Lookup-table-based inverse model for human skin reflectance spectroscopy: two-layered Monte Carlo simulations and experiments

Xiewei Zhong; Xiang Wen; Dan Zhu

doi:10.1364/OE.22.001852

1. Introduction

Diffuse reflectance spectroscopy in the visible and near-infrared (NIR) spectral ranges is an effective technique for non-invasive measurement of optically related molecular components and the content of chromophores in biological tissues. It has shown great potential for the investigation and characterization of skin physiology [1–3], for instance, discriminating between pathological and normal skin [4,5], and evaluating the performance of skin care products [6–8]. There are chromophores (melanin, hemoglobin, and water), and scatterers (collagenous fibrils, fibers, and various subcellular structures), in skin that show strong absorption or scattering of visible and NIR light, respectively [9–11]. The changes in their contents can directly affect the skin’s diffuse reflectance.

In order to extract the physiological properties of skin from the measured reflectance spectrum, some inverse models have been proposed, including the diffusion approximation [12]; the semi-empirical analytical solution [13,14]; and the experiment-based [15–17] and Monte Carlo (MC) simulation-based [18,19] look-up tables (LUTs). As a pilot study for diffuse reflectance analysis, the diffusion approximation has been employed to model the reflectance spectrum of colon polyps [12], as well as to evaluate the melanin and hemoglobin content of skin [20]. However, the diffusion approximation is only suitable for multi-scattering events, and thus is not appropriate for compact source and detection fibers with very short distances between them [21]. Combining the diffusion approximation with a tissue phantom experiment, a semi-empirical analytical solution was developed that was valid for short distances between the source and detection fibers [13,14]. But if the applicable range of optical properties were expanded, it was reported that the accuracy of the approach would need more parameters for calibration [22].

The experiment-based LUT becomes an alternative description for probe reflectance of different optical properties of turbid media [15,23], and was generated by measuring the reflectance of tissue phantoms with known optical properties. However, the accuracy of optical properties determined from this inverse model not only suffers from the finite number of tissue phantoms, but also from the difference between tissue phantoms and real human skin. In contrast, the Monte Carlo simulation – a gold standard of modeling that has been widely used to investigate complex systems and processes, validate approximate theoretical models, and evaluate new techniques [24,25] – can also be used for analyzing diffuse reflectance spectra [18]. Although the initial simulation is time-consuming, the established MC-based LUT can provide rapid spectrum analysis. For such MC-based LUT, it is necessary to consider the geometry of the fiber probe in the initial simulation, and the performance of the LUT should be tested by experiment.

Human skin is multilayered, with the epidermis, dermis, and subcutaneous tissue. For skin diagnosis purposes, reflectance probes are usually optimized for detecting both melanin in the epidermis and blood in the dermis. Zonios et al. proposed a two-layered tissue reflectance model, and demonstrated its validity for “no absorption” in the upper and lower layers [26,27]. However, there is melanin packed in the melanosomes of the epidermis, and hemoglobin in the blood vessels of the dermis, which result in strong absorption in the range of visible light. A semi-empirical model for skin analysis has been proposed [28] that is valid for a monochromatic, collimated, and normally incident beam. However, for practical reflectance fiber probes, the validity of the semi-empirical model needs to be further verified. Considering the dermal layer as two layers (with an upper dermal layer and a lower dermal layer), a three-layered model based on the different path length distributions was proposed, and simulated by Monte Carlo with different epidermis thicknesses, variable scattering, and no absorption [29]. Ultimately though, any change in the upper layer absorption will alter the path length of the lower layers.

In this work, in order to approximate the model to human skin, two-layered Monte Carlo (MC) simulations for a commercial reflectance probe (R400-7, Ocean Optics, USA), were performed for different absorptions in the epidermal and dermal layers. An inverse model based on the 3D-LUT from MC simulations was also presented for spectroscopy analysis. The difference between the simulated and measured reflectance was calibrated by tissue phantom experiments, and the validity of the inverse model for extracting the melanin contained in human skin, or blood oxygen parameters, was verified by human skin experiments.

2. Theory and methods

2.1 Instrumentation

An optical fiber spectrometer (Ocean Optics, USA), including a light source (HL-2000), an optical fiber probe (Ocean Optics R400-7), and a spectrophotometer (USB-4000), was used here. Figure 1 shows the face of the probe, from which a central fiber collects reflected light and the surrounding six fibers deliver light. The diameter of the fibers was 400 μm (NA = 0.22), the center-to-center and the largest separations between the source and the detection fibers were 480 μm and 880 μm, respectively. Spectrum in the range of 500–900 nm with a resolution of 1.5 nm (FWHM) was used for analysis. All spectra were calibrated on a certified reflectance standard (SRS-99-020, Labsphere, USA). Calibration spectra were collected from a fixed 5 cm height above the standard. A Mexameter (MX 18, Courage-Khazaka, Germany) probe was used to measure the melanin index and erythema index of skin. In the measurement, the both probes were vertical and soft touched on the forearm skin and then the signal was recorded.

Fig. 1 The face of the optical fiber probe. Six 400 μm diameter source fibers surrounded a single central 400 μm diameter detection fiber. The center-to-center spacing was 480 μm. Beyond r = 740 μm, a metal flange supported the fibers (not shown here).

