(D-152) Development of an Image Analysis Pipeline for Unbiased Quantification of Skeletal Muscle Cross-Sectional Area

Friday, October 13, 2023

3:30 PM – 4:30 PM PDT

Location: Exhibit Hall - Row D - Poster # 152

Presenting Author(s)

KV

Keerthana Vijayaragavan

Undergraduate Student
University of Virginia
Chantilly, Virginia, United States

Co-Author(s)

MH

Megan Haase

PhD Candidate
University of Virginia, United States
JD

Jacob Dunn

Lab Specialist
University of Virginia, United States

Primary Investigator(s)

SB

Silvia Blemker

Professor of Biomedical Engineering
University of Virginia, United States

Introduction:: Preferred Subtrack: Biomechanics of Rehabilitation/Injury

Following an acute injury, skeletal muscle has a remarkable regenerative capability which can be assessed by analyzing changes in cross-sectional area (CSA) which are known to be correlated with functional force recovery1. Therefore, CSA quantification is important for understanding muscle regeneration. Previous studies have shown that the software, Cellpose is a useful tool for unbiased quantification of myofibers2. However, there is still a lack of knowledge and consistency across studies surrounding the best way to measure injury with histology. Current methods generally measure injury from cropped images; however, the process of selecting crops is not systematic or clear which can introduce bias and error. The goal of this project was to develop and evaluate a new framework for quantification of fiber CSA during muscle generation. The method uses a novel and semi-automatic muscle fiber quantification method. The proposed pipeline for CSA quantification has strict guidelines to maintain consistency across images and an unbiased method for selecting a location of the image for quantification through analyzing both whole and localized sections to determine the significance of the area quantified. We hypothesize that unbiased cropping and full section analysis will show differences in CSA quantification. This research will aid in accurately processing data to better understand skeletal muscle regeneration through the changes in CSA.

Materials and Methods:: Mice gastrocnemius (n=3) were injected with cardiotoxin to induce acute injury, and tissue was harvested at days 1 and 7 post injury. The tissue was then sectioned at 10 um thickness and stained with DAPI and Laminin-alpha 2. The slides were imaged using confocal microscopy at 20X magnification with a resolution of 512 x 512. The images were then processed via a pretrained Cytoplasm Cellpose model to develop masks of the muscle sections. Images were processed as whole muscles as well as superficial, the outer region of the muscle, and deep, the inner region of the muscle, cropped sections3. The cropped images were selected using a custom MATLAB script and a standard operating procedure was developed to maintain consistency across images. This included rules such as that any longitudinal fibers were removed as part of the data and a target ROI of 200-250 per image was set. The MATLAB script outputted 5 crops and the image with the greatest number of ROIs was chosen. The masks were used in FIJI’s plugin Labels to ROIs to quantify CSA. The masks were eroded by a scale of 4 to account for thickness of the mask fiber outline as previous studies have indicated that this makes a significant difference in CSA quantification2. A MATLAB script was used to find the average area and average minimum feret diameter from the Labels to ROIs output. The difference in location was compared as well as lateral versus medial to determine the uniformity of damage.

Results, Conclusions, and Discussions:: There were appreciable differences in CSA of lateral and medial muscles for both injured and healthy tissue. Looking at the fiber size distribution plot, comparing uninjured to injured muscles, it is seen especially in the medial sections that injured tissue is right skewed, indicating that more fibers have smaller average CSA. As seen in the figure, in the injured medial, 178 fibers fell within the bin of 100-300 um2, and as the CSA increased the number of fibers that fell within each range gradually decreased. In the uninjured medial, however, there was a more normal distribution of fiber number in relation to CSA. This finding was similar in both day 1 and day 7 time points. The medial tissue showed a shift from day 1 to day 7 in the average number of fibers decreasing per CSA. This trend did not hold for the lateral tissue which can be attributed to the fact that the whole tissue section was analyzed and lateral tissue is significantly larger than medial tissue. This indicates that the injury was more uniform in the medial tissue compared to the lateral.This experiment showed that a strict list of conditions needs to be set when measuring skeletal muscle CSA to provide unbiased data as conditions such as the muscle (lateral and medial) and type of image (whole and cropped) can alter the results. As seen in the results, the injured medial tissue showed a steeper decline in average CSA. A consistent method of data quantification can aid in producing accurate results across muscle sections. This new method for finding skeletal muscle CSAs can be used for future work in accurately studying muscle regeneration outcomes as well as streamline data processing and maintaining consistency across different experiments and researchers.

Acknowledgements (Optional): : Research was supported by funding from the Dean's Undergraduate Research Fellowship. A special thanks to members of the M3 Lab including Brendan Shea and Josiah Calhoun.

References (Optional): :

Häkkinen K., et al. (1991), Eur J Appl Physiol Occup Physiol.

Waisman, A., et al. (2021), Sci Rep

Mintz E., et al. (2020), Tissue Engineering Part A