Development of a Method for the Determination of Human Skin

Sep 12, 2001 - μm thick, except on the palms and soles, and is composed of partially flattened and keratinized layers.2 If, as a result of environmen...
4 downloads 0 Views 150KB Size
Download PDF
Anal. Chem. 2001, 73, 4964-4971

Development of a Method for the Determination of Human Skin Moisture Using a Portable Near-Infrared System Young-Ah Woo, Jhii-Weon Ahn, In-Koo Chun, and Hyo-Jin Kim*

College of Pharmacy, Dongduk Women’s University, Seoul, Korea

In this study, a portable near-infrared (NIR) system was newly integrated with a photodiode array detector that has no moving parts, and this system has been successfully applied for the evaluation of human skin moisture. The good correlation between NIR absorbance and the absolute water content of separated hairless mouse skin, in vitro, was showed, depending on the water content (7.484.9%) using this portable NIR system. Partial least squares (PLS) regression was used for calibration with the 1150-1650-nm wavelength range. For practical use for the evaluation of human skin moisture, the PLS model for human skin moisture was developed in vivo using the portable NIR system on the basis of the relative water content values of stratum corneum from the conventional capacitance method. The PLS model showed a good correlation. This study indicated that the portable NIR system, as compared to conventional methods, could be a powerful tool for human skin moisture, which may be much more stable to environmental conditions, such as temperature and humidity. Furthermore, to confirm the performance of the newly integrated portable NIR system, a scanning-type conventional NIR spectrometer was used in the same experiments, and the results were compared. Human skin consists of epidermis and dermis. The softness and pliability of skin are the main characteristic factors for protecting the body and assisting in motion.1 These factors are dependent on the amount of moisture contained in the stratum corneum, which is the outermost layer of epidermis, and are controlled by the barrier function that maintains an adequate water content in the skin layer. The stratum corneum is about 10-40 µm thick, except on the palms and soles, and is composed of partially flattened and keratinized layers.2 If, as a result of environmental changes, the health of the stratum corneum is not maintained, the efficiency of the barrier and moisture-maintaining functions of the skin will drop off. As a result, the skin becomes easily dried, roughened, and even more susceptible to infection. Therefore, it is very important to maintain sufficient moisture in the stratum corneum for healthy skin. Electric conductance,3-6 transepidermal water loss (TEWL),7 and infrared spectroscopy by attenuated total reflectance (ATR)8-10 have been used as conventional skin moisture measuring devices. * Corresponding author. Phone: 82-2-940-4525. Fax: 82-2-943-9578. E-mail: [email protected]. (1) Obata, M.; Tagami, H. J. Soc. Cosmet. Chem. 1990, 41, 135. (2) Rushmer, R. F.; Buettner, K. J. K.; Short, J. M.; Odland, G. F. Science 1966, 154, 343.