Download Full Size | PDF

2.2 Establishment of a two-layered MC-based LUT

Monte Carlo simulations are the golden standard for simulating light transport in tissue and have been widely used to simulate light distribution in turbid media [30–33]. In order to reduce the simulation time, a GPU acceleration technique was adopted, and considering the size of the source fiber, Wang et al.’s CONV program [34] was used to obtain diffuse reflectance (R(r)). The reflectance was then integrated across the detection fiber to determine the probability that a photon traveling a fixed distance would be collected (for the current probe’s geometry). The simulated reflectance collected by the detection fiber (M_R, dimensionless), was determined by:

M_{R} = 6 \int_{d - r}^{d + r} 2 R (x) \cdot x \cdot arc \cos \frac{d^{2} + x^{2} - r^{2}}{2 d \cdot x} d x

where the factor of 6 accounts for the six source fibers surrounding the detection fiber, R is the reflectance obtained after processing by the CONV program, d means the distance between the center of source and detector fibers, which is 480 μm, and r represents the radius of the fiber, which is 200 μm.

Here, it was assumed that the number of simulated photons was 10⁶; the refractive index of the tissue and the probe was 1.4 and 1.52 respectively; the anisotropy factor was 0.9; and the epidermal layer thickness was 70 μm [35] (the reflectance spectrum was measured from the forearms of young adults). In total, the reflectance of tissue with 24000 different combinations of absorption coefficients in the epidermis (40 values), µ_a,epi, and the dermis (20 values), µ_a,derm, and a reduced scattering coefficient (30 values) in both layers, µ_s’, was simulated (This covered 0 < µ_a,epi < 100 cm⁻¹, 0 < µ_a,derm < 50 cm⁻¹, and 2 < µ_s’ < 70 cm⁻¹). The values of M_R were used to establish a two-layered MC-based LUT for the probe response.

2.3 Simulation of the skin reflectance spectrum

The main contributions to epidermal absorption are melanin packaged in melanosomes, and the flesh [28]:

μ_{a, e p i} (λ) = μ_{a, m e l} (λ) M + μ_{a, b a c k} (λ) (1 - M)

where M is the melanin volume fraction, and

μ_{a, m e l} (λ)

(cm⁻¹) and

μ_{a, b a c k} (λ)

(cm⁻¹) are the absorption coefficients of melanin and the background (due to flesh), respectively [36]:

μ_{a, m e l} (λ) = 6.60 \times 10^{11} λ^{- 3.33}

μ_{a, b a c k} (λ) = 7.84 \times 10^{8} λ^{- 3.255}

Hemoglobin contained in the blood is the main contributor to absorption by the dermis. The absorption spectrum of oxygenated hemoglobin also differs significantly from deoxygenated hemoglobin, thus the absorption coefficient of the dermis is influenced by both the blood content and the oxygen saturation, which can be expressed as:

μ_{a, d e r m} (λ) = B (μ_{a, o x y} (λ) S + μ_{a, d e o x y} (λ) (1 - S))

where B and S refer to the volume fraction and saturation of blood, respectively;

μ_{a, o x y} (λ)

and

μ_{a, d e o x y} (λ)

are the absorption coefficients of oxygenated and deoxygenated hemoglobin in the blood [37], respectively.

The reduced scattering coefficients of the epidermis and dermis (µ_s’) were assumed to be equal, and can be described by the power law dependence on the wavelength [38]:

μ_{s}' (λ) = μ_{s}' (λ_{0}) \cdot {(\frac{λ}{λ_{0}})}^{- b}

Here, λ₀ = 630 nm,

μ_{s}' (λ_{0})

is the scattering magnitude at λ₀, and the b is called scattering power in Ref [39].

Through an interpolation program (named “getR”), the simulated R_s values were yielded from the two-layered MC-based LUT that was established previously.

R_{s} = g e t R (μ_{a, e p i}, μ_{a, d e r m}, μ_{s}')

2.4 Calibration of the simulated spectrum

Due to the simulated reflectance R_s is normalized, R_s must be multiplied by a constant, K, before comparison with the measured reflectance. Tissue phantoms with four concentrations of Intralipid (20%-Intralipid, SichuanKelun Pharmaceutical Co., Ltd, China), and seven concentrations of Indian ink (Indian ink, Solarbio, China), were made to establish the experimental reflectance, $R_{e} (μ_{a}, μ_{s}')$ , and obtain the constant K. Four concentrations of Intralipid, 20%, 5%, 1.25%, and 0.3125%, were used to obtain a range of $μ_{s}'$ from 1.17 to 260 cm⁻¹. The values of $μ_{s}'$ were obtained by fluence rate measurements coupled with the “adding absorber” method [40,41]. The absorption coefficient of all stock India ink solutions had been measured with a high-accuracy spectrophotometer (LAMBDA 950, Perkin Elmer, USA) before usage. Adding a 200 μl stock India ink solution into 200 ml of each of the four Intralipid solutions seven times, resulted in a range of the absorption coefficients of the phantoms from 0.0032 to 31.6 cm⁻¹. After each addition, the reflectance of the phantom was added to the data set $R_{e} (μ_{a}, μ_{s}')$ .

As the tissue phantoms were homogeneous, the results of Monte Carlo simulations for the same absorption coefficients in the epidermis and dermis were fitted with the phantom-measured reflectance. Equation (8) was then used to extract the constant K, which minimized χ (K = 10^3.254).

