Image processing for TPEF/SHG laser scanning microscope

• Main Author: Tham, Yi Hui
• Other Authors: Zhang Yilei
• Format: Final Year Project
• Language: English
• Published: 2014
• Subjects: DRNTU::Engineering::Mechanical engineering
• Online Access: http://hdl.handle.net/10356/60314
• Institution: Nanyang Technological University

Collagen is the most abundant protein in vertebrate animals. It makes up 25% to 35% of the whole-body protein content and plays important structural and functional roles in tissues and organs. For example, collagen can be generally found in connective tissues such as skin, tendon and ligament, providing strength and elasticity. Besides providing structural strength, it has been found that collagen there exist a relationship between collagen organization in tissue and its function. For example, imaging has shown that in normal tendons, collagen fibers are organized regularly. On the other hand, organization of collagen in injured tendons is random. Imaging of collagen has been developed to allow and enhance biological and medical research. Second harmonic generation (SHG) microscopy emerged as an established and powerful contrast mechanism for visualising fibrillar collagen in tissues. The disadvantage of this imaging method is that it has an extremely small field of view. In order to obtain meaningful information about collagens, a series of overlapping SHG images has to be captured to allow one to see the distribution of collagen. This project aims to design a MATLAB programme to process and stitch SHG images of collagen together effectively. The method used to build this programme is by putting together several algorithms and making them work together. Algorithms of histogram equalization and median filtering are first applied to the SHG images to adjust the contrast so that details in the images are visible. Following that, interest points are detected in images using Harris Corner Detector. These interest points are then described by SIFT algorithm for the purpose of matching interest points of one image to that of another image. Using RANSAC, a homography transformation between two images is estimated from the set of matching interest points. Lastly, the images are projected using the estimated homography transformation to produce the final stitch result. This final stitch results provides a wider view of tissue sample, and thus enhances the study of collagen.