Automated Analysis Method for Determining Tendon Collagen Fiber Orientation from Second Harmonic Generation Images


  • Thomas Elia, Biomedical Engineering, University of Delaware

Faculty Mentor(s)

  • Dawn Elliott, Biomedical Engineering, University of Delaware


Tendons are fibrous tissues that connect muscle to bone and for movement. Tendons consist of ropelike structures composed of collagen to transfer large forces. Overloading tendons can cause irreversible microstructural damage. Previously collagen organization has been assessed using histology staining such as picrosirius red which requires dehydration/rehydration of the tissue during staining and can alter the collagen structure. Second Harmonic Generation (SHG) is an imaging technique that is often used to produce high resolution images of collagen rich tissues without staining the collagen structure. Previously, collagen organization was determined by relying on blinded human graders to provide a semi-quantitative measure of fiber orientation. Therefore, the purpose of this study was to create an automated analysis method to determine collagen orientation in SHG images.

In this study, I developed an algorithm using Fourier transform analysis in MATLAB to quantify collagen fiber orientation in SHG images of rat plantaris tendon. Each image was split into blocks and a Fourier transform was applied and an ellipse was fitted to the frequency data. The minor axis of the ellipse was used to establish the fiber orientation angles. I validated this method using SHG images of a healthy plantaris tendon and a tendon that was cyclically loaded from 1.9N to 6.3N for 1000 cycles. Fiber orientations were displayed using a quiver plot and superimposed on the original image for manual validation. Additionally, the arrows were scaled based on the ratio between the major and minor axes length to provide a confidence index for each angle that was computed. Using this automated image analysis, we can determine fiber angles in SHG images, providing a quantitative measure of tendon fiber orientation. Future work will utilize this automated image analysis to compare fiber organization in overloaded plantaris tendons compared with controls.