χ = \sum {[K \cdot \frac{R_{s} (μ_{a}, μ_{a}, μ_{s}')}{R_{e} (μ_{a}, μ_{s}')} - 1]}^{2}

2.5 Extracting physiological and optical parameters from skin spectrum

After measuring the simulated spectrum, $R_{s} (λ)$ , and the measured spectrum, $R_{m} (λ)$ , a least square fitting method was used to obtain the minimal squared residuals $δ$ , as expressed by:

δ = \sum_{λ = 500 n m}^{900 n m} {[K \cdot R_{s} (λ) - R_{m} (λ)]}^{2}

Equation (9) was used in an “fminsearch” function in MATLAB to extract the biological and optical parameters from the skin spectrum, i.e., B, S, M, $μ_{s}' (λ_{0})$ , and b. The initial values of the above parameters (B, S, M, $μ_{s}' (λ_{0})$ , and b) were chosen from parameter data set randomly. If the minimal squared residual was still too big, the chosen would be repeated again. The parameters data set was established based on the literatures [28,42].

2.6 Measurement of the melanin values

In order to evaluate the validity of the inverse model, a verified instrument [43], Mexameter MX 18, was used to measure the melanin index of skin. This instrument has previously been used to evaluate the color of skin [44,45]. The melanin measurements were carried out on the volar part of the forearm ten times at each position, by both the optical fiber spectrometer and the Mexameter. All of the experimental data were from 27 healthy volunteers, including 9 white people (Russian), 12 yellow people (Chinese) and 6 brown people (3 Lebanese and 3 Peruvian), with a mean age of 23 ± 2 years.

2.7 Measurement of the blood volume fraction and saturation values

To validate the fraction of oxygenated hemoglobin and deoxygenated hemoglobin extracted from this spectroscopic analysis, two human experiments were conducted, forearm venous occlusion and ischemia experiment, and hot compress experiment. The volunteer for these measurements was taken from the above-mentioned experimental group.

Considering the individual difference, a suitable cuff pressure (40 mmHg) in venous occlusion for the volunteer was found by increasing the pressure from 10 mmHg to 50 mmHg in steps of 10 mmHg gradually.

The measurement sequence of the former experiment was as follows. First, a pneumatic cuff was placed around the arm above the elbow and the forearm was positioned above the heart level to allow for rapid venous drainage. Next, the probe was placed on the inner forearm and a 7-min rest was given to ensure the blood flow was stable. After 5 min (of this rest period) measurement of the spectrum of the skin commenced (measuring each second). When the rest time was exhausted, the cuff was inflated to a pressure of about 40 mmHg for 100 s, in order to obtain a venous occlusion, and then released. Another 4 min rest period was allowed to recover the state of the arm, after which an ischemia was induced by inflating the cuff to a pressure of about 240 mmHg for 45 s. The pressure was then released and a final rest period of 4 min was given before the end of the measurement.

In the hot compress experiment, the fiber probe was soft touched on the inner forearm at first, and then a 7 min rest was given. After 5 min (of this rest period), start recording the spectrum of the skin (measuring each second). When the rest time was exhausted, stop recording the spectrum and place a hot-water bag with 51°C to the measured point for 25 s. And then, the hot-water bag was removed, and 5-min spectrum measurement was conducted before the end of the measurement. For further comparing, another setup, Mexameter, was also used to measure the erythema index. The procedure was similar, a 7-min rest (begin measuring at last 2-min), 25-s hot compress, and 5-min measurement (measuring each ten second).

All of the above experiments were repeated three times.

3. Results

3.1 The two-layered MC-based LUT for spectrum analysis

The lookup table, established by the two-layered MC simulations with various optical properties, is given in Fig. 2.

Fig. 2 The two-layered MC-based LUT for fiber probe reflectance.

Download Full Size | PDF

All of these optical properties, µ_a,epi, µ_a,derm and µ_s’, have the same dimension: cm⁻¹. The color-bar indicated the range of reflectance intensity, M_R, where red denoted high intensity and blue denoted low intensity. According to the variation of the color, it can be found that when the dermal absorption coefficient, µ_a,derm, and the reduced scattering coefficient, µ_s’, were both fixed, the reflectance, M_R, decreased with an increase in the epidermal absorption coefficient, µ_a,epi. For a constant µ_a,epi, the reflectance also decreased with an increase in µ_a,derm when µ_s’ was fixed, and for a constant µ_a,derm, increasing µ_s’ increased M_R, too.

3.2 Typical spectrum and parameters of different skin types

Figure 3(a)-3(c) show typical spectrum of forearm skin with different types, white, yellow and brown, respectively. The black circles were the measurement data, and the red line was the fit data with the established spectroscopic analysis, which showed a good consistency. The parameters including mean and deviation were calculated from the measured spectra (measuring ten times at chosen point), and all mean data were in a reasonable range [28]. The sequence of the melanin content in various skin, white (2.1%) < yellow (4.5%) < brown 8.2%), was in coincidence with reality. It also can be found that the reflectance intensity from darker skin was much lower than lighter skin in the visible light, but the both were similar in the near infrared light. This may because of the strong absorption property of melanin in the shorter wavelength. The results also show that the blood volume fraction increases and blood saturation decreases when the color of skin becomes darker.

