Poster J13 - MatLab-Based Automated Analysis Tool for Evaluating Fiber Alignment in Tendinous Scaffolds

Friday, October 25, 2024

3:30 PM - 4:30 PM EST

Location: Exhibit Hall E, F & G

Co-Author(s)

AM

Andrea M. McPherson

Undergraduate Student
University of Florida - Biomedical Engineering, United States

Presenting Author(s)

Harrison B. Broadaway (he/him/his)

PhD Student
University of Florida - Biomedical Engineering
Gainesville, Florida, United States

Primary Investigator(s)

Brittany L. Taylor, PhD (she/her/hers)

Assistant Professor
University of Florida - Biomedical Engineering, Florida, United States

Introduction: One of the main distinguishing factors between healthy and diseased tendons is the alignment and organization of the collagenous matrix. Healthy tendons consist of an aligned matrix composed of primarily collagen I fibers while diseased tendons exhibit disorganized collagenous fibers, mostly of collagen type III. To achieve biomimicry in our nanofibrous scaffolds, the rotations per minute (RPMs) of our scaffold fabrication protocol were adjusted when electrospinning, with literature demonstrating that a higher RPM generally corresponds to increased fiber alignment. If the orientation is quantified, we can use aligned scaffolds as a therapeutic to promote tendon repair, and scaffolds of both alignments can be used to model healthy and diseased tendons, respectively. As such, the objective of this study is to create an automated fiber alignment analysis tool that would allow us to determine whether scaffolds are representative of a healthy or injured tendon. It is hypothesized that polycaprolactone (PCL) scaffolds produced at 550 RPM will demonstrate more aligned characteristics than those spun at 150 RPM, demonstrated via quantification of fiber orientation.

Materials and

Methods: 33%, 50%, and 66% (w/v) PCL solutions were created by vortexing PCL with dichloromethane (DCM) and dimethylformamide (DMF) overnight. These solutions were then extruded onto a rotating cylindrical mandrel at a flow rate of 2.5 mL/hr with a 2kV voltage differential. The mandrel was spun at 550 and 150 RPM to produce aligned and unaligned scaffolds, respectively. Upon completion, scaffolds were stored in a desiccator for further usage. Scaffold samples (N = 18) were cut into 5 x 5 mm squares, sputter coated with gold-palladium nanoparticles, and imaged with Scanning Electron Microscopy (SEM) at 500x, 1200x, and 2500x objectives. The frequency of fibril alignment within 0° to 180° was quantified using a custom MATLAB program. The key components of the fiber orientation algorithm are SEM micrograph pre-processing, fiber edge detection, identification of key points on the fiber edge detection, and then angle calculation between the closest key points (Figure 1). Quantified fiber orientation for a perfectly aligned scaffold would display detected fibers in one degree, which would present as a vertical line at 0° (since it was adjusted for the largest number of angle instances to shift to 0°). For a perfectly unaligned scaffold, there would be an equal amount of angle instances in each degree, which would present as a horizontal line. Scaffolds spun at 550 RPM (n = 9) and 150 RPM (n = 9) with varying scaffold densities and SEM objectives were blindly evaluated utilizing this program.

Results, Conclusions, and Discussions: SEM micrographs were divided with respect to magnification, gaussian-filtered, and improved in contrast with the imadjust function (Figures 2A-B and 3). Canny edge detection with a threshold of 0.1 was then performed on the filtered, grayscale subimages (Figure 3). Morphological operations were applied to clean up skeletonized subimages, including removal of small objects, isolated pixels, and spurious branches (Figures 2C and 3). The key point detection function was used to find corners and long columns. Corner detection finds where black pixels and white pixels are diagonal to one another a square formation and saves the top white pixel as a key point. Column detection finds where there are five white pixels in a row and saves the third pixel as a key point. The closest neighbor was found for each key point and saved as pairs so the tangent angle could be taken between each unique key point pairing. The summed angle instances were then normalized by shifting the degree of maximum angle instances to 0° and calculated for the percentage of angles in each degree.
An example of 150 and 550 RPM scaffold histograms are shown in Figure 4, with the lower RPM angle summations being more dispersed, as expected. The average percentage of fibers within 10° of the neutral axis is 28.412% and 38.086% for 150 and 550 RPM scaffolds, respectively. This confirms the algorithm correctly quantifies increased alignment in the higher RPM scaffolds relative to the lower RPM scaffolds. 10%, however, is a relatively small difference in alignment, which may be attributed to outlier readings. Regardless, the algorithm performed best with 50% (w/v) scaffolds imaged at 1200x and 2500x magnification. Images that did not have sufficient image quality or were taken at 500x did not produce viable results due to negatively influencing image processing steps of the algorithm.
The results suggest the automated fiber orientation analysis tool has the potential to accurately quantify fiber orientation. Although currently best performing with a specific magnification and image quality, future implementations such as fiber tracking may improve algorithm processing and sequential quantification.

Acknowledgements (Optional):