Rupture of cerebral aneurysm is one of the major causes of subarachnoid hemorrhage, and quantifying patient-specific hemodynamics is a clinically important challenge. Because vascular geometry and blood flow conditions vary individually, computational fluid dynamics (CFD) based on medical images is a powerful tool for such quantification, but its high computational cost poses a problem in clinical applications. In this talk, I will introduce three practical strategies for hemodynamic quantification methods based on CFD-based data assimilation and physics-based neural network with a fine-tuning strategy.