Fig. 3 Typical spectrum and parameters of forearm skin with different types, (a) white skin, (b) yellow skin, and (c) brown skin.

Download Full Size | PDF

3.3 Evaluation of the validity of the melanin volume fraction

To verify the inverse model for calculating melanin content, the optical fiber spectrometer and the melanin probe (Mexameter) were used to measure the melanin content of human skin. Figure 4 shows the melanin contents determined by both methods, which indicated linear behavior. The red, blue and black circles were the melanin contents measured from white, yellow and brown skin, respectively. The melanin volume fraction was extracted from the reflectance spectrum based on an analysis of the absorption of skin and ranged from 2.0% to 8.2% (2.0% to 2.6% for white, 3.2% to 6.2% for yellow, and 6.2% to 8.2% for brown), whilst the Mexameter gave an index of melanin content ranging from 91 to 334 (91 to 128 for white, 161 to 241 for yellow, and 244 to 334 for brown). The data from both methods resulted in a high Pearson correlation coefficient of 0.97, which indicates good agreement. And the vertical and horizontal segment denoted the error bar in the melanin measurement by the established spectroscopic analysis method and the Mexameter, respectively.

Fig. 4 Correlation between the melanin index and the melanin volume fraction.

Download Full Size | PDF

3.4 Validation of blood volume fraction and blood saturation measurements

To demonstrate the validity of the reflectance spectroscopy for the measurement of the blood volume fraction (B) and the blood saturation values (S), the forearm venous occlusion and ischemia measurement, and hot compress experiment, outlined in section 2.7, were conducted. For these measurements, the dynamic reflectance spectrum was measured with the optical fiber spectrometer, and the blood oxygen parameters were deduced from the inverse model.

Figure 5(a)-5(b) show the typical changes in B and S during venous occlusion and ischemia, where blue circles were the forearm venous occlusion period, and red circles were the forearm ischemia period. During the 100-s occlusion period (blue circles), the B gradually increased, and the S decreased from 40% to 25%. After releasing the pressure, the B and the S immediately returned to the initial value. A 4-min rest period followed, after which the 45-s ischemia period (red circles) ensued, where both the B and the S showed a decrease, although with differing trends; the B exhibited an exponential decrease, whereas the S levels exhibited a linear decrease. The subsequent release of pressure resulted in an increase in the B and the S levels. About 30 s later both values reached a peak that exceeded the baseline value, after which they both gradually relaxed.

Fig. 5 The tracings of the blood volume fraction, B, and the blood saturation, S, changes during forearm venous occlusion and ischemia. Blue and red circles were the period of forearm venous occlusion and forearm ischemia, respectively. (a) Typical values of B, (b) typical values of S, (c) relative values of B, and (d) relative values of S.

Download Full Size | PDF

Figure 5(c)-5(d) show relative changes in both parameters (B and S) with error bars during the experiment. The data at every 10 s were chosen in order to make points and error bars in the Fig. clearly. The tracing of the relative changes were similar with above-mentioned (typical changes in B and S).

Figure 6 shows the tracings of B, S (with our method), and erythema index (with Mexameter), changes during the hot compress experiment. The upper panel and the lower panel were typical and relative values, respectively. We can observe all of the values (B, S and erythema index) were increased after 25-s hot compress, and then recovered to the baseline. And the relative increase in the B and the S was larger than that in the erythema index.

Fig. 6 The changes in the blood volume fraction, B, the blood saturation, S, and the erythema index during hot compress. The upper panel was typical result, and the lower panel was relative result.

Download Full Size | PDF

4. Discussion

A 3D-LUT based on the inverse model for human skin reflectance spectroscopy and two-layered Monte Carlo simulations was presented, and used to calculate reduced scattering coefficient, human melanin and blood oxygen parameters. The typical result (in subsection 3.2) shows the deviation of the extracted blood oxygen parameters was a little higher than that of other parameters. This might be because of the fiber probe pressure in measurements. It should be noted that the blood saturation of health human is over 95%, which is much higher than the measurements based on the reflectance spectrum in this work. This is because the source-detector distance is too short to detect the deep information of tissue, and veins are usually located in superficial layer of skin and arteries are located in deeper tissue. Therefore, the measured saturation level is much smaller than that of body.

Human skin is organized into different types of tissues and vessels with a complex structure. For instance, the dermis under the epidermis is consisting of upper blood net dermis, reticular dermis, and deep blood net dermis [46]. For brown skin, photon will main located in the relative superficial zone, such as upper blood net dermis which has more blood vessels than reticular dermis. That is why the blood volume fraction in the brown skin was larger than that in lighter skin.

The melanin measurements indicated that the melanin volume fraction of skin by our proposed inverse model has a consistency with the melanin index by the commercial Mexameter. Although the both are diffuse-reflectance-based, the latter has been verified to be a high repeatable and sensitive device with other methods, DermaSpectrometer (Cortex Technology) (based on diffuse reflectance) and Chromameter CR 200 (Minolta) (based on the CIE L*a*b* color system) [43]. One thing to note is that the study subjects reported in this study (white, yellow and brown skin) are all relatively light skin colors. Darker skin has higher absorption for its higher melanin content, which results in lower reflectance levels increasing the error of the system. Therefore, the further investigations need to be studied.

