Data-driven Multiscale Modeling in Biological and Physical Systems
Biomedical Engineering Seminar
Alriza Yazdani, PhD
Abstract: First, I will discuss the recent developments in the study of complex biological systems at both micro- and macroscopic scales. Specifically, I will present data-driven multiscale modeling strategies to address the process of thrombus formation in aortic dissections using in-vivo and in-vitro data collected for murine dissections. This numerical framework elucidates hemodynamic and biochemical conditions under which a thrombus forms. Furthermore, inspired by the recent developments in physics informed learning machines to assimilate data, we have extended deep neural networks to a whole new set of problems for which the underlying conservation laws (i.e. mass, momentum, and energy) can be incorporated to infer hidden quantities of interest in different health and industrial applications. I will introduce a novel Navier-Stokes informed neural network that encodes the governing equations of fluid motion and infers hidden quantities of interest such as velocity and pressure fields merely from spatio-temporal visualizations of a passive scaler (e.g. dye or smoke) transported in arbitrarily complex domains. Finally, I will present multifidelity modeling with the help of Gaussian processes that allows us to use simple but partially accurate models with low computational cost, and to effectively enhance their accuracy by injecting a small set of high-fidelity observations. This new paradigm is significant as it allows us to efficiently quantify uncertainty in our models.
Dr. Alriza Yazdani is a faculty candidate in the Department of Biomedical Engineering
Tuesday, February 26, 2019 at 3:00 p.m.
Minerals and Materials Engineering Building (M&M), U113
1400 Townsend Drive, Houghton, MI 49931