4964 Analytical Chemistry, Vol. 73, No. 20, October 15, 2001

TEWL measures the rate of evaporation of water from the skin surface, which is extremely environment-sensitive and requires several minutes of equilibration time for stable readings. Infrared spectroscopy is a direct moisture measuring method at a certain wavelength range. Because this is not only very expensive but also difficult to operate, it is rarely used for commercial purposes. In addition, ATR measurements depend on the ambient conditions and are restricted to the uppermost stratum corneum. Electric conductance measuring devices, including the capacitance method, has been widely used until now. When the alternating current with a constant frequency is applied to the skin, electric conductance is measured, and the skin moisture is calculated from the electric conductivity, which is dependent on the water content of the skin. However, these devices are influenced by external temperature and humidity, which requires that they be kept at a constant temperature and humidity. In addition, the amount of electrolyte that skin contains can change the conductance value without regard to water content. The uses of near-infrared reflectance spectroscopy for measurement of skin moisture have been reported.11-14 In vitro or in vivo, they showed a good correlation between water content and NIR absorbance in addition to the advantages of NIR analysis over other measurements.15,16 These studies were performed using conventional NIR instruments that were equipped with an integrating sphere or moving grating, which in practical terms are hard to move and handle. According to a recent overview on NIR spectroscopy of skin,17,18 diffuse reflectance spectra recorded with (3) Elsner, P.; Berardesca, E.; Maibach, H. Bioengineering of the Skin: Water and the Stratum Corneum; CRC Press, Boca Raton, FL, 1994; Chapter 14. (4) Martinsen, O. G.; Grimnes, S.; Karlson, J. Skin Pharmacol. 1995, 8, 237. (5) Salter, D. C. Skin Bioeng. 1998, 26, 38. (6) Tagami, H. Acta Derm. Venerol. Suppl. (Stockh.) 1994, 185, 29. (7) Potts, R. O. J. Soc. Cosmet. Chem. 1986, 37, 171. (8) Lucassen, G. W.; Veen, G. N. A.; Jansen, J. A. J. J. Biomed. Optics 1998, 3, 267. (9) Prasch, TH.; Knu ¨ bel, G.; Schmidt-Fonk, K.; Ortanderl, S.; Nieveler, S.; Fo ¨rster, TH. Int. J. Cos. Sci. 2000, 22, 371. (10) Wichrowski, K.; Sore, G.; Khaı¨at, A. Int. J. Cosmet. Sci. 1995, 17, 1. (11) Rigal, J.; Losch, M. J.; Bazin, R.; Camus, C.; Sturelle, C.; Descamps, V.; Le´veˆque, J. L. J. Soc. Cosmet. Chem. 1993, 44, 197. (12) Martin, K. J. Soc. Cosmet. Chem. 1993, 44, 249. (13) Sowa, M. G.; Payette, J. R.; Mantsch, H. M. J. Surg. Res. 1999, 86, 62. (14) Libnau, F. O.; Kvalheim, O. M.; Christy, A. A.; Toft, J. Vib. Spectrosc. 1994, 7, 243 (15) Walling, P. L.; Dabney, J. M. J. Soc. Cosmet. Chem. 1989, 40, 151. (16) K. Martin, Appl. Spectrosc. 1998, 52, 1001. (17) Heise, H. M. In Infrared and Raman Spectroscopy of Biological Materials; Gremlich, H. U., Yan, B., Eds.; Marcel Dekker: New York, 2000. (18) Kumar, G.; Shumitt, J. M. Appl. Optics 1997, 36, 2286. 10.1021/ac0102563 CCC: $20.00

© 2001 American Chemical Society Published on Web 09/12/2001

Figure 1. Schematic diagram of newly integrated portable NIR system.

the use of an optical fiber probe were valuable for clinical diagnostics. In our study, a new portable NIR system was integrated to determine the water content of skin rapidly and in a stable manner, even under severe environmental conditions. To elucidate the relationship between water content and NIR absorbance using partial least squares (PLS) regression,19,20 separated hairless mouse skin was used in vitro. We also developed a PLS model using NIR spectra of human skin and confirmed the potential that a NIR portable system with PLS model can be practically used for the rapid and stable measurement of human skin moisture. EXPERIMENTAL SECTION Portable NIR System. The newly integrated system has been focused on the development of a compact spectrometer equipped with an optical fiber using microchip technology. A schematic diagram for the portable NIR system is presented in Figure 1. This system includes a tungsten halogen lamp for the light source, an InGaAs photodiode array for the microspectrometer, an internal battery of portable size (DC 12V), and software for interfacing and chemometrics. RS-232c cable was used for the connection to a computer. Figure 2 shows the schematic diagram of the fiber-optic probe (microParts, Germany) that was used. The fiber-optic probe has a total of nine reflectance fiber-optic bundles and eight surrounding bundles for illumination of the sample and one center bundle for receiving the light from the sample (Figure 2a). To receive the light effectively and to avoid contacting the skin surface directly, a 0.3-mm gap was maintained between the fiber terminal and the skin surface by the use of a holder, which could strongly support the fiber-optic bundle to obtain a stable spectrum (Figure 2b). A (19) Sharaf, M. A.; Illman, D. L.; Kowalski, B. R. Chemometrics; John Wiley: New York, 1986. (20) Beebe, K. R.; Pell, R. J.; Seasholtz, M. B. Chemometrics: A Practical Guide; John Wiley: New York, 1998.

Figure 2. Schematic diagram of fiber-optic probe that was used: (a) terminal view, and (b) lateral section view.

ceramic was used as a reference standard to convert raw data into reflectance spectra of samples. Each reflectance spectrum Analytical Chemistry, Vol. 73, No. 20, October 15, 2001

4965

Figure 3. Water content changes of hairless mouse skin. S* means after soaking 1 h in water.

