2.1 Phantom measurement results
The absorption spectra of the bound water phantom near the second overtone and first overtone of pure water absorption peaks at 25 ℃ were determined using the inverse adding doubling method and are shown in Figs. 1(a) and 1(b), respectively, as dotted lines. Due to high water absorption strength and the sensitivity limitation of our spectrometer, data collected at wavelengths longer than 1,380 nm were noisy and could not be used for calculating the absorption coefficient. We used the pure water absorption spectrum to fit the bound water phantom absorption spectra, and the results are shown as solid lines in Fig. 1. It can be seen in Fig. 1 that in the wavelength ranges between 945 and 1,000 nm as well as between 1,230 to 1,360 nm, the bound water phantom absorption spectrum has a discernable difference from the pure water absorption spectrum. Herrera-Gómez et al. showed that the FWHM (full width at half maximum) of water absorption band at 6,060 nm increased with slightly decreased peak wavelength as the amount of bound water increased.23 Likewise, Eneh et al. showed that the water absorption spectrum at the wavelength of 3,900 nm was blue shifted and broadened as bound water fraction increased.22 Chung et al. reported that at 975 nm the bound water absorption spectrum was red shifted as compared to the pure water absorption spectrum.24 Our measurement result at the second water absorption overtone region, depicted in Fig. 1(a), has a red shifting effect similar to that described in the Fig. 1(c) of ref.24. On the other hand, it can be seen in Fig. 1(b) that the bound water phantom absorption spectrum between 1,360 to 1,380 nm is blue shifted from the pure water absorption spectrum. This could be caused by the first overtone absorption peak broadening. It is worth noting that we fabricated another Lipofundin phantom which had no gelatin powder added, and its absorption spectrum was almost identical to the pure water absorption spectrum (data not shown). Our phantom study results elucidate that the bound water effect induces absorption spectrum deviation from the pure water absorption spectrum both at the first and second overtones.
2.2. Human skin measurement results
2.2.1 Typical psoriasis absorption spectra
To understand the characteristics of psoriasis absorption spectra, a representative subject with typical psoriatic lesions was selected from the 21 subjects recruited for the analyses in this subsection. The typical absorption spectra of the psoriatic lesion site and its adjacent uninvolved skin site of a subject near 970 nm are displayed in Figs. 2(a) and 2(b), respectively. It can be seen in Fig. 2 that the magnitude of absorption of psoriatic lesion is slightly larger than that of adjacent uninvolved skin. This phenomenon was universal for all psoriasis subjects recruited in this study and we believe this was caused by the higher skin blood content of psoriatic lesion than the uninvolved skin. The best pure water spectrum fitting to the absorption spectra are displayed as solid lines in the figures. It can be observed in Fig. 2 that the absorption spectra of the lesion site as well as the adjacent uninvolved site deviate from the pure water absorption spectrum. This phenomenon is similar to the bound water phantom measurement results displayed in Fig. 1(a). The pure water fitting residuals of the lesion and uninvolved skin absorption spectra between 940 and 1,000 nm were 0.55 and 0.47, respectively. In this case, the fitting residual of lesion site is higher than that of uninvolved skin.
The typical psoriatic lesion and adjacent uninvolved skin absorption spectra near 1,300 nm and their best pure water fitting spectra are depicted in Fig. 3. In general, at this first water absorption overtone band, the absorption spectra of both skin sites demonstrate water absorption peak broadening, resembling the bound water phantom measurement results shown in Fig. 1(b). The pure water fitting residuals of the psoriasis and normal skin absorption spectra between 1,230 and 1,380 nm were 7.61 and 2.77, respectively. It is apparent that the fitting residual values at the first overtone band are much larger than those at the second overtone band. Besides, qualitative observation of the spectra difference between psoriasis and normal skin (Fig. 2(a) to Fig. 2(b) vs. Fig. 3(a) to Fig. 3(b)) reveals that the difference is more prominent at the first overtone band. In addition, we noted that the skin absorption spectra of normal subjects (not depicted here) were comparable to those of uninvolved skin as displayed in Figs. 2(b) and 3(b).
2.2.2 Statistical analyses of skin absorption spectra
We carried out one-way ANOVA to investigate the differences between the pure water fitting residuals of psoriatic lesion sites, the adjacent uninvolved sites and uninvolved upper inner arms of the 21 psoriasis subjects, and normal upper inner arms of the 21 normal subjects. Box-and-whisker plots displayed in Fig. 4 summarized the statistics results of one-way ANOVA. We first noted that, in Fig. 4(a), the distribution of fitting residuals of lesion sites, adjacent uninvolved sites, uninvolved upper inner arms are comparable. The P-value of one-way ANOVA of the four groups was 0.21, indicating that the population means were not significantly different. It can be deduced that although at second overtone band the bound water effect can be observed in the spectrum analysis such as those shown in Fig. 2, the fitting residuals at this band could not be used as an effective parameter for distinguishing the lesion site from the uninvolved skin of a psoriasis subject nor from the normal skin. On the other hand, it can be seen in Fig. 4(b) that the distribution of pure water fitting residuals of lesion sites is significantly higher than those of the other three sites. The P-value of one-way ANOVA of the four groups was 9.65E-15, indicating that the population means were significantly different. By carrying out Scheffѐ tests, we found that fitting residuals of psoriatic lesion at the first overtone band was significantly higher than the other three sites, and the uninvolved skin adjacent to the lesion site had significantly higher fitting residuals than those of upper inner arm skin of normal subjects. At this first overtone band, the fitting residuals of the upper inner arm skin of psoriasis and normal subjects were not significantly different. Our results suggest that pure water fitting residual near the first water absorption overtone could be an effective indicator of psoriasis.
