Analysis of muscle tissue in vivo using fiber-optic autofluorescence and diffuse reflectance spectroscopy

Abstract

Significance: Current methods for analyzing pathological muscle tissue are time consuming and rarely quantitative, and they involve invasive biopsies. Faster and less invasive diagnosis of muscle disease may be achievable using marker-free in vivo optical sensing methods.

Aim: It was speculated that changes in the biochemical composition and structure of muscle associated with pathology could be measured quantitatively using visible wavelength optical spectroscopy techniques enabling automated classification.

Approach: A fiber-optic autofluorescence (AF) and diffuse reflectance (DR) spectroscopy device was manufactured. The device and data processing techniques based on principal component analysis were validated using in situ measurements on healthy skeletal and cardiac muscle. These methods were then applied to two mouse models of genetic muscle disease: a type 1 neurofibromatosis (NF1) limb-mesenchyme knockout (Nf1Prx1 - / - ) and a muscular dystrophy mouse (mdx).

Results: Healthy skeletal and cardiac muscle specimens were separable using AF and DR with receiver operator curve areas (ROC-AUC) of >0.79. AF and DR analyses showed optically separable changes in Nf1Prx1 - / - quadriceps muscle (ROC-AUC >0.97) with no differences detected in the heart (ROC-AUC <0.67), which does not undergo gene deletion in this model. Changes in AF spectra in mdx muscle were seen between the 3 week and 10 week time points (ROC-AUC = 0.96) and were not seen in the wild-type controls (ROC-AUC = 0.58).

Conclusion: These findings support the utility of in vivo fiber-optic AF and DR spectroscopy for the assessment of muscle tissue. This report highlights that there is considerable scope to develop this marker-free optical technology for preclinical muscle research and for diagnostic assessment of clinical myopathies and dystrophies.

Keywords: autofluorescence spectroscope; diffuse reflectance spectroscopy; in vivo fiber-optics; in vivo spectroscopy; muscle; muscular dystrophy; myopathy.

Fig. 1

(a) Schematic of the fiber-optic AF/DR system. 405-nm laser (violet) and halogen lamp (yellow) light sources were coupled into proximal ends of two multimode (MM) optical fibers via servo-mechanical shutters and a bandpass (BP) filter to further narrow the AF excitation wavelength range. Tissue DR and AF light was collected via a central fiber at the probe’s distal end (red) and passed through a 425-nm long-pass (LP) filter before detection at the spectrometer. Additional fibers within the fiber-optic probe (white) were unused. (b) The fiber-optic probe head was positioned directly onto the muscle surface during optical measurements, here showing a DR measurement on the murine gastrocnemius.

Fig. 2

(a)–(d) AF and (e)–(h) DR spectroscopy enabled classification of muscle types (Quad, Gastr, TA, and heart) in healthy B6 mice using PCA. (a), (e) Average normalized spectra (solid lines) show spectral differences between quadriceps, gastrocnemius, tibialis anterior, and cardiac muscle (95% CIs shown by dotted lines and shaded areas). (b), (f) Coefficient weightings of each principal component describe sources of variance across all wavelengths, including each PC’s total percentage explained variance (see legend). (c), (g) PC1 and PC2 scores show clustering of muscle types, which enabled their classification using decision boundaries (shaded regions) designed by a QDC. (d), (h) Classification ROC curves provide a quantitative measure of muscle type separability, where an AUC of 1 describes a perfect classifier and 0.5 a weakest classifier (see legend for muscle-specific ROC-AUCs).

Fig. 3

(a)–(d) AF and (e)–(h) DR spectroscopy of quadriceps muscle enables classification of NF1 mice from B6 controls (both 10 weeks) using PCA scores; however, when applied to cardiac muscle, the same analysis yields no separability. Normalized AF spectra of (a) quadriceps and (c) cardiac muscle in B6 and NF1 mice with [(b), (d)] corresponding PC score and QDC decision boundary plots. Normalized DR spectra of (e) quadriceps and (g) cardiac muscle in B6 and NF1 mice with [(f), (h)] corresponding PC score and QDC decision boundary plots.

Fig. 4

Histopathology of quadriceps muscle from 10-week-old NF1 mice differs from that observed in B6 mice. H&E staining shows endomysial fibrosis, and Oil Red O staining illustrates the intramyocellular lipid seen in NF1 mice. Scale bars indicate $200\mu \mathrm{m}$.

Fig. 5

(a)–(c) AF and (d)–(f) DR spectroscopy on quadriceps muscle enabled classification of mdx mice from B10 controls using PCA at both 3 week and 10 week time points. Normalized AF and DR spectra (top row) indicate differences in optical properties between these groups. These differences are highlighted by clustering in PCA space (middle row) as well as in their classification ROC curves (bottom row).

Fig. 6

Representative histopathology of quadriceps muscle from 10-week-old mdx mice revealed several dystrophic features, including centralized nuclei and muscular atrophy by H&E not evident in 3-week-old mdx mice and aged B10 controls. Scale bars indicate $200\mu \mathrm{m}$.