A0687
Title: Topological data analysis for statistical analysis of structure and dynamics in imaging
Authors: Andrew Thomas - University of Iowa (United States) [presenting]
Michael Jauch - Florida State University (United States)
David Matteson - Cornell University (United States)
Peter Crozier - Arizona State University (United States)
Abstract: The purpose is to discuss two separate statistical applications of a topological data analysis method introduced called detecTDA for detecting structure and change in a noisy image series. The main image series considered are extremely noisy and highly dynamic catalytic nanoparticle videos from transmission electron microscopy. First, the ability of the topological method to identify structure within the frames of these nanoparticle videos is examined, and whether an image is statistically distinct from one consisting purely of noise is assessed. The method is also demonstrated, along with the newly introduced changepoint method called bclr, pinpoints the statistically significant structural/topological features that govern a change in the state of the nanoparticle video (e.g. from ordered to disordered), concluding with a brief note on the software developed for these tasks.