was derived from a mean of 50 scans collected between 1100 and 1750 nm at 2 nm intervals. Conventional Scanning-Type NIR Spectrometer. To test the performance of the developed portable system, conventional a scanning-type NIR spectrometer was used, and the calibration results were compared. We used a NIRSystems model 6500 spectrometer (Foss NIRSystems Inc., Silver Spring, MD), which has been widely used. This system is equipped with a quartz halogen lamp, silicon detectors, and PbS detectors. NIR reflectance spectra were collected over the 700-2200-nm spectral region using reflectance fiber-optics (Foss NIRSystems Inc., Silver Spring, MD). The spectra were collected using 2-nm data intervals. Each sample spectrum was obtained by averaging 50 scans. All of the spectra were recorded as log(1/R) with respect to a ceramic reference standard. Hairless Mouse Skin. Dorsal skin from two hairless male mice was used. The epidermal parts were separated from the dermal tissues and were cut into 6 pieces. The size of each piece was ∼1 × 10-4 m2, and the thickness was 313 ( 79 µm. It was backed by aluminum foil for the measurement. The pieces were labeled A to F. The labeled skin pieces were weighed and soaked

for 1 h so that they would contain a maximum amount of water. After that, we weighed them at 1-h intervals while drying in the desiccator with silica gels. After 18 h, the skin pieces were dried at 105 °C until they were of constant weights. The total drying time was 90 h. NIR spectra of each piece of skin were acquired by using the portable NIR system and a scanning-type NIR spectrometer with reflectance fiber-optic right after every weighing. Through this procedure, the 138 NIR spectra of the hairless mouse skin samples with various ranges from 7.4 to 84.9% were acquired. Human Skin. To develop a PLS model for the determination of human skin moisture using the portable NIR system, 200 NIR spectra were collected from the arms of 10 persons. A capacitance method using Corneometer CM 825 (Courage-Khazaka, Ko¨ln, Germany) was used for reference values for the relative water content of human skin. The relative water content for all of the samples ranged from 44.3 to 82.0. NIR reflectance spectra and reference values using the Corneometer CM 825 were acquired at the same sites on the arms. PLS regression was applied to describe the correlation between NIR absorbance and relative water content value of the skin. All experiments were performed in the laboratory, which was controlled at 20 °C and 45% to obtain stable reference values, because the capacitance method is environment-sensitive. Evaluation of the NIR Model. The samples were divided into a calibration set for modeling and a prediction set for the evaluation of the developed model. The prediction set consisted of samples that were not used for the calibration set. Each developed NIR model was evaluated as the standard error of calibration (SEC) for the calibration set and the standard error of prediction (SEP) for the prediction set.

SEC )

x

n

∑

(yˆi - yi)2

i)1

n-p

,

SEP )

x

n

∑(yˆ - y )

2

i

i

i)1

n

where yˆi is the NIR predicted value, yi is the reference value, n is the number of data spectra, and p is the number of factors used.

Figure 4. (a) NIR spectra of hairless mouse skin by using portable NIR system. (b) Second derivative NIR spectra of hairless mouse skin by using scanning-type NIR spectrometer. 4966

Analytical Chemistry, Vol. 73, No. 20, October 15, 2001

Figure 5. Effect of the number of PLS factors on the standard error of calibration (SEC) and the standard error of prediction (SEP) for the calibration model by using the portable NIR system. Table 1. Calibration Results of the Water Content of Hairless Mouse Skin in the 1150-1650-nm Wavelength Range Using the Portable NIR System and a Scanning-Type NIR Spectrometer spectrometer portable scanning-type

spectral treatment none 1D 2D none 1D 2D

no. factors

SEC, %

SEP, %

9 9 7 6 4 6

5.8 6.2 6.2 4.6 5.4 5.1

5.9 6.8 7.2 5.7 6.6 5.6

Repeatability Test. When acquiring NIR spectra of human skin, repeatability tests of spectral measurement were implemented for both the portable NIR system and the scanning-type NIR spectrometer. Six Consecutive spectral measurements were acquired by using the fiber-optic probe. The relative standard deviation was calculated in the 1150-1650-nm wavelength range for both the portable NIR system and the scanning-type spectrometer. In addition, for the scanning-type spectrometer, the relative standard deviation was derived from the 1150-2200-nm wavelength region. RESULTS AND DISCUSSION Hairless Mouse SkinsPortable NIR System. Figure 3 shows the water content changes of hairless mouse skin, depending on the time. Since the hairless mouse skin was first soaked in water for 1 h in order to acquire a data set with a broad water content range, every second point had the highest water content. The loss of water of separated hairless mouse skin was slow when the skin was placed for in the desiccators charged with silica gel. After 18 h, the skin samples were dried at 105 °C in an oven until they were of constant weight, and the water content declined rapidly. We acquired the hairless mouse skin samples with various water content ranges, and the near-infrared reflectance spectra of the hairless mouse skin samples were collected. Only four NIR spectra from discrete drying steps after baseline normalization at 1100 nm were presented for the clear comparison, as shown in Figure 4a. The huge band at 1450 nm in the spectrum is due to the overtone of OH band stretching of water. The water

