Quantitative evaluation of skeletal muscle defects in second harmonic generation images

Abstract

Skeletal muscle pathologies cause irregularities in the normally periodic organization of the myofibrils. Objective grading of muscle morphology is necessary to assess muscle health, compare biopsies, and evaluate treatments and the evolution of disease. To facilitate such quantitation, we have developed a fast, sensitive, automatic imaging analysis software. It detects major and minor morphological changes by combining texture features and Fourier transform (FT) techniques. We apply this tool to second harmonic generation (SHG) images of muscle fibers which visualize the repeating myosin bands. Texture features are then calculated by using a Haralick gray-level cooccurrence matrix in MATLAB. Two scores are retrieved from the texture correlation plot by using FT and curve-fitting methods. The sensitivity of the technique was tested on SHG images of human adult and infant muscle biopsies and of mouse muscle samples. The scores are strongly correlated to muscle fiber condition. We named the software MARS (muscle assessment and rating scores). It is executed automatically and is highly sensitive even to subtle defects. We propose MARS as a powerful and unbiased tool to assess muscle health.

Fig. 1

Flow chart of the imaging processing. See Sec. 2 for details.

Fig. 2

$R$ and $S$ are sensitive to small and large defects in muscle fiber images. The left column shows SHG images from five human muscle biopsies. (a–c) are from normal adults; (d) and (e) from infants with Pompe disease. The image sizes are all $512\times 1024$. The middle column shows the texture correlation plots in space domain (solid line) with curve fitting (dotted line) and the $S$ scores. All figures were plotted with texture correlation ($T_d$) against the spacing (d) in pixels. The right column shows the texture correlation plots in the frequency domain and the $R$ scores. The plots are the amplitude ($T_u$) of the Fourier transform of the texture correlation against the spacing frequency (u) in ${\mathrm{pixel}}^{-1}$. From (a), which is a practically perfect muscle fiber, to (e), which is a very damaged muscle fiber, both $R$ and $S$ decrease by about three orders of magnitude. The stippled boxes in (d) and (e) show the region of interest which was selected for analysis in order to exclude the gap between fibers. The images were acquired under very similar magnification. The mean sarcomere spacing is 28 to 33 pixels and the number of GLCM matrices is close to 120. Scale bar: 10 μm.

Fig. 3

$R$ and $S$ are unaffected by changes in brightness. Panel a shows an SHG image whose average brightness was increased (b) or decreased (c) in ImageJ while keeping pixel brightness values between 0 and 255. Histograms are shown in the second column. The average brightness and standard deviations are in the third column. Both $R$ and $S$ show only negligible changes from (a) to (c).

Fig. 4

$R$ and $S$ are little affected by image noise. Panel a shows an SHG image before addition of 5% (b) or 50% (c) Gaussian noise. $R$ and $S$ change less than 2% as the noise level increases.

Fig. 5

$R$ and $S$ values for the same muscle fiber under different imaging conditions. (a) Muscle fiber was imaged under different conditions by adjusting laser power, PMT gain, and averaging method. Laser power was doubled from image a to b; $R$ and $S$ had less than 3% difference. PMT gain was set from 1075, to 800, to 600 in images b, c, and d, and the dramatic changes in the average brightness (from 150 to 19) did not lead to major changes in $R$ and $S$ values (within 20%). Most confocal and two-photon microscopes have line and frame averaging methods to reduce noise level. Two-line average and two-frame average were used in image c, which has obvious improvement on image quality by comparing to image e, which had no average method used while other conditions were the same. The $R$ and $S$ difference is very small (within 10%).

Fig. 6

$R$ and $S$ are affected by large magnification changes. Three samples with different muscle condition were imaged. To assess the effect of magnification, the same area was imaged with one objective lens ($40\times$) and six different zoom values from 1 to 6. The plots show the scores expressed in %, with the $6\times$ magnification results taken as 100%. $R$ and $S$ decrease gradually as a function of magnification. The changes vary among different samples and are stronger between $6\times$ and $3\times$ (26% average decrease for $R$; 19% for $S$). The results are also presented in Table 2.

Fig. 7

$R$ and $S$ scores are strongly correlated. To generate this plot, all $R$ and $S$ values compiled in the present work were pooled. A logarithmic plot was used to ease the representation of the large dynamic range of values. For the middle part of the plot ($S$ values from 10 to 0.1), the dots practically fall on a straight line which is along $\log S=\log R-1.7$, indicating a strong correlation. However, the distributions are scattered at the two ends, especially for fibers in very bad condition ($S<0.1$) indicating different sensitivities of $R$ and $S$.

Fig. 8

$S$ values highlight differences in muscle condition in a variety of human and mouse samples. SHG images were collected from different human and mouse samples, as indicated below the horizontal axis. From each sample at least 35 fiber images were analyzed and plotted in KaleidaGraph. The middle line in the box represents the median and the height of the box represents the standard deviation. Since $S$ and $R$ values show very similar trends, only $S$ is displayed here.