1.

Introduction

Diagnostic optical modalities detect and analyze backscattered photons from illuminated tissue. The common principle for these techniques is that the detected light contains information on tissue scattering and absorption from the interrogated tissue volume. To untangle this information, one needs to understand how the optical properties are related to photon propagation in tissue. Therapeutic modalities on the other hand, depositing light energy in tissue, are also influenced by the means for light delivery and tissue light transport governed by the optical properties.

Photon propagation in tissue is commonly modeled using either the diffusion approximation or Monte Carlo simulations. Although the former can be calculated analytically, the latter is considered most accurate.¹ Both approaches predict remitted light intensity given that the absorbing and scattering properties are known. In skin tissue, melanin and hemoglobin are the most dominant absorbing chromophores in the visible wavelength range. The main sources of tissue scattering are filamentous proteins such as keratins in the epidermis and collagen in the dermis,² or intracellular structures such as nuclei and mitochondria.³ The size of the scattering particles also has a direct impact on the scattering angle, affecting both the scattering phase function for a given wavelength and how the reduced scattering coefficient, ${\mu }_{\mathrm{s}}^{\prime }$, varies with wavelength.⁴ In some cases, the scattering properties themselves have been shown to contain physiological information of high importance, for example, in cancer⁵ and when assessing burn wounds.⁶ It has also been suggested that the fraction of large and small scattering particles in skin tissue can have a diagnostic value.³ Hence, there is a need to further determine how Mie and Rayleigh scattering varies in a larger cohort.

Previous characterizations of tissue scattering applied inverse photon propagation algorithms to sets of measured light intensities to estimate the optical properties of tissue (i.e., absorption and reduced scattering coefficients). Typically based on either the diffusion approximation or Monte Carlo simulations, these algorithms vary in complexity and may include: multiple layers with different ${\mu }_{\mathrm{s}}^{\prime }$,^7,8 absorbing chromophores with wavelength constraints,^9,10 and different degrees of Mie and Rayleigh scattering to properly account for the different scattering particle sizes found in tissue.¹⁰ It has been shown that the accuracy of an inverse algorithm depends on how well the photon propagation model can mimic the detected light intensity.^11,12

Different modalities for illumination and detection of light are commonly used when assessing optical properties. For in vitro studies, tissue samples may be analyzed using single- and double-integrating sphere setups.¹³ In vivo setups may have a temporally¹⁴ or spatially¹⁵ modulated light source, or utilize multiple detection points in combination with a steady-state single-point light source.^16,17 In both in vivo and in vitro studies, the number of analyzed subjects or samples has been limited.¹⁸ Furthermore, in vitro samples may have a higher ${\mu }_{\mathrm{s}}^{\prime }$, partly due to sample preparation.⁴ Hence, to develop diagnostic and therapeutic optical modalities based on light transport modeling, there is a need for a more comprehensive dataset on ${\mu }_{\mathrm{s}}^{\prime }$.

The aim of this study was to assess light scattering in the forearm skin of the subjects in the Linköping cohort of the Swedish CArdioPulmonary bioImage Study (SCAPIS). This is, to our knowledge, the first time tissue optical scattering has been estimated in such a large study. The magnitude and variation in a normal population can serve as input parameters when designing new inverse algorithms or when predicting the effect of therapeutic optical techniques.

Light scattering data were determined as the reduced scattering coefficient and the fraction of Mie and Rayleigh scattering. The scattering properties were assessed using an instrument integrating spatially resolved diffuse reflectance spectroscopy and laser Doppler flowmetry. Data analysis included an inverse Monte Carlo algorithm that in addition to scattering accounts for geometrical and absorbing properties of the tissue.

2.

Methods

3.

Results

Twenty-seven subjects were excluded due to data acquisition failure and four subjects decided to end the measurement before the end of the protocol. The age of the remaining 1734 study subjects was $57.7\pm 4.4$ years (mean ± SD).

The average reduced scattering coefficient for baseline values in the wavelength range 475 to 850 nm, and their normal (±SD) variation are shown in Fig. 1. The mean ${\mu }_{\mathrm{s}}^{\prime }$ ranged from 3.16 (SD = 0.72) ${\mathrm{mm}}^{-1}$ at 475 nm to 1.13 (0.27) ${\mathrm{mm}}^{-1}$ at 850 nm (Table 1). The corresponding 95% confidence intervals for the reduced scattering coefficient are [1.71 to 4.60] ${\mathrm{mm}}^{-1}$ and [0.60 to 1.67] ${\mathrm{mm}}^{-1}$, respectively (Table 1).