Figure 6. (a) Scatter plot showing the correlation between the NIR value and the water content of hairless mouse skin using the 11501650-nm range (portable NIR system). (b) Scatter plot showing correlation between the NIR value and the water content of hairless mouse skin using the 1150-2200-nm range (scanning-type NIR spectrometer).

band changes depending on the water content were clearly observed. However, the baseline of NIR spectra of hairless mouse skin was extremely shifted as a result of the scattering effect of skin. Despite the correlation between the NIR absorbance and the water content around 1450 nm, the scattering effect on the spectra was more dominant. It was difficult to use a classical univariate calibration method for water content of skin because of the complex sample system. In addition, one or several wavelength(s) related to water content could not be found because of the interferences of the other components of skin on the NIR spectra, as well as the scattering effect of the skin surface. Partial least squares (PLS) regression was, therefore, used for the development of a calibration model for the water content of skin, which is a powerful multivariate calibration method21 used to elucidate the relationship between NIR absorbance and concentration of interest, even in the complex system. Derivative techniques were carried out to remove baseline shift and enhance the spectral features before PLS modeling. To develop a robust (21) Martens, H.; Næs, T. M. Multivariate Calibration; John Wiley: New York, 1989

Analytical Chemistry, Vol. 73, No. 20, October 15, 2001

4967

Figure 7. Enlarged noise spectra, as obtained by subtracting two consecutively recorded log R skin spectra using (a) the portable NIR system and (b) the scanning-type NIR spectrometer.

model, several conditions, such as first and second derivatization, were considered. The spectral ranges from 1150 to 1650 nm and 1350 to 1500 nm were investigated. However, using only the range 1350-1500 nm, the calibration results were much worse than when using the whole range. The PLS models using 1150-1650 nm were developed using log 1/R, first derivative, and second derivative spectra, as listed in Table 1. The optimum number of PLS factors was identified as the number of factors that gives a minimum SEP. As the number of factors increased, the SEP decreased gradually until there were nine factors and then increased again to 10 factors while the SEC decreased continuously, as shown in Figure 5. The SEP is a good index to avoid over-fitting. Therefore, nine factors could be the optimum for modeling in this case. For the development of a PLS model for water content of hairless mouse skin, the calibration and prediction results were better with a SEC of 5.8% and a SEP of 5.9%, respectively, when log 1/R spectra were used. Figure 6a represents the scattering plot showing the correlation between the NIR predicted value and the water content using the best PLS model. Open and solid circles indicate calibration and prediction data, respectively. The calibration and prediction data showed good correlation with absolute water content of hairless mouse skin. 4968 Analytical Chemistry, Vol. 73, No. 20, October 15, 2001

Overall, the calibration of water content of hairless skin mouse was successfully performed by the newly integrated portable NIR system. Hairless Mouse SkinsScanning-Type NIR Spectrometer. To compare the performance of the developed system, conventional scanning-type NIR spectrometer that has been widely used was employed. Calibration of water content of hairless mouse skin using scanning-type NIR spectrometer was accomplished using the same method as implemented for the portable NIR system. Figure 4b shows the second-derivative NIR spectra in the 7002200-nm spectral range using the scanning-type NIR spectrometer. The log 1/R spectra were separated into two groups because of considerable baseline shift. However, we could acquire more enhanced spectral features of hairless mouse skin in the broad wavelength region from scanning-type NIR spectrometer, thanks to high resolution and a second derivatization. Water bands around 970, 1450, and 1900 nm were shown resulting from the second overtone, the first overtone of O-H stretching, and the combination of O-H stretching and deformation, respectively. The weak bands appeared clearly around 1510 and 2050 nm, which were attributed to the N-H bond of the protein of skin. The 1725-nm peaks correspond to alkyl CH groups in skin lipids and protein.

