SMASH - semi-automatic muscle analysis using segmentation of histology: a MATLAB application

Abstract

Background: Histological assessment of skeletal muscle tissue is commonly applied to many areas of skeletal muscle physiological research. Histological parameters including fiber distribution, fiber type, centrally nucleated fibers, and capillary density are all frequently quantified measures of skeletal muscle. These parameters reflect functional properties of muscle and undergo adaptation in many muscle diseases and injuries. While standard operating procedures have been developed to guide analysis of many of these parameters, the software to freely, efficiently, and consistently analyze them is not readily available. In order to provide this service to the muscle research community we developed an open source MATLAB script to analyze immunofluorescent muscle sections incorporating user controls for muscle histological analysis.

Results: The software consists of multiple functions designed to provide tools for the analysis selected. Initial segmentation and fiber filter functions segment the image and remove non-fiber elements based on user-defined parameters to create a fiber mask. Establishing parameters set by the user, the software outputs data on fiber size and type, centrally nucleated fibers, and other structures. These functions were evaluated on stained soleus muscle sections from 1-year-old wild-type and mdx mice, a model of Duchenne muscular dystrophy. In accordance with previously published data, fiber size was not different between groups, but mdx muscles had much higher fiber size variability. The mdx muscle had a significantly greater proportion of type I fibers, but type I fibers did not change in size relative to type II fibers. Centrally nucleated fibers were highly prevalent in mdx muscle and were significantly larger than peripherally nucleated fibers.

Conclusions: The MATLAB code described and provided along with this manuscript is designed for image processing of skeletal muscle immunofluorescent histological sections. The program allows for semi-automated fiber detection along with user correction. The output of the code provides data in accordance with established standards of practice. The results of the program have been validated using a small set of wild-type and mdx muscle sections. This program is the first freely available and open source image processing program designed to automate analysis of skeletal muscle histological sections.

Keywords: Histological muscle analysis; Image segmentation; Standardized quantitative analysis; mdx mouse.

Figure 1

Image selection. (A) Dialog box allowing selection of a single function to perform on the selected muscle section. (B) Example muscle section of 1-year-old soleus muscle from mdx mouse. Red is laminin stain, green is slow muscle myosin heavy chain, and blue is DAPI. (C) Enlarged portion of (B).

Figure 2

Initial segmentation. (A) Input box with options for fiber segmentation. (B) The example image with white lines drawn along fiber boarders based on segmentation parameters. (C) Dialog box that appears after separating a fiber. (D) Zoomed in of B. Red dotted lines shows where two fibers are separated by the user. Yellow dotted lines shows where a fiber is not separated from intestinal space.

Figure 3

Segmentation examples. Example muscle section of 1-year-old soleus muscle from mdx mouse. Red is laminin stain, green is slow muscle myosin heavy chain, and blue is DAPI. (A) Segmentation filter value of 2 produces over segmentation of original image. (B) Segmentation filter value of 5 produces appropriate segmentation. (C) Segmentation filter value of 12 produces under segmentation of original image. (D) Reducing the brightness of laminin staining by 50% (representing poor staining) produces under segmentation. (E) Enhancing the brightness of laminin by 400% (representing over exposure) produces over segmentation. (F) Overexposure can be compensated for by increasing the segmentation filter value to 12. Yellow arrow heads show areas of over segmentation. Pink arrow heads show areas of under segmentation.

Figure 4

Fiber filter. (A) Input box with options for filtering fibers. (B) Image generated in which fibers passing the filter are randomly colored to more easily distinguish adjacent fibers. (C) Zoomed in of (B). Yellow arrows point to possible user selections as non-fiber objects. Pink arrows show fibers that did not pass the filter. Lower left pink arrow did not pass due to improper segmentation as depicted in Figure 2D. (D) After selection of fibers the user is asked if they would like to indeed delete them or complete the function operation.

Figure 5

Fiber properties. (A) Input box with options for fiber properties. (B) Histograms showing minimum Feret diameter (top) and fiber CSA (bottom). (C) Truncated example of Excel output from running fiber properties function.

Figure 6

Fiber type. (A) Input box with options for fiber type function. (B) Truncated example of Excel output from running fiber type function. (C) Portion of immunofluorescent image. Red is laminin stain, green is slow muscle myosin heavy chain, and blue is DAPI (top left). Image showing negative fibers in gray and positive fibers in white (top right). Histogram of average fiber intensity for fiber type stain with threshold value (bottom left). Histogram of fiber CSA with positive fiber in red and all fibers in blue (bottom right).

Figure 7

Centrally nucleated fibers. (A) Input box with options for CNF function. (B) Truncated example of Excel output from running CNF function. (C) Portion of immunofluorescent image. Red is laminin stain, green is slow muscle myosin heavy chain, and blue is DAPI (top left). Image showing filtered nuclei (top right). Image with boarder regions in red and nuclei above threshold in blue (bottom left). Image in which only CNFs are depicted in white (bottom right).

Figure 8

Object counter. (A) Input box with options for object counter function. (B) Truncated example of Excel output from running object counter function. (C) Portion of immunofluorescent image. Red is laminin stain, green is PECAM, and blue is DAPI (top left). Image showing only the objects of interest, here PECAM from the green channel (top right). Image showing objects of interest after smoothing filter is applied (bottom left). Binary image showing discrete objects that pass threshold value (bottom right).

Figure 9

Comparison of output data from 1-year-old C57 and mdx mouse soleus muscle. (A) Mean of cross-sectional area of fibers in C57 and mdx muscle sections. (B) Standard deviation of CSA of fibers. (C) Mean of minimum Feret diameter (MFD) of fibers. (D) Standard deviation of MFD diameter. (E) Percentage of Type I fibers in C57 and mdx muscle sections. (F) The ratio of mean CSA of type I fibers to type II fibers. (G) The percentage of fibers with centrally nucleated fiber (CNF)s. (H) The mean CSA of mdx peripherally nucleated fibers (PNF) and CNFs. *P <0.05 for mdx compared to C57, † P <0.05 for CNFs compared to PNFs within the same muscle.

Figure 10

Comparison of mask file from legacy method and SMASH. (A) Original image of a laminin stained muscle section. (B) Fibers are colored randomly from SMASH output mask with dark regions corresponding to fiber area using a simple threshold. Dark gray areas are interstitial in SMASH and fibers using legacy methods while white areas are interstitial in both masks. Figure demonstrates the larger fiber area obtained with SMASH compared to the legacy method of using a simple threshold for fiber area.