corroborate well with clinical features that can be extracted from 4D Flow MRI data and provide additional details which are not possible to obtain from even the most advanced in-vivo measurements. The paradigm of hemodynamic analytics, as outlined here, is likely to be of fundamental importance in providing physiological insight into patients suffering from various cardiovascular ailments such as aortic dissection, coarctation, and aneurysm, providing complementary physics-based insight into clinical decision-making, in addition to the more commonly practiced experience- and evidence-based principles.