Synthetic dataset generation for the analysis and the evaluation of image-based hemodynamics of the human aorta

Here, we consider the issue of generating a suitable controlled environment for the evaluation of phase contrast (PC) MRI measurements. The computational framework, tailored to build synthetic datasets, is based on a two-step approach, i.e., define and implement (1) an accurate CFD model and (2) an image generator able to mime the overall outcomes of a PC MRI acquisition starting from datasets retrieved by the computational model. About 20 different datasets were built by changing relevant image parameters (pixel size, slice thickness, time frames per cardiac cycle). Focusing our attention on the thoracic aorta, synthetic images were processed in order to: (1) verify to which extent the fluid dynamics into the aortic arch is influenced by the image parameters; (2) establish the effect of spatial and temporal interpolation. Our study demonstrates that the integral scale of the aortic bulk flow could be described satisfactorily even when using images which are nowadays acquirable with MRI scanners. However, attention must be paid to near-wall velocities that can be affected by large inaccuracy. In detail, in bulk flow regions error values are well bounded (below 5% for most of the analyzed resolutions), while errors greater than 100% are systematically present at the vessel's wall. Moreover, also the data interpolation process can be responsible for large inaccuracies in new data generation, due to the inherent complexity of the flow field in some connected regions.