2.2.3 Reduced scattering
The reduced scattering coefficients of the typical psoriatic lesion and its adjacent uninvolved site of the representative subject at the second overtone of water absorption are illustrated in Figs 5(a) and 5(b), respectively. The reduced scattering coefficients of uninvolved site are generally higher than the lesion counterpart. This observation agrees with what we have reported earlier where the analyzed wavelength range was from 500 to 900 nm. 15 The scattering power law (μs' = a*λ-b) was employed to smooth the reduced scattering spectra in the 940-1,000 nm range, and the fitting results are shown as solid lines in Fig. 5. The magnitude “a” and wavelength exponent “b” of the scattering power law fitting were 315.565 and 0.808, respectively, for the typical lesion shown in Fig. 5(a), and were 330.316 and 0.777, respectively, for the adjacent uninvolved site shown in Fig. 5(b).
Consistent with the results observed in the second overtone band, the reduced scattering coefficients of psoriatic lesion were lower than those of uninvolved site at the first overtone band, as can be seen in Fig 6. Due to strong water absorption near the first overtone band, the scattering coefficients at 1,370 nm drop drastically. The magnitude “a” and wavelength exponent “b” of the scattering power law fitting in the 1,230–1,380 nm range were 257,912.881 and 1.762, respectively, for the typical lesion site shown in Fig. 6(a), and were 484.744 and 0.823, respectively, for the adjacent uninvolved site shown in Fig. 6(b).
2.2.4 Statistical analyses of skin reduced scattering spectra
One-way ANOVA was performed to investigate the reduced scattering coefficients differences between various skin sites of the 21 psoriasis subjects and 21 normal subjects. We only selected the reduced scattering coefficients at 970 nm and 1,300 nm for the analyses. The analysis results are summarized in the box-and-whisker plots illustrated in Fig. 7. One-way ANOVA indicated that the 970 nm reduced scattering coefficients of the four measurement sites were significantly different (P = 6.61E-7). Employing further Scheffѐ tests revealed that the psoriatic lesions had significantly lower 970 nm reduced scattering coefficients than those of adjacent uninvolved skin, uninvolved upper inner arm skin, and normal upper inner arm skin. The complete Scheffѐ test results are listed in the inset of Fig. 7(a). On the other hand, at 1,300 nm wavelength, the reduced scattering coefficients of the uninvolved arm group had highest mean as shown in Fig. 7(b). This data distribution is qualitatively similar to that of 970 nm reduced scattering coefficients shown in Fig. 7(a). One-way ANOVA indicated that the 1300 nm reduced scattering coefficients of the four measurement sites were significantly different (P = 3.76E-5). Scheffѐ test results indicate that at 1,300 nm the reduced scattering coefficients of psoriatic lesions were not significantly lower than those of normal skin (P = 0.19), which is different from the results shown in Fig. 7(a). At 1,300 nm wavelength, the reduced scattering coefficients of psoriatic lesion were significantly lower than those of adjacent uninvolved skin and uninvolved upper inner arm skin (P = 0.01 and P = 7.38E-5). Our results suggest that while the light scattering property of skin at either the first or the second overtone of water absorption band could be used for psoriasis diagnosis, using second overtone band would be more effective than the first overtone band. This is reasonable since the skin reduced scattering coefficient decreases as wavelength increases, and the difference between the light scattering property of psoriatic lesion and normal skin would be more prominent at the shorter wavelengths.
We further analyzed the values of the magnitude “a” and the exponent “b” of the scattering power law that fit to the skin reduced scattering spectra. Surprisingly, we found one-way ANOVA results that neither “a” nor “b” values derived from 940–1,000 nm or 1,230–1,380 nm showed significant differences between the measurement sites. We observed that the span of “a” and “b” values of the 21 psoriatic lesion sites were quite large as compared with those of the other sites. The “a” and “b” values in the scattering power law were linked to the scatterers’ number density and average size, respectively.25 In addition, it has been indicated that the light scattering of biological tissues is majorly induced by the organelles and membranes of cells which provide index of refraction perturbation.25 Our finding suggests that the size of fine light-scattering structures in the epidermis of psoriasis skin have great variation as compared with those of the adjacent uninvolved or normal skin.