A semi‐automated pipeline for quantitation of Pax7+, myonuclei, and cross‐sectional area by fiber type

FIGURE 1. Vastus lateralis cross‐section histology. Laminin stain is shown in white, and parameters of interest are shown in red. Images taken at 20×. Scale bars = 100 μm. (a) Red overlay shows segmentation of myofibers by CellProfiler, after image pre‐processing. (b) DAPI. (c) Myosin Heavy Chain I indicating type I fibers. (d) White arrows within callout indicate Pax7+ satellite cells.

FIGURE 2. Imperfect images can be segmented by the program with pre‐processing. Red staining shows laminin (cell borders) and green staining shows DAPI (nuclei). Scale bar = 100 μm. (a) Raw image overlay of the imperfect edge of a tissue section at 20×. (b) The same image overlay post‐cleaning in Fiji. (c) Labeled cells identified by the pipeline from tissue section shown in a.

FIGURE 3. Myofiber segmentation with Cellpose “cyto2” detection model. (a) White staining shows fluorescent laminin signal, with red overlay showing cells accepted by the pipeline. This overlay is saved for each image set to allow for manual quality control. Scale bar = 100 μm. (b) Grayscale object masks generated and saved by the pipeline for use in the Python script. Scale bar = 100 μm. (c) Masks identified by the pipeline are numbered by the Python script so that each cell within each image has a unique identifier, allowing for manual quality controls. Scale bar = 100 μm.

FIGURE 4. Nuclear segmentation with Cellpose “nuclei” detection model, after second iteration of RunCellpose module. (a) White staining shows fluorescent DAPI signal, with red overlay showing nuclei accepted by the pipeline. This overlay is saved for each image set to allow for manual quality control. Scale bar = 100 μm. (b) Callout from red box in (a). Scale bar = 100 μm. (c) Grayscale object masks generated and saved by the pipeline for colocalization with myofibers in the Python script. Scale bar = 100 μm.

FIGURE 5. Satellite cell segmentation with IdentifyPrimaryObjects module. (a) White staining shows fluorescent Pax7+ signal, with red overlay showing satellite cells accepted by the pipeline. This overlay is saved for each image set to allow for manual quality control. Scale bar = 100 μm. (b) Callout from red box in (a). Scale bar = 100 μm. (c) Grayscale object masks generated and saved by the pipeline for colocalization with myofibers in the Python script. Scale bar = 100 μm.

FIGURE 6. Intensity threshold for automated fiber typing. Analysis of intensities within segmented myofibers from 3722 muscle cells from 12 human biopsies was used to establish a threshold between type I and type II fibers. The frequency distribution shows two groups, with type II fibers being the group below an intensity of 20 and type I fibers being the group above an intensity of 20. The dotted line represents the threshold at pixel value 20.

FIGURE 7. Comparison of nuclei/cell counted manually versus by the pipeline (N = 310). (a) Bland Altman 95% limits of agreements plot comparing manual nuclear counting to pipeline nuclear counting. The frequency of data points is represented by point size. (b) Average nuclei/cell as counted by manual analysis and by the pipeline. Paired t‐test resulted in a two‐tailed p‐value of 0.9686. Data is presented as mean ± standard deviation. (c) Frequency distribution of nuclei/cell for manual (black bars) and pipeline (white bars) analysis. (d) Correlation between nuclei/cell counted by the pipeline and by manual analysis; r = 0.4926 and p < 0.0001. Red line represents the line of identity.

FIGURE 8. Comparison of satellite cells/cell counted manually versus by the pipeline (N = 310). (a) Bland Altman 95% limits of agreements plot comparing manual satellite cell counting to pipeline satellite cell counting. The frequency of data points is represented by point size. (b) Average satellite cells/cell as counted by manual analysis and by the pipeline. Paired t‐test resulted in a two‐tailed p‐value of 0.5325. Data is presented as mean ± standard deviation. (c) Frequency distribution of satellite cells/cell for manual (black bars) and pipeline (white bars) analysis. (d) Correlation between satellite cells/cell counted by the pipeline and by manual analysis; r = 0.6762 and p < 0.0001. Red line represents the line of identity.

FIGURE 9. Comparison of fiber type assessed manually versus assessed by the pipeline (N = 310). (a) Bland Altman 95% limits of agreements plot comparing manual fiber typing to pipeline fiber typing. The frequency of data points is represented by point size. (b) Proportion of type I fibers as counted by manual analysis and by the pipeline. Paired t‐test resulted in a two‐tailed p‐value of 0.6555. Data is presented as mean ± standard deviation. (c) Frequency distribution of type I and type II fibers. For manual (black bars) and pipeline. Type I fibers have a binary fiber type of 1 (presence of signal), type II fibers have a binary fiber type of 0 (no signal present). (d) Correlation between pipeline and manual fiber typing; r = 0.9390 and p < 0.0001. Red line represents the line of identity.

FIGURE 10. Comparison of cross‐sectional area (CSA) measured manually versus by the pipeline (N = 310). (a) Bland Altman 95% limits of agreements plot comparing manual cross‐sectional area to pipeline cross sectional area. (b) Average cross‐sectional area counted by manual analysis and by the pipeline. Paired t‐test resulted in a two‐tailed p‐value less than 0.0001. Data is presented as mean ± standard deviation. (c) Frequency distribution of cross‐sectional area for manual (black bars) and pipeline (white bars) analysis. (d) Correlation between cross‐sectional area from the pipeline and measured manually; r = 0.9592 and p < 0.0001. Red line represents the line of identity.

FIGURE 11. Comparison of perimeter measured manually versus by the pipeline (N = 310). (a) Bland Altman 95% limits of agreements plot comparing manual perimeter pipeline perimeter. (b) Average perimeter counted by manual analysis and by the pipeline. Paired t‐test resulted in a two‐tailed p‐value less than 0.0001. Data is presented as mean ± standard deviation. (c) Frequency distribution of perimeter for manual (black bars) and pipeline (white bars) analysis. (d) Correlation between perimeter from the pipeline and measured manually; r = 0.9354 and p < 0.0001. Red line represents the line of identity.

FIGURE 12. Manual versus CellProfiler cell outlines. (a) Manually drawn outlines shown in white, drawn over the raw laminin stain. Brush width used in Fiji was 10 pixels. Scale bar = 100 μm. (b) Automated outlines shown in red, drawn over the raw laminin stain shown in (a). Line width used by CellProfiler was 1 pixel. Scale bar = 100 μm. (c) Callout from white box in (a). Scale bar = 100 μm. (d) Callout from white box in (b). Scale bar = 100 μm.