Quantification of biopolymer filament structure

The quantitative analysis of a polymer network is important for understanding its role in biological function. We developed a Matlab program to recognize and segment filaments in a 2-D image, and measure and describe the structure. Our algorithm improved the speed of the Lichtenstein Fiberscore segmentation algorithm by using matrix convolutions, compared filament length by the algorithms of Kulpa, Lichtenstein, and Kimura, and measured the number of branchpoints and Euler number. A user interface was added to easily manipulate algorithm parameters, select images, and visualize results. We used the program to compare the DNA biopolymer network of cystic fibrosis (CF) sputum with mucus from patients without respiratory problems. We also examined an image of fibrin. The images were taken with a laser scanning confocal microscope after staining the specimens with Yo-Yo-1. Computation using matrix convolutions reduced the execution time (Pentium III) of a 512 x 512 TIF image from 18 min to 15s. The Kimura length estimation appeared best at describing filament length because it varied least with filament orientation. The image of CF sputum showed increased filament length, more branchpoints, and more negative Euler number compared to the normal sample. These quantitative descriptions of the network can be correlated to material, mechanical, diffusion, or flow properties, physiological processes, or therapy.