Automated segmentation of collagen fibers from confocal reflectance images

• Main Author: Vaikundam Sriram
• Other Authors: Justin Dauwels
• Format: Theses and Dissertations
• Language: English
• Published: 2016
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: http://hdl.handle.net/10356/65901
• Institution: Nanyang Technological University

When cells migrate in the extracellular matrix (ECM), the fibers in the ECM gel are degraded, remodeled or realigned. Studying how the fibers remodel and degrade, is important to understand various biological processes such as tissue regeneration and wound healing. Experiments of cell migration in 3D gel have been performed in microfluidic devices. Time lapse microscopy images of the gel have been acquired at 60X magnification. Our goal is to detect the length and orientation of the fibers from these confocal reflectance images. The fiber extraction is challenging due to low signal to noise ratio, low resolution and cross linking of the fibers. Two methods for fiber extraction have been explored namely, Hough transform segmentation and directional region growing. The Hough transform was used to identify straight lines and separate fibers which are crossed linked. However, this method fails to detect fibers which are curved. Region growing can be used to segment curved fibers but fails to separate crossed-linked fibers. A directional region growing algorithm is proposed which limits the region grown by a range of angles. The angular range is updated as the curvature of the fiber changes. A comparative study of both the methods have been discussed.