A previous investigation showed that the range of melanin index for Asian and Indian skin measured using the Mexameter was 150-600, and for Caucasian skin was 50-250 [47]. Here, the melanin index for the brown skin fell within a range of 244-334, and 161-261 for yellow, 91-128 for white. The melanin content of skin was also analyzed previously using an optical fiber spectrometer, and was typically 1.3-6.3% for light-skinned adults, 11-16% for moderately pigmented adults, and 18-43% for darkly pigmented adults [36]. Whilst the measured melanin volume fractions presented here (in the range of 6.2-8.2% for brown skin, 3.2-6.2% for yellow, and 2.0-2.6% for white) might fall in the lower range compared to the previous investigation, this is expected due to different inverse models and parameters of the skin studied. The average epidermal thickness of the forearm for the young volunteers (mean age 23 ± 2 years) was about 70 μm [35], which is significantly larger than the 60 μm cited in Ref [36]. Such a difference would result in an overestimate of the melanin in the epidermis.

With regards to the blood oxygen parameters measurement, an increase in the blood volume fraction was observed during venous occlusion due to prevention of blood flow back to the heart. The increasing blood volume fraction could prevent the fresh blood from flowing in, thus the continuous oxygen consumption could result in a decrease in the blood saturation. Similarly, in common sense, during the ischemia, the blood from the heart could not flow into the forearm, and the blood in the forearm could not flow back to the heart, so the blood volume fraction should not be changed (with respect to initial values), as was shown in Ref [48]. However, in this work, the blood volume fraction was decreased during the ischemia. This distinction might result from the different source-detection distance (480 μm here, which was much shorter than 3.5 cm in Ref [48].). The shorter source-detection distance can detect more superficial information.

To check whether the information in the deep blood net dermis can be obtained by the established setup, a Monte Carlo simulation was conducted to quantify the probability of the emitted photon reaching the depth from 5 to 1500 μm in steps of 5 μm, as shown in Fig. 7. Here, the largest separation between the source fiber and detector, 880 μm, was used in the simulation, rather than 480 μm, center-to-center distance. When the wavelength of the incident light was 500 nm, the peak value of the probability was appeared at 195 μm, and that was occurred at 285 and 385 nm with 700 and 900 nm incident light, respectively. The simulated result shows, in the range of 500-900 nm incident light, the largest depth that the most photons can reach was within 385 μm (the layer of reticular dermis: in depth 330-2020 μm). This means the detected photons had little chance to reach the next layer, deep blood net dermis (the depth and the structure information of skin were according to the Ref [46]). Thus, our setup was much more sensitive to the superficial vessels.

Fig. 7 The probability of the emitted photon (880 μm away from the light incident point) reaching the depth from 5 to 1500 μm in steps of 5 μm by Monte Carlo simulation.

Download Full Size | PDF

During the ischemia, the blood in the deeper dermis would lack the dynamic to flow to superficial areas without boot flow to the forearm. It would lead to a decrease in the observed blood volume fraction from the upper dermis. The blood content in the deeper dermis would increase to prevent upper superficial blood from flowing down, thus further decreasing the speed (of decrease) of the blood volume fraction. Moreover, the blood saturation would decrease at a fast and continuous rate due to continuous metabolism during this period. After ischemia, both the blood volume fraction and blood saturation values exceeded the baseline due to immediate release of blood into the forearm. Therefore, the obtained results were in well agreement with physiological changes. In addition, the physiological changes during the hot compress experiment were also well reflected in the variation of the blood volume fraction and blood saturation.

All the obtained values, melanin, blood volume fraction, B, and blood saturation, S, can influence the color and appearance of skin. With above parameters, the established spectroscopic analysis can provide an objective appraisal on the efficacy of the cosmetics, so it has potentiality to be applied in cosmetology. In addition, the peripheral vascular diseases, such as venous thrombosis, can induce the abnormal of two parameters (B and S) in the superficial skin. Therefore, our proposed method is expected to be a supplementary diagnostic. And results from the hot compress demonstrate that our analysis method is more sensitive for obtaining the blood parameters than the Mexamter because the relative change of blood parameters by former (50%) was larger than that by latter (30%). Thus, the superficial information of the skin obtained successfully with the established setup has great potential in the cosmetics and medical industry.

Here, the inverse model was developed based on the 3D-LUT with Monte Carlo simulations. Compared with the experiment-based LUT obtained by Tunnell et al. [15], the two models show negative correlation between the reflectance, M_R, and the absorption coefficient, µ_a; and positive correlation between M_R and the reduced scattering coefficient, µ_s’. However, for the experiment-based LUT, the homogeneous assumption for skin limits the accuracy of calculating the optical properties and physiological parameters. This is because the melanin is found mainly in the epidermal layer, whose contents affect the absorption property of the epidermis, and the blood only lies in the dermal layer. For the skin, which is not too dark, one-layer skin model may induce more errors.

