Characterizing reduced scattering coefficient of normal human skin across different anatomic locations and Fitzpatrick skin types using spatial frequency domain imaging

Abstract. Significance: Spatial frequency domain imaging (SFDI), a noncontact wide-field imaging technique using patterned illumination with multiple wavelengths, has been used to quantitatively measure structural and functional parameters of in vivo tissue. Using SFDI in a porcine model, we previously found that scattering changes in skin could potentially be used to noninvasively assess burn severity and monitor wound healing. Translating these findings to human subjects necessitates a better understanding of the variation in “baseline” human skin scattering properties across skin types and anatomical locations. Aim: Using SFDI, we aim to characterize the variation in the reduced scattering coefficient (μs′) for skin across a range of pigmentation and anatomic sites (including common burn locations) for normal human subjects. These measurements are expected to characterize baseline human skin properties to inform our use of SFDI for clinical burn severity and wound healing assessments. Approach: SFDI was used to measure μs′ in the visible- and near-infrared regime (471 to 851 nm) in 15 subjects at 10 anatomical locations. Subjects varied in age, gender, and Fitzpatrick skin type. Results: For all anatomical locations, the coefficient of variation in measured μs′ decreased with increasing wavelength. High intersubject variation in μs′ at visible wavelengths coincided with large values of the melanin extinction coefficient at those wavelengths. At 851 nm, where intersubject variation in μs′ was smallest for all anatomical locations and absorption from melanin is minimal, significant intrasubject differences in μs′ were observed at the different anatomical locations. Conclusions: Our study is the first report of wide-field mapping of human skin scattering properties across multiple skin types and anatomical locations using SFDI. Measured μs′ values varied notably between skin types at wavelengths where absorption from melanin was prominent. Additionally, μs′ varied considerably across different anatomical locations at 851 nm, where the confounding effects from melanin absorption are minimized.


Introduction
Diffuse optical spectroscopic (DOS) techniques have been widely used to obtain in vivo tissue optical properties. 1,24][5] Specifically, DOS techniques quantify the wavelength-dependent tissue reduced scattering (μ 0 s ) and absorption (μ a ) coefficients, which can be used to deduce subsurface structural and functional information.0][11] Although these results showed promise for a potential new approach to rapidly assess burn severity and prognosticate wound healing, translating this technique to human subjects necessitates an understanding of baseline μ 0 s values in human skin.Thus, it is important to document μ 0 s values of normal skin at commonly used DOS wavelengths (visible-and nearinfrared) for various anatomical locations and levels of pigmentation.
Prior to DOS, many studies have contributed to documenting human skin optical properties through in vitro and ex vivo measurements using integrating spheres.In 2011, Bashkatov et al. 12 thoroughly catalogued many of these contributions in their review work of in vitro, ex vivo, and in vivo optical properties of human skin, adipose, and muscle.These studies offered valuable insights toward complete characterization of human skin optical properties, for both whole skin and separated epidermis, dermis, and adipose layers.However, the reported values from these ex vivo measurements are not necessarily representative of in vivo tissues. 13,14Previous in vivo studies to quantify μ 0 s of healthy skin in different anatomical locations have employed several different DOS techniques.Doornbos et al. 15 utilized a fiber-based spatially resolved diffuse reflectance spectroscopy system to obtain in vivo optical properties of human skin and the underlying tissue.Tseng et al. 16,17 applied steady-state frequency domain photon migration to perform highly localized measurements of μ a and μ 0 s of in vivo volar forearm, palm, dorsal forearm, and upper inner arm for human subjects across a range of Fitzpatrick skin types.In 2015, Saager et al. 18 compared multiphoton microscopy and spatial frequency domain spectroscopy for measurement of melanin and reduced scattering on dorsal forearm and volar upper arm regions of 12 subjects of various skin types.In a study on volar forearm of 1765 Caucasian subjects (i.e., skin types I and II), Jonasson et al. 19 obtained scattering parameters over a range from 475 to 850 nm using a commercial diffuse reflectance spectroscopic system.Kono et al. used reflection spatial profile measurement to measure optical properties at 450 to 800 nm and 950 to 1600 nm for 198 subjects on the inner forearm, cheek, and dorsal hand between thumb and forefinger. 20owever, these studies had two limitations: (1) they only covered a small range of anatomical locations for measurements of scattering properties and (2) the measurement systems were restricted to point-based or single-line measurements that required multiple measurements to characterize the heterogeneity of large regions on the body.In summary, clinical translation of DOS requires a broader characterization of in vivo human skin that spans multiple anatomical locations and pigmentation levels, while also accounting for the heterogeneous nature of each sampled region.
In this study, we employ spatial frequency domain imaging (SFDI) to characterize and document μ 0 s of normal skin for 15 subjects with various pigmentation levels (Fitzpatrick types I to VI 21 ) at 10 anatomical locations.SFDI is a noncontact, wide-field DOS imaging technique that uses spatially modulated illumination in combination with models of light-tissue interaction to determine optical properties of in vivo tissue. 5,22,23Dognitz and Wagnieres 22 first developed and used a variation of SFDI to obtain in vivo skin optical properties at 400, 500, and 700 nm.Cuccia et al. 5,23,24 further developed the technique to expand the imaging spectrum to include near infrared wavelengths and enabled clinical translation of SFDI to skin ulcer imaging.Here, we document μ 0 s values across all subjects and anatomical locations at imaging wavelengths, ranging from visible to near-infrared.These measurements were derived from the semi-infinite homogeneous model described previously. 5We then compare μ 0 s values between subjects at each wavelength and identify 851 nm as the wavelength with the least variation in μ 0 s between subjects.We posit that this result is due to melanin being highly absorbing at visible wavelengths and localized in a thin layer at the base of the epidermis, which leads to a confounding effect in separating μ a and μ 0 s in the visible spectrum for subjects with darker skin.The decreasing intersubject μ 0 s variation with increasing wavelength suggests that pigmentation effects on μ 0 s determined by SFDI are the least at longer wavelengths (i.e., near-infrared and beyond).Finally, we show that baseline μ 0 s values vary with anatomical location, using 851 nm (where absorption from melanin is the lowest) as the wavelength for this analysis.This study serves as the first report for categorization of normal human skin scattering properties across multiple skin types and anatomical locations using SFDI.These findings are important for establishing the natural variation in baseline μ 0 s that must be accounted for when DOS techniques are translated to a clinical setting for applications such as burn and wound healing triage.

