Date

2016

Document Type

Dissertation

Degree

Doctor of Philosophy

Department

Computer Engineering

First Adviser

Huang, Xiaolei

Other advisers/committee members

Vavylonis, Dimitrios; Spletzer, John; Lopresti, Daniel

Abstract

Curvilinear biopolymer networks pervade living systems. They are routinely imaged by fluorescence microscopy to gain insight into their structural, mechanical, and dynamic properties. Image analysis can facilitate understanding the mechanisms of their formation and their biological functions from a quantitative viewpoint. Due to the variability in network geometry, topology and dynamics as well as often low resolution and low signal-to-noise ratio in images, segmentation and tracking networks from these images is challenging. In this dissertation, we propose a complete framework for extracting the geometry and topology of curvilinear biopolymer networks, and also tracking their dynamics from multi-dimensional images. The proposed multiple Stretching Open Active Contours (SOACs) can identify network centerlines and junctions, and infer plausible network topology. Combined with a $k$-partite matching algorithm, temporal correspondences among all the detected filaments can be established. This work enables statistical analysis of structural parameters of biopolymer networks as well as their dynamics. Quantitative evaluation using simulated and experimental images demonstrate its effectiveness and efficiency. Moreover, a principled method of optimizing key parameters without ground truth is proposed for attaining the best extraction result for any type of images. The proposed methods are implemented into a usable open source software ``SOAX''. Besides network extraction and tracking, SOAX provides a user-friendly cross-platform GUI for interactive visualization, manual editing and quantitative analysis. Using SOAX to analyze several types of biopolymer networks demonstrates the potential of the proposed methods to help answer key questions in cell biology and biophysics from a quantitative viewpoint.

Share

COinS