Seminar Series
Home → Seminar Series → Presentation

ICS import...
ICal Link

Friday April 6, 2018 -- Uncertainty-aware Personalized Modeling: Quantifying Parameter Uncertainty in Personalized Cardiac Models

CANCELLED, 11:45 am

speaker photo

Speaker: Linwei Wang, Associate Professor, Rochester Institute of Technology

Dr. Linwei Wang is an Associate Professor in the PhD Program of Computing and Information Sciences at the Rochester Institute of Technology in Rochester, NY. Her research interests center around statistical inference and its application to biomedical signal and image analysis, with a focus on improving patient care in cardiac arrhythmia and other heart diseases She currently directs the Computational Biomedical Lab at RIT. Her research is funded by the National Science Foundation and the National Institutes of Health. She is a recipient of the NSF CAREER Award in 2014. Dr. Wang obtained her bachelor degree in Optic-Electrical Engineering from Zhejiang University (China) in 2005, her master degree in Electronic and Computer Engineering from Hong Kong University of Science and Technology in 2007, and her PhD in Computing and Information Sciences from RIT prior to joining the faculty of RIT in 2009.

Presentation Abstract:

Personalized computer models are reaching the maturity to enter the era of predictive medicine to predict progression of disease and responses to therapeutic interventions, all done on a patient-specific basis at little or no risk to the patient. However, a key barrier to the widespread clinical adoption of simulation models arises from the uncertainty of the model predictions. One critical challenge towards attaching a confidence measure to these predictions is the difficult to quantify the sources of uncertainty within the model as they are being customized from patient-specific data. This is especially true for patient-specific tissue properties (in the form of model parameters), which are often not directly measurable but have to be inferred from indirect clinical data. Probabilistic estimation of these model parameters remains an unresolved challenge, because standard Markov Chain Monte Carlo sampling requires repeated model simulations that are computationally infeasible. In this talk, we describe our recent effort in probabilistic estimation of patient-specific tissue properties in a patient-specific cardiac model. We introduce a strategy of active Gaussian process (GP) construction to use a minimal amount of model simulations to build an efficient surrogate of the expensive posterior probabilistic density functions of the model parameters. We discuss alternative approaches to reach a low-dimensional (LD) representation of the high-dimensional (HD) parameter space, including a coarse-to-fine GP construction based an explicit multi-scale structure of the parameter space, versus

Faculty Host: Frank Sachse Contact: