Simulations and machine learning going hand in hand for clinical medicine

The 23rd iTHES Theoretical Science Colloquium

Recent rapid progress of AI technologies has strongly affected the medical community, profoundly enhancing medical image analysis as well as improving decision-making in clinical practice. Nevertheless, black-box systems cannot be accepted easily in clinical medicine because of issues related to accountability and incorporation of new and rapidly developing medical technologies.

This talk presents a bilateral approach to cardiovascular problems consisting of (1) machine learning approach for estimation of fluid dynamical forces such as wall shear stress and oscillatory shear index by using geometrical information of the vessels; and (2) simulation approach for understanding physical mechanisms, from vessel geometry to wall forces distributions via flow patterns, using fluid–structure interaction analysis based on partial differential equations. This work was conducted as part of our JST-CREST project: "New challenges for mathematical modeling in clinical medicine."