Recently, a similar work using diffuse reflectance spectroscopy to eliminate optical and structural parameters of the skin with Monte Carlo simulations and a multilayer skin model was reported [29]. A Monte Carlo simulation with different epidermis thicknesses and variable scattering was used to extract the distribution of the path lengths, but no absorption properties were considered. As the skin has suction chromophores, such as superficial melanin and deeper blood, the alteration of upper-layer absorption would result in a different path length in the lower layer. On the other hand, the inverse model used here not only considers the multilayer structure of the skin, but also combines the absorption and scattering properties in the Monte Carlo simulations. Thus, the inverse model may be more accurate for analyzing skin spectra. Note, however, that because the measurements in this work were all conducted on the forearms of late teens, the thickness of the epidermis was set as 70 μm [35]; this limits the applicability of the current inverse model for other kinds of skin. Thus, to apply this inverse model to other parts of the skin (e.g. away from the forearm), another MC-based LUT for a different thickness of the epidermis, or a four-dimensional LUT containing variable thicknesses of the epidermis, would be required. Although the analysis method proposed here remains universal, this will be the focus of future study.

5. Conclusion

In this paper, an inverse model based on a LUT was developed to extract the physiological parameters and the optical properties of skin (i.e., melanin volume fraction, blood volume fraction, blood saturation, reduced scattering coefficient at 630 nm, and scattering power). The LUT was established by two-layered Monte Carlo simulations that considered absorption and scattering. Comparison of melanin measured by an optical fiber spectrometer and a Mexameter, and the dynamic blood volume fraction and blood saturation during two experiments (a forearm venous occlusion and ischemia, and hot compress), were performed in order to verify the validity of the inverse model. The Pearson coefficient for the measured melanin volume fraction and the melanin index was 0.97, indicating high correlation. Good agreement between the measured and theoretical values of the blood volume fraction and blood saturation was also demonstrated. In addition, our setup was much sensitive than the commercial Mexameter in blood oxygen parameters measurement. Thus, this inverse model has great potential for the evaluation of the characteristics of skin that could be very useful for non-invasive diagnoses of skin diseases, or evaluation of aesthetic efficacy.

Acknowledgments

This study was supported by the National Nature Science Foundation of China (Grant Nos. 81171376 & 91232710), the Research Fund for the Doctoral Program of Higher Education of China (Grant No. 20110142110073), and the Science Fund for Creative Research Group (Grant No. 61121004).

References and links

1. I. Bodén, D. Nilsson, P. Naredi, and B. Lindholm-Sethson, “Characterization of healthy skin using near infrared spectroscopy and skin impedance,” Med. Biol. Eng. Comput. 46(10), 985–995 (2008). [CrossRef] [PubMed]

2. J. O’Doherty, J. Henricson, C. Anderson, M. J. Leahy, G. E. Nilsson, and F. Sjöberg, “Sub-epidermal imaging using polarized light spectroscopy for assessment of skin microcirculation,” Skin Res. Technol. 13(4), 472–484 (2007). [CrossRef] [PubMed]

3. G. N. Stamatas, J. Nikolovski, M. C. Mack, and N. Kollias, “Infant skin physiology and development during the first years of life: a review of recent findings based on in vivo studies,” Int. J. Cosmet. Sci. 33(1), 17–24 (2011). [CrossRef] [PubMed]

4. R. Marchesini, N. Cascinelli, M. Brambilla, C. Clemente, L. Mascheroni, E. Pignoli, A. Testori, and D. R. Venturoli, “In vivo spectrophotometric evaluation of neoplastic and non-neoplastic skin pigmented lesions. II: Discriminant analysis between nevus and melanoma,” Photochem. Photobiol. 55(4), 515–522 (1992). [CrossRef] [PubMed]

5. B. W. Murphy, R. J. Webster, B. A. Turlach, C. J. Quirk, C. D. Clay, P. J. Heenan, and D. D. Sampson, “Toward the discrimination of early melanoma from common and dysplastic nevus using fiber optic diffuse reflectance spectroscopy,” J. Biomed. Opt. 10(6), 064020 (2005). [CrossRef] [PubMed]

6. Th. Forster, U. Issberner, and H. Hensen, “Lipid/surfactant compounds as a new tool to optimize skin-care properties of personal-cleansing products,” J. Surfactants Deterg. 3(3), 345–352 (2000). [CrossRef]

7. L. Kilpatrick-Liverman, P. Kazmi, E. Wolff, and T. G. Polefka, “The use of near-infrared spectroscopy in skin care applications,” Skin Res. Technol. 12(3), 162–169 (2006). [CrossRef] [PubMed]

8. Q. Sun, M. Tran, B. Smith, and J. D. Winefordner, “In-situ evaluation of barrier-cream performance on human skin using laser-induced breakdown spectroscopy,” Contact Dermat. 43(5), 259–263 (2000). [CrossRef] [PubMed]

9. R. R. Anderson and J. A. Parrish, “The optics of human skin,” J. Invest. Dermatol. 77(1), 13–19 (1981). [CrossRef] [PubMed]

10. M. J. C. Van Gemert, S. L. Jacques, H. J. C. M. Sterenborg, and W. M. Star, “Skin optics,” IEEE Trans. Biomed. Eng. 36(12), 1146–1154 (1989). [CrossRef] [PubMed]

11. A. N. Bashkatov, E. A. Genina, and V. V. Tuchin, “Optical properties of skin, subcutaneous, and muscle tissues: a review,” J. Innov. Opt. Health Sci. 4(01), 9–38 (2011). [CrossRef]

