Computational fluid dynamics
Harnessing the power of CFD to
solve complex clinical problems.
Benefit from biofluid mechanics modelling that facilitates breakthroughs in personalized medicine.
AI-enhanced CFD assessment of CAD in CCTA
We tackled issues related to coronary artery disease diagnosis and monitoring and introduced an end-to-end deep learning-powered pipeline for automated analysis of coronary computed tomography angiography which additionally exploits computational fluid dynamics to capture the functional vessel characteristics. Our experiments, performed over clinically acquired scans, revealed that our segmentation approach (that takes advantage of centerlines instead of performing full manual segmentation) not only outperforms state-of-the-art nnU-Nets but also leads to the blood-flow parameters which are in substantial agreement with those elaborated for the ground-truth delineations.
Here green streamlines are Ground Truth (GT), red: Centerlines (CL) straightened and blue is nnUnet. Flow is steady and driven by the same pressure. Blue streamline is significantly faster. We consider the coronary tree from the inlet to the points where the diameter is 1.5mm.
Computing hemodynamic parameters
Hemodynamic parameters such as velocity, pressure, vortices, wall shear stress can be computed during computer simulation and so on, and are considered for the characterization of the flow field. Results showed that the vortex location and inflow area depend strongly on even a subtle changes of vessel geometry.
Blood flow simulation in cerebral aneurysm
Vortex structure reveals abnormal areas where flow detachment occurs