Spatial Frequency Domain Imaging
The OxImager RS™ (Modulim, Inc., Irvine, California) was used for SFDI measurements. 23his device measures calibrated diffuse reflectance over a 20 × 15 cm 2 field of view with a resolution of ∼1.5 mm.The system employs LEDs at eight center wavelengths (471, 526, 591, 621, 659, 731, and 851 nm) at maximum power of 0.5 mW∕cm 2 , and projects structured patterns at five evenly spaced spatial frequencies between 0 and 0.2 mm −1 , as described previously. 25The exposure time varies based on wavelength and the pigmentation of the imaged surface, but are typically between 5 and 60 ms.To mitigate motion artifacts related to respiration, each region was imaged three consecutive times, and repetitions with notable motion artifacts were disregarded.A single acquisition, which includes a single image taken of each of the 8 wavelengths at all 5 spatial frequencies, takes ∼30 s to complete.Using the software that accompanies the instrument (Modulim Inc.), data processing of three repetitions for all 10 anatomical locations on a single patient takes ∼10 min.All further analysis was performed using a single repetition typical of each region.A polydimethylsiloxane-based tissue-simulating reference phantom with known optical properties was measured at each imaging time point under the same lighting conditions as that of the subjects.Raw reflectance images from the subjects were calibrated against the images of the reference phantom and processed using the MI-Analyze software suite (Modulim, Inc., Irvine, California) to obtain μ 0 s and μ a at each wavelength.This calculation assumed a semi-infinite medium with homogeneous optical properties throughout the imaged tissue volume and used a Monte Carlo-based transport forward model. 5The model generated a 768 × 768 element look-up-table spanning an absorption coefficient range of 0 ≤ μ a ≤ 3.0 and a scattering range of 0.01 ≤ μ 0 s ≤ 4.0, with anisotropy and refractive index values fixed at 0.8 and 1.4, respectively.

Subjects
Subjects (N ¼ 15; 8 male and 7 female) were recruited and imaged under Institutional Review Board (IRB) protocol (IRB# 2011-8370).Subjects had skin types ranging from I to VI on the Fitzpatrick scale and no known dermatological complications.The majority of the subjects were young adults.Twelve subjects were in the age range of 18 to 35 years, whereas three were in the range of 36 to 55 years.During recruitment, we took care to ensure that subjects were distributed as evenly as possible across a wide range of skin types.However, we did not perform any a priori analysis to predefine the exact number of patients of each skin type to recruit.Measurements were obtained at 10 anatomical locations on each subject.Locations included cheek, ventral forearm, dorsal forearm, shin, palm, lower back, and chest (near collar bone), which are common areas for burn injuries.Measurements were also taken of the forehead, upper arm (bicep), and posterior neck (near the hairline).For regions not located on the midline, such as cheek, arm, and shin, we chose to image the subject's dominant side.Fitzpatrick skin types were determined using subject surveys (Table S4 in the Supplemental Materials) and clinically verified by Dr. Sharif.In order to supplement the low-quality webcam images that are captured by the commercial SFDI device, color images were taken prior to each measurement, using a digital camera (NEX-3, Sony Corporation of America, New York, New York).Instrumentation and measured anatomical locations are shown in Fig. 1.