12. G. Zonios, L. T. Perelman, V. Backman, R. Manoharan, M. Fitzmaurice, J. Van Dam, and M. S. Feld, “Diffuse Reflectance Spectroscopy of Human Adenomatous Colon Polyps In Vivo,” Appl. Opt. 38(31), 6628–6637 (1999). [CrossRef] [PubMed]

13. G. Zonios and A. Dimou, “Modeling diffuse reflectance from semi-infinite turbid media: application to the study of skin optical properties,” Opt. Express 14(19), 8661–8674 (2006). [CrossRef] [PubMed]

14. G. Zonios and A. Dimou, “Light scattering spectroscopy of human skin in vivo,” Opt. Express 17(3), 1256–1267 (2009). [CrossRef] [PubMed]

15. N. Rajaram, T. H. Nguyen, and J. W. Tunnell, “Lookup table-based inverse model for determining optical properties of turbid media,” J. Biomed. Opt. 13(5), 050501 (2008). [CrossRef] [PubMed]

16. N. Rajaram, T. J. Aramil, K. Lee, J. S. Reichenberg, T. H. Nguyen, and J. W. Tunnell, “Design and validation of a clinical instrument for spectral diagnosis of cutaneous malignancy,” Appl. Opt. 49(2), 142–152 (2010). [CrossRef] [PubMed]

17. S. F. Bish, N. Rajaram, B. Nichols, and J. W. Tunnell, “Development of a noncontact diffuse optical spectroscopy probe for measuring tissue optical properties,” J. Biomed. Opt. 16(12), 120505 (2011). [CrossRef] [PubMed]

18. G. M. Palmer and N. Ramanujam, “Monte Carlo-based inverse model for calculating tissue optical properties. Part I: Theory and validation on synthetic phantoms,” Appl. Opt. 45(5), 1062–1071 (2006). [CrossRef] [PubMed]

19. M. C. Skala, G. M. Palmer, K. M. Vrotsos, A. Gendron-Fitzpatrick, and N. Ramanujam, “Comparison of a physical model and principal component analysis for the diagnosis of epithelial neoplasias in vivo using diffuse reflectance spectroscopy,” Opt. Express 15(12), 7863–7875 (2007). [CrossRef] [PubMed]

20. G. Zonios, J. Bykowski, and N. Kollias, “Skin melanin, hemoglobin, and light scattering properties can be quantitatively assessed in vivo using diffuse reflectance spectroscopy,” J. Invest. Dermatol. 117(6), 1452–1457 (2001). [CrossRef] [PubMed]

21. J. C. Finlay and T. H. Foster, “Hemoglobin oxygen saturations in phantoms and in vivo from measurements of steady-state diffuse reflectance at a single, short source-detector separation,” Med. Phys. 31(7), 1949–1959 (2004). [CrossRef] [PubMed]

22. G. Zonios and A. Dimou, “Modeling diffuse reflectance from homogeneous semi-infinite turbid media for biological tissue applications: a Monte Carlo study,” Biomed. Opt. Express 2(12), 3284–3294 (2011). [CrossRef] [PubMed]

23. L. Lim, B. Nichols, N. Rajaram, and J. W. Tunnell, “Probe pressure effects on human skin diffuse reflectance and fluorescence spectroscopy measurements,” J. Biomed. Opt. 16(1), 011012 (2011). [CrossRef] [PubMed]

24. Z. Qian, S. S. Victor, Y. Gu, C. A. Giller, and H. Liu, “Look-Ahead Distance of a fiber probe used to assist neurosurgery: Phantom and Monte Carlo study,” Opt. Express 11(16), 1844–1855 (2003). [CrossRef] [PubMed]

25. D. Zhu, W. Lu, S. Zeng, and Q. Luo, “Effect of light losses of sample between two integrating spheres on optical properties estimation,” J. Biomed. Opt. 12(6), 064004 (2007). [CrossRef] [PubMed]

26. G. Mantis and G. Zonios, “Simple two-layer reflectance model for biological tissue applications,” Appl. Opt. 48(18), 3490–3496 (2009). [CrossRef] [PubMed]

27. G. Zonios and A. Dimou, “Simple two-layer reflectance model for biological tissue applications: lower absorbing layer,” Appl. Opt. 49(27), 5026–5031 (2010). [CrossRef] [PubMed]

28. D. Yudovsky and L. Pilon, “Rapid and accurate estimation of blood saturation, melanin content, and epidermis thickness from spectral diffuse reflectance,” Appl. Opt. 49(10), 1707–1719 (2010). [CrossRef] [PubMed]

29. I. Fredriksson, M. Larsson, and T. Strömberg, “Inverse Monte Carlo method in a multilayered tissue model for diffuse reflectance spectroscopy,” J. Biomed. Opt. 17(4), 047004 (2012). [CrossRef] [PubMed]

30. C. Jiang, H. He, P. Li, and Q. Luo, “Graphics processing unit cluster accelerated Monte Carlo simulation of photon transport in multi-layered tissues,” J. Innov. Opt. Health Sci. 5(02), 1250004 (2012). [CrossRef]

31. C. Zhu and Q. Liu, “Validity of the semi-infinite tumor model in diffuse reflectance spectroscopy for epithelial cancer diagnosis: a Monte Carlo study,” Opt. Express 19(18), 17799–17812 (2011). [CrossRef] [PubMed]

