Impact of one-layer assumption on diffuse reflectance spectroscopy of skin

Abstract

Diffuse reflectance spectroscopy (DRS) can be used to noninvasively measure skin properties. To extract skin properties from DRS spectra, you need a model that relates the reflectance to the tissue properties. Most models are based on the assumption that skin is homogenous. In reality, skin is composed of multiple layers, and the homogeneity assumption can lead to errors. In this study, we analyze the errors caused by the homogeneity assumption. This is accomplished by creating realistic skin spectra using a computational model, then extracting properties from those spectra using a one-layer model. The extracted parameters are then compared to the parameters used to create the modeled spectra. We used a wavelength range of 400 to 750 nm and a source detector separation of 250 μm. Our results show that use of a one-layer skin model causes underestimation of hemoglobin concentration [Hb] and melanin concentration [mel]. Additionally, the magnitude of the error is dependent on epidermal thickness. The one-layer assumption also causes [Hb] and [mel] to be correlated. Oxygen saturation is overestimated when it is below 50% and underestimated when it is above 50%. We also found that the vessel radius factor used to account for pigment packaging is correlated with epidermal thickness.

Fig. 1

A representative fit showing the good agreement between the two-layer and one-layer spectra.

Fig. 2

Two-layer [mel] versus the one-layer extracted [mel]. This plot was created by varying the two-layer [mel] used to create the spectra and fixing all other parameters at three different values for $Z_0$ (50, 100, and $200\mu \mathrm{m}$). [Hb] was fixed at $1\mathrm{mg}/\mathrm{ml}$, ${\mu }_s^{\prime }$ was fixed at $20{\mathrm{cm}}^{-1}$, ${\mathrm{SO}}_2$ was fixed at 100%, and $B$ was fixed at $-1.5$. [mel] ranged from 0 to $5\mathrm{mg}/\mathrm{ml}$ in 20 increments.

Fig. 3

Two-layer [Hb] versus the one-layer extracted [Hb]. This plot was created by varying the two-layer [Hb] used to create the spectra and fixing all other parameters at three different values for $Z_0$ (50, 100, and $200\mu \mathrm{m}$). [Hb] was fixed at $1\mathrm{mg}/\mathrm{ml}$, ${\mu }_s^{\prime }$ was fixed at $20{\mathrm{cm}}^{-1}$, ${\mathrm{SO}}_2$ was fixed at 100%, and $B$ was fixed at $-1.5$. [mel] ranged from 0 to $3\mathrm{mg}/\mathrm{ml}$ in 20 increments.

Fig. 4

Two-layer ${\mathrm{SO}}_2$ versus the one-layer extracted ${\mathrm{SO}}_2$. This plot was created by varying the two-layer ${\mathrm{SO}}_2$ used to create the spectra and fixing all other parameters at three different values for $Z_0$ (50, 100, and $200\mu \mathrm{m}$). [mel] was fixed at $1\mathrm{mg}/\mathrm{ml}$, ${\mu }_s^{\prime }$ was fixed at $20{\mathrm{cm}}^{-1}$, [Hb] was fixed at $1\mathrm{mg}/\mathrm{ml}$, and $B$ was fixed at $-1.5$. ${\mathrm{SO}}_2$ ranged from 0 to 100% in 20 increments.

Fig. 5

$Z_0$ versus the vessel radius parameter used in the one-layer inverse model. This plot was created by varying $Z_0$ in the two-layer model used to create the spectra and fixing all other parameters. [mel] was fixed at $1\mathrm{mg}/\mathrm{ml}$, [Hb] was fixed at $1\mathrm{mg}/\mathrm{ml}$, ${\mu }_s^{\prime }$ was fixed at $20{\mathrm{cm}}^{-1}$, ${\mathrm{SO}}_2$ was fixed at 100%, and $B$ was fixed at $-1.5$. $Z_0$ ranged from 0 to $300\mu \mathrm{m}$ in 20 increments.

Fig. 6

(a) Random pairs of [Hb] and [mel] used to generate the two-layer spectra. (b) Extracted one-layer values for [Hb] and [mel].