SFDI Data Aggregation
Further data analysis was performed in MATLAB ® (R2018a, MathWorks, Natick, Massachusetts).A 40 × 40 pixel (∼1 cm 2 ) region of interest (ROI) was chosen from each anatomic location to avoid regions susceptible to artifacts from abrupt changes in curvature (e.g., wrinkles).For each location on each subject, ROIs of the same size were obtained from similar positions relative to the subject's individual anatomy.The measured μ 0 s values within the ROI were then used to perform statistical comparisons.

Statistical Analysis
At each anatomical location and wavelength, an intrasubject average value of μ 0 s was obtained from the 1600 (40 × 40) sampled pixels.Then intersubject means and standard deviations in μ 0 s values over all 15 subjects were calculated from the intrasubject averages (Table 1 and Table S2   in the Supplemental Materials).These values are then used to calculate an intersubject coefficient of variation in μ 0 s (standard deviation divided by mean) and are shown in Table S3 in the Supplemental Materials.We chose this statistical analysis method to best demonstrate the intersubject variation (i.e., the spread) in μ 0 s .Furthermore, μ 0 s values measured at 851 nm were compared between locations using a oneway repeated measures of analysis of variance (ANOVA) (Table 2).A post hoc Tukey's honest significant difference test was used to further compare differences between paired anatomical locations (Table 3).A p value <0.05 was considered statistically significant for this study.Box and whisker plots for the mean reduced scattering coefficient for each anatomical location are presented as Fig. 3 for all 15 subjects.The bottom and top of the boxes show the first and third quartiles, the bar inside the box marks the second quartile (the median), + indicates the mean, and the ends of the whiskers represent the minimum and maximum values.

SFDI Measurements of Reduced Scattering Coefficients at Each Anatomical Location for Visible to Near-Infrared Wavelengths
To illustrate the variation in μ 0 s with wavelength and skin type, Figs.The intersubject coefficients of variation for μ 0 s across all 15 subjects were calculated for each anatomical location at each wavelength (Table S3 in the Supplemental Materials).These coefficients decreased with increasing wavelength for all 10 anatomical locations (0.554 to 0.682 mm −1 at 471 nm; 0.0789 to 0.111 mm −1 at 851 nm).These values showed the least intersubject variation at 851 nm.The decrease in intersubject coefficient of variation of μ 0 s with increasing wavelengths coincides with the monotonically decreasing eumelanin extinction coefficient. 26,27This result suggests that variation in μ 0 s at shorter wavelengths is largely due to the inability of the semi-infinite homogeneous light transport model to adequately extract optical properties in subjects with darker skin types.In Table 1, we show the intersubject μ 0 s means and standard deviations at 851 nm, the measured wavelength that we believe is the least confounded by pigmentation.
It should be noted that the palm also showed the decreasing trend in coefficient of variation for μ 0 s with wavelength, but the decrease was less pronounced (0.168 to 0.0596 mm −1 over the same range of wavelengths; Table S3 in the Supplemental Materials).This is most likely due to the palm possessing the lowest melanin concentration in comparison to other anatomical locations. 28,29Thus the palm μ 0 s values are least confounded by pigmentation.Intersubject means and standard deviations of μ 0 s and μ a at each anatomical location are documented in Tables S2 and S4 in the Supplemental Materials.Figure S1 in the Supplemental Materials shows μ a values for each patient measured from the dorsal forearm.

Interanatomical Location Variations of μ 0
s Among all Subjects at 851 nm Based on the low intersubject variation in μ 0 s reported above, we chose 851 nm as the wavelength of interest for documenting skin scattering properties at the 10 anatomical locations for all 15 subjects.Table 2 shows the results of the one-way ANOVA test of comparing all 10 locations.Table 3 shows p values obtained from Tukey's tests between μ 0 s values at 851 nm for each pair of anatomical locations.Representative μ 0 s maps with ROIs are shown in Fig. 3, and average μ 0 s values at 851 nm for all subjects are shown in Fig. 4.