Fig. 1

Average reduced scattering coefficient during baseline (middle line) with the shaded area marking $\pm 1$ SD.

Table 1

The reduced scattering coefficient in baseline as mean [mm−1], SD, and 95% confidence interval.

Mean baseline, occlusion, and reperfusion values for the parameters describing the reduced scattering coefficient [Eq. (1)] are shown in Table 2. There was a significant difference in the parameters ${\alpha }^{\prime }$ and $\gamma$ between the occlusion and reperfusion phases compared with baseline ($p<0.001$).

Table 2

Parameters describing the reduced scattering coefficient during the three phases.

The baseline histograms for ${\alpha }^{\prime }$, $\beta$, and $\gamma$ are shown in Figs. 2(a)–2(c). It can be observed that the distribution for $\beta$ is right skewed, whereas that for $\gamma$ is left skewed.

Fig. 2

Histograms for the parameters (a) ${\alpha }^{\prime }$, (b) $\beta$, and (c) $\gamma$. Time-averaged parameter value during 5-min baseline.

All the following analyses were performed only for baseline values. Age was correlated to ${\alpha }^{\prime }$ ($p<0.001$, $r=-0.12$). The total amount of melanin (melanin concentration times epidermis thickness) was on average $0.48\%\pm 0.36\%$ mm and correlated to $\alpha$ ($p<0.001$, $r=-0.32$), $\beta$ ($p<0.001$, $r=0.57$), and $\gamma$ ($p=0.004$, $r=-0.069$).

There were significant differences in all scattering parameters between men and women ($p<0.001$), Table 3. No significant difference in the amount of melanin among men and women was observed.

Table 3

Differences in scattering parameters between men and women.

4.

Discussion

In this study, we present in vivo scattering properties of human skin in a large Swedish cohort. The results, describing the magnitude and variation in the reduced scattering coefficient in a normal population, can be used as an input when developing inverse algorithms, when predicting the effect of therapeutic optical techniques, or when evaluating traumatzed skin including burn wound assessment.⁶ In addition, we have also assessed the influences of gender and age on the parameters describing the reduced scattering coefficient in tissue and the correlation between the reduced scattering coefficient and the amount of melanin.

The EPOS system, along with its skin model and inverse Monte Carlo method, has been designed to measure oxygen saturation, RBC tissue fraction, and speed-resolved perfusion. Previous studies using the EPOS system have only reported these specific output parameters.^19,21,23,24 This article, however, is focused on extracting and reporting scattering properties using this system. A previous validation has shown that the accuracy of the EPOS method in estimating the reduced scattering coefficient was within 15% using two-layered silicon phantoms including ${\mathrm{TiO}}_2$ as a scattering agent.²⁵ We have reasons to believe that the error is smaller than this as the optical phantoms in the evaluation contained a hemoglobin mimicking absorber with much less prominent absorption peaks than that of real hemoglobin. This makes the data fitting of the inverse Monte Carlo method more difficult. Furthermore, it should be noted that the values of reduced scattering in the validation study (range 1.25 to $3{\mathrm{mm}}^{-1}$) resembled those observed in the present study.

Previously reported values on ${\mu }_{\mathrm{s}}^{\prime }$ originate from both in vitro and in vivo studies. Bashkatov et al.¹³ examined 23 in vitro postmortem skin samples in a double-integrating sphere system. They observed a ${\mu }_{\mathrm{s}}^{\prime }$ ranging from 4.1 to $1.9{\mathrm{mm}}^{-1}$ in the 475- to 850-nm wavelength range. Salomatina et al.⁵ analyzed eight dermis samples in vitro. Their ${\mu }_{\mathrm{s}}^{\prime }$ ranged from 4.8 to $2.1{\mathrm{mm}}^{-1}$. Jacques⁴ compiled data from eight studies (both in vitro and in vivo) including Bashkatov¹³ and found ${\mu }_{\mathrm{s}}^{\prime }$ in the range 5.4 to $2.2{\mathrm{mm}}^{-1}$. Bosschaart et al.¹⁷ examined 60 neonates’ dorsal hand skin using in vivo spatially resolved diffuse reflectance and found ${\mu }_{\mathrm{s}}^{\prime }$ ranging approximately from 2.7 to $1.3{\mathrm{mm}}^{-1}$ in 450 to 600 nm. Our range of 3.16 to $1.13{\mathrm{mm}}^{-1}$ for the reduced scattering coefficient in the 475- to 850-nm wavelength range is in line with these studies. The difference between studies is larger regarding the individual parameters (${\alpha }^{\prime }$, $\beta$, and $\gamma$) describing the reduced scattering coefficient.⁴