32. B. Luo and S. He, “An improved Monte Carlo diffusion hybrid model for light reflectance by turbid media,” Opt. Express 15(10), 5905–5918 (2007). [CrossRef] [PubMed]

33. L. Wang, S. L. Jacques, and L. Zheng, “MCML-Monte Carlo modeling of light transport in multi-layered tissues,” Comput. Meth. Prog. Bio. 47(2), 131–146 (1995). [CrossRef]

34. L. Wang, S. L. Jacques, and L. Zheng, “CONV-convolution for responses to a finite diameter photon beam incident on multi-layered tissues,” Comput. Meth. Prog. Bio. 54(3), 141–150 (1997). [CrossRef]

35. T. Gambichler, R. Matip, G. Moussa, P. Altmeyer, and K. Hoffmann, “In vivo data of epidermal thickness evaluated by optical coherence tomography: Effects of age, gender, skin type, and anatomic site,” J. Dermatol. Sci. 44(3), 145–152 (2006). [CrossRef] [PubMed]

36. S. L. Jacques, “Skin Optics,” (1998), http://omlc.ogi.edu/news/jan98/skinoptics.html.

37. S. A. Prahl, “Optical Absorption of Hemoglobin,” http://omlc.ogi.edu/spectra/hemoglobin/index.html.

38. R. T. Zaman, N. Rajaram, B. S. Nichols, H. G. Rylander 3rd, T. Wang, J. W. Tunnell, and A. J. Welch, “Changes in morphology and optical properties of sclera and choroidal layers due to hyperosmotic agent,” J. Biomed. Opt. 16(7), 077008 (2011). [CrossRef] [PubMed]

39. A. Cerussi, N. Shah, D. Hsiang, A. Durkin, J. Butler, and B. J. Tromberg, “In vivo absorption, scattering, and physiologic properties of 58 malignant breast tumors determined by broadband diffuse optical spectroscopy,” J. Biomed. Opt. 11(4), 044005 (2006). [CrossRef] [PubMed]

40. H. J. van Staveren, C. J. M. Moes, J. van Marie, S. A. Prahl, and M. J. C. van Gemert, “Light scattering in Intralipid-10% in the wavelength range of 400-1100 nm,” Appl. Opt. 30(31), 4507–4514 (1991). [CrossRef] [PubMed]

41. F. Martelli and G. Zaccanti, “Calibration of scattering and absorption properties of a liquid diffusive medium at NIR wavelengths. CW method,” Opt. Express 15(2), 486–500 (2007). [CrossRef] [PubMed]

42. S. L. Jacques, “Spectral Imaging and Analysis to Yield Tissue Optical Properties,” J. Innov. Opt. Health Sci. 2(02), 123–129 (2009). [CrossRef]

43. P. Clarys, K. Alewaeters, R. Lambrecht, and A. O. Barel, “Skin color measurements: comparison between three instruments: the Chromameter^®, the DermaSpectrometer^® and the Mexameter^®,” Skin Res. Technol. 6(4), 230–238 (2000). [CrossRef] [PubMed]

44. J. W. Shin, D. H. Lee, S. Y. Choi, J. I. Na, K. C. Park, S. W. Youn, and C. H. Huh, “Objective and non-invasive evaluation of photorejuvenation effect with intense pulsed light treatment in Asian skin,” J. Eur. Acad. Dermatol. Venereol. 25(5), 516–522 (2011). [CrossRef] [PubMed]

45. Y.-H. Li, Y. Wu, J. Z. S. Chen, X. Zhu, Y.-Y. Xu, J. Chen, G.-H. Dong, X.-H. Gao, and H.-D. Chen, “A Split-Face Study of Intense Pulsed Light on Photoaging Skin in Chinese Population,” Lasers Surg. Med. 42(2), 185–191 (2010). [CrossRef] [PubMed]

46. T. Maeda, N. Arakawa, M. Takahashi, and Y. Aizu, “Monte Carlo simulation of spectral reflectance using a multilayered skin tissue model,” Opt. Rev. 17(3), 223–229 (2010). [CrossRef]

47. Courage & Khazaka, “Information and operating instructions for the Cutometer MPA 580 and its probe,” Koln, Germany: CK electronic GmbH (2005).

48. R. A. De Blasi, N. Almenrader, P. Aurisicchio, and M. Ferrari, “Comparison of two methods of measuring forearm oxygen consumption (VO₂) by near infrared spectroscopy,” J. Biomed. Opt. 2(2), 171–175 (1997). [CrossRef] [PubMed]

Lookup-table-based inverse model for human skin reflectance spectroscopy: two-layered Monte Carlo simulations and experiments

Abstract

1. Introduction

2. Theory and methods

2.1 Instrumentation

2.2 Establishment of a two-layered MC-based LUT

2.3 Simulation of the skin reflectance spectrum

2.4 Calibration of the simulated spectrum

2.5 Extracting physiological and optical parameters from skin spectrum

2.6 Measurement of the melanin values

2.7 Measurement of the blood volume fraction and saturation values

3. Results

3.1 The two-layered MC-based LUT for spectrum analysis

3.2 Typical spectrum and parameters of different skin types

3.3 Evaluation of the validity of the melanin volume fraction

3.4 Validation of blood volume fraction and blood saturation measurements

4. Discussion

5. Conclusion

Acknowledgments

References and links

Cited By

Figures (7)

Equations (9)

Optics Express