Quantitative Biosciences Thesis Proposal
Hector Augusto Velasco-Perez
School of Physics
Advisor: Dr. Flavio H. Fenton (Schools of Physics)
Open to the Community
Spatio-temporal reconstruction of electrical reentrant waves in cardiac tissue
Monday, September 24th, 2018
4:00 pm
Howey Building, Room N110
Committee Members:
Dr. Shahriar Iravanian, School of Medicine; Emory University
Dr. Simon Sponberg, School of Physics; Georgia Institute of Technology
Abstract:
Atrial fibrillation affects millions of people worldwide and, left untreated, it increases the risk of stroke, heart failure and death. Clinical procedures are restricted to only extracellular measurements on the endocardial surface of the atria using unipolar or bipolar electrodes. These electrodes allow the clinicians to have accurate voltage time series recordings, but they do not allow them to sample many areas simultaneously. We aim to reconstruct the spatio-temporal electrical wave dynamics by applying the dynamic mode decomposition (DMD) to reduce the dimensionality of the system and analyze the relevant features. This process will consist of three steps: First, data will be generated in silico by running computer simulations with a realistic physiological model. The data collected will be used to generate time series at certain locations of the tissue to resemble electrode recordings. We will consider the spatio-temporal data and electrode data as separate sets of data. At this stage, the DMD method will generate an optimal spatio-temporal reduced model approximation of the system. Next, we will compare and correlate the features of the modes and eigenvalues between both data sets and on themselves. This process will be repeated with experimental data, were we will generate the spatial data by preforming optical mapping on the tissue. We will identify the relevant features and the relationship between them to reconstruct the spatio-temporal solution assuming we only have access to electrode data. Since we are targeting medical applications, we are concerned about the speed of our computations. As preliminary work, we have developed high speed and accurate simulation tools with parallel graphic processing units architectures. Our software can incorporate complex geometries, fiber anisotropy, and solves the equations with several numerical schemes.