Discussion
In this study, we used SFDI to obtain reduced scattering coefficient values (μ 0 s ) of normal skin at 10 anatomical locations of 15 subjects ranging from Fitzpatrick skin types I to VI (Table 1).Our data showed lower μ 0 s values for subjects with higher Fitzpatrick skin type, which agrees with existing literature.Specifically, both Saager et al. 18 and Jonasson et al. 19 noted a decreasing μ 0 s for subjects with higher melanin fraction compared to μ 0 s measured in subjects having low melanin fraction.Tseng et al. 16,17 attributed this decrease in μ 0 s to limited probing depth due to increase in absorption by melanin, leading to fewer photons reaching the collagen and elastin matrix in the dermis, which can contribute strongly to μ 0 s values in the near-infrared wavelength range.Although we use a physical tissue-simulating phantom (with optical properties measured using an integrating sphere) as a calibration to obtain the diffuse reflectance, Jonasson et al. normalized an Inverse Monte Carlo modeled spectrum with the average measured spectral intensity to avoid the need for absolute calibration of the intensity recorded by their spectrometers.Furthermore, we also imaged skin for all Fitzpatrick skin types.This will contribute to greater variation in our data relative to the Swedish cohort. 19inally, it should be noted that the μ 0 s spectra for Fitzpatrick skin types I and II appear typical for biological tissue and can be described by Rayleigh and Mie scattering. 19,31,33,34However, our measured μ 0 s from darker skin (Fitzpatrick skin types III to VI) tended to increase with wavelength for almost all anatomical locations except the palm.We attribute this observation to the high absorption of visible light by melanin found only in the epidermis.The localization of melanin to the epidermis contributes to a highly inhomogeneous depth-resolved absorption profile.Our simple semi-infinite homogeneous model cannot resolve such complex geometry. 35vidence of this limitation can be seen from an examination of the data obtained for the palm.The palm has a very low melanin fraction for all skin types.For palm skin, we observed a decrease in μ 0 s spectra with increasing wavelength [Fig.2(b)].We are currently investigating the effects of melanin confined to the epidermis on SFDI derived optical properties determined from a semi-infinite homogeneous model and developing new models to account for such effects.

Conclusion
In this study, we have for the first time documented the reduced scattering coefficient properties μ 0 s of normal skin at 10 anatomical locations, for subjects having pigmentation variations across all Fitzpatrick skin types (i.e., I to VI), using SFDI.Examining the measured μ 0 s at an anatomical location (i.e., dorsal forearm) at visible wavelengths showed a decreasing trend with higher Fitzpatrick skin type.However, there were no significant differences of this kind observed between any of the skin types seen at the longest wavelength measured (851 nm).Furthermore, significant differences in measured μ 0 s at 851 nm were observed between different anatomical locations.These findings regarding the reduced scattering properties across various anatomical locations of subjects of all Fitzpatrick skin types will aid in establishing baseline SFDI measurement for future clinical studies.

Disclosures
Dr. Durkin has a financial interest in Modulim Inc. (formerly known as Modulated Imaging, Inc.), which developed the SFDI device employed in the study.However, Dr. Durkin does not participate in the management of Modulated Imaging, and has not shared these results with that company.Conflicts of interest have been disclosed and managed in accordance with University of California and NIH policies.The other authors have no financial interests or commercial associations that might pose or create a conflict of interest with the information presented in this article.
2(a) and 2(b) show the μ 0 s spectra (averaged over an ROI) of the dorsal forearms and palms, respectively, for all 15 subjects, classified by Fitzpatrick skin type.Representative color images and μ 0 s maps of the dorsal forearm and palm of subjects with various Fitzpatrick skin types are shown in Figs.2(c) and 2(d), respectively.Overall, at shorter wavelengths, subjects with Fitzpatrick skin types indicative of lower pigmentation (I, II, and III) had higher measured μ 0 s than subjects with skin types corresponding to more pigmentation (IV and V/VI), likely due to confounding effects from absorption of light by melanin.

Fig. 2
Fig. 2 (a) μ 0 s distribution (471 to 851 nm) for the dorsal forearm across all wavelengths of all 15 subjects shown with their Fitzpatrick skin types.(b) μ 0 s distribution (471 to 851 nm) for the palm.(c) Representative color and μ 0 s image examples of the dorsal forearm on subjects of Fitzpatrick types I, III, and V/VI.Scale bar = 1 cm.(d) Representative color images and μ 0 s maps of the palm.Scale bar = 1 cm.

Fig. 3
Fig. 3 Examples of μ 0 s maps from each anatomic location at 851 nm for a subject of Fitzpatrick skin type I.The 1-cm2 ROIs used for analysis are shown in white squares.

Fig. 4
Fig. 4 Box and whisker graph of μ 0 s values measured at 851 nm from all 15 subjects for each anatomical location.

Table 1
Summary of μ 0 s values obtained at 10 anatomical locations for all 15 subjects at 851 nm.The mean values were reported along with the standard deviations [mean (standard deviation)].
Phan et al.: Characterizing reduced scattering coefficient of normal human skin. . .

Table 2
Results of a one-way ANOVA performed on μ 0 s values for each pair of anatomical locations in all 15 subjects at 851 nm.