There was a small, but significant, deviation from baseline values in the parameter ${\alpha }^{\prime }$ and $\gamma$ during the arterial occlusion and reperfusion phases. This could be an effect of the scattering model, where all layers are assumed to have equal scattering properties. If scattering varies as a function of depth in skin, a change in the sampling volume might change the scattering parameters during the provocation as the model cannot account for different scattering properties in different layers. This could result from a different amount of blood due to a redistribution of blood between the layers during the occlusion and reperfusion phases. More work is needed to determine whether the change is a model effect or relates to an actual change in scattering properties during a provocation. Even though the deviation from baseline is small, it is present for the majority of the individuals included in this study.

Melanin is recognized as an important absorbing component in skin, but the impact of melanin on tissue scattering has not been fully clarified. Sardar et al.²⁶ investigated the optical properties of melanin in vitro and found that the scattering coefficient of melanin was much higher than the absorption coefficient for wavelengths between 476 and 633 nm. On the other hand, Zonios and Dimou²⁷ found no difference in scattering properties when comparing normal skin and pigmented nevi, and concluded that melanin had negligible impact on skin scattering. The number of subjects in that study was; however, small ($n=3$) and the linear model for scattering did not account for Rayleigh scattering. We found a correlation between total amount of melanin, estimated from the model, and the scattering properties of skin, especially for the ${\alpha }^{\prime }$ and $\beta$ parameters, where ${\alpha }^{\prime }$ characterizes scattering density and $\beta$ is the size of the scattering particles. Our findings showed that the reduced scattering coefficient and the size of the scattering particles decreased with an increased amount of melanin, which is in line with previous findings by Tseng et al.²⁸ They related the reduction in ${\mu }_{\mathrm{s}}^{\prime }$ to a smaller sampling volume as more light gets absorbed as the amount of melanin increases, where fewer photons would have passed a smaller volume of scattering particles like collagen and elastin in the dermis. An increase in the $\beta$ indicates a shift toward smaller structures scattering within the sampled volume. Riesz et al.²⁹ found that the scattering coefficient of melanin in vitro applies to Rayleigh theory for wavelengths between 210 and 325 nm. In vivo melanin is located in melanosomes, with various sizes and distributions; however, the relationship between in vitro and in vivo scattering of melanin is not clear.

In this study, we found both a correlation between the reduced scattering coefficient and age, and significant differences in all scattering parameters between men and women, resulting in a lowered reduced scattering coefficient for women. Saidi et al.³⁰ found a positive correlation between ${\mu }_{\mathrm{s}}^{\prime }$ and age in neonates, where the reduced scattering coefficient increased as the density of collagen fibers increased during maturation. Bosschaart et al.¹⁷ found no correlation between age in neonates and ${\mu }_{\mathrm{s}}^{\prime }$. They related the lack of correlation to an age range outside the one used in their study as ${\mu }_{\mathrm{s}}^{\prime }$ is expected to increase at some point during maturation and be higher in adults.⁴ The negative correlation between ${\mu }_{\mathrm{s}}^{\prime }$ and age could originate from the decrease in collagen concentration in skin during aging.³¹ Women also have a lower collagen concentration compared with men,^31,32 which could cause the differences between men and women in all scattering parameters. Few previous studies have reported gender differences in the reduced scattering coefficient in human skin, probably due to small numbers of subjects. Calabro et al.³³ showed a lower reduced scattering coefficient in the skin of female mice, which was found to be related to a thinner dermal layer for females caused by less collagen.

5.

Conclusions

In this study, in vivo values for tissue scattering are presented from a cohort including 1734 subjects. This is the first time these type of data have been presented for such a large number of included subjects. Within the cohort, we found a negative correlation between the reduced scattering coefficient and age, and a significant difference between men and women in the reduced scattering coefficient as well as in the fraction of small scattering particles. Our presented mean values for reduced scattering are well in line with many previously reported values. In addition, we also report on the distribution of the ${\alpha }^{\prime }$, $\beta$, and $\gamma$ parameters, describing the reduced scattering coefficient, which may serve as reference values for future studies.