"Data-driven approaches for predicting and reconstructing cardiac electrical dynamics"
Oct 13, 2021 Schedule:
- Tea Time - F2F ( 417 DSL) / Virtual ( Zoom)
- 03:00 to 03:30 PM Eastern Time (US and Canada)
- Colloquium - F2F ( 499 DSL) / Virtual ( Zoom)
- 03:30 to 04:30 PM Eastern Time (US and Canada)
Meeting # 942 7359 5552
Integration of data to enhance or even replace mechanistic mathematical models of physical and biological processes is an important focus of modern scientific analysis. In this talk, we focus on applications to the heart, where complex dynamics including rapidly rotating spiral and scroll waves of electrical activity can inhibit contraction and may be lethal if untreated. Despite the prevalence of qualitative modeling studies, experimental limitations have hindered a detailed quantitative understanding of the specific mechanisms responsible for reentrant wave formation and breakup in practice. To further this effort, we assess the usefulness of data assimilation to reconstruct time series of electrical states in cardiac tissue. Using model-generated synthetic observations from a numerical experiment, we quantify the performance of an ensemble Kalman filter in reconstructing time series of both observed and unobserved variables. We illustrate how extensions using stochastic approaches can further reduce error in state estimates. As time permits, we will also show that machine-learning approaches can successfully predict complex single-cell cardiac voltage data and compare the performance of recurrent neural networks and echo state networks in such forecasting tasks.