Figure 8. (a) NIR spectra of human skin using the portable NIR system. (b) First derivative NIR spectra of human skin using scanning-type NIR spectrometer. Table 2. Calibration Results of the Relative Water Content of Human Skin Using the Portable NIR System spectral range, nm 1150-1650 1150-1650 1150-1650 1350-1500 1350-1500 1350-1500

spectral treatment none 1D 2D none 1D 2D

no. factors

SEC

SEP

13 9 8 8 11 10

4.5 4.7 4.7 6.0 5.9 6.3

4.9 5.0 5.3 6.7 6.9 7.4

In the NIR range from 700 to 2200 nm, three water absorption band candidates for skin moisture could be observed. However, the absorption band around 970 nm is weak and has much more light scattering effect due to short wavelength. Moreover, it could not be valuable for the determination of water content in the

stratum corneum because of too much penetration into the inner skin. Since the detector changed at 1100 nm, from a silicon to a PbS detector, the spectral range above 1150 nm was considered for the development of a calibration model. We used the whole region, 1150-2200 nm, and the same spectral region, 1150-1650 nm, that was used for the portable system. When comparing the calibration from the 1150-2200-m range with that from 11501650 nm, the calibration results were not so much different. Thus, to compare the performance of two NIR spectrometers, only the calibration results from the 1150-1650-nm range are listed in Table 1. When second derivative spectra with the 1150-1650-nm wavelength range were used, the better calibration result was acquired with SEC of 5.1% and SEP of 5.6%. The performance of the calibration of conventional scanning-type NIR spectrometer was approximately equivalent to that of the portable NIR system. Analytical Chemistry, Vol. 73, No. 20, October 15, 2001

4969

Table 3. Calibration Results of the Relative Water Content of Human Skin Using a Scanning-Type NIR Spectrometer

Figure 9. (a) Scatter plot showing the correlation between the NIR value and the reference value for the water content of human skin using the 1150-1650-nm range (portable NIR system). (b) Scatter plot showing correlation between the NIR value and the reference value for the water content of human skin using the 1150-2200-nm range (scanning-type NIR spectrometer).

Figure 6b shows the scattering plot showing the good correlation between NIR value and water content using the best PLS model. Repeatability. The relative standard deviation was calculated by obtaining the NIR spectra of human skin 6 times. For the portable NIR system, the average value of relative standard deviation (RSD) in the region from 1150 to 1650 nm was 1.27 ( 0.04%. For the scanning-type spectrometer, the RSD in the region from 1150 to 1650 nm and in the region from 1150 to 2200 nm was 1.45 ( 0.18% and 1.69 ( 0.48%, respectively. In addition to the repeatability test, an enlarged noise spectrum was obtained by subtracting two consecutively recorded log R skin spectra from 50 scanning times. It was acquired in the 1150-1650-nm wavelength region from each NIR spectrometer. The portable NIR system gave a better stable result, thanks to the optimized fiberoptic probe for human skin, as shown in Figure 7a. The enlarged noise spectrum from the scanning-type spectrometer had more noises in the water absorption band region. The baseline was drifted in Figure 7b. It is thought to be due to scanning time and the 4970 Analytical Chemistry, Vol. 73, No. 20, October 15, 2001

spectral range, nm

spectral treatment

no. factors

SEC

SEP

1150-2200 1150-1650 1350-1500 1850-1950

1D 1D 1D 1D

8 8 5 5

3.2 3.9 5.8 4.8

4.9 5.5 5.9 5.6

nonspecified fiber-optic probe for human skin. It required ∼45 s to acquire a spectrum by scanning-type NIR spectrometer, but 2 s by the portable NIR system. It could be considered that delicate fluctuations of skin sample existed during the scanning time. However, the noise level was low for both the portable and the scanning-type NIR spectrometers and spectral treatment such as derivatives could be applied for complementary measures. Human SkinsPortable NIR System. The portable NIR system was applied for the determination of the water content, which is an important factor for healthy skin, of the stratum corneum layer of human skin. We used the conventional capacitance method for the reference value and investigated the correlation between the NIR value and the reference value. The NIR reflectance spectra were acquired from the inside of the arm of 10 persons; 20 spectra were collected from each person. Figure 8a shows the human skin spectra. Two hundred spectra were separated into two equivalent sets for the development of PLS modeling and the evaluation of the PLS models. PLS regression was performed in the same method as used for the hairless mouse skin moisture calibration. We investigated the calibration results, depending on the spectral region and spectral treatment, to find a robust model. Two regions were used. One is the 1150-1650nm region, and the other is the 1350-1500-nm region, the water absorption band. The calibration result is listed in Table 2. Similar to the result of hairless mouse skin, there are no improvements in calibration using the derivative spectra. The calibration model with log 1/R spectra was better with a SEC of 4.5 and a SEP of 4.9. Figure 9a represents the scatter plot showing good correlation between the NIR and reference values. Human SkinsScanning Type-NIR Spectrometer. The NIR spectra were acquired using the scanning-type NIR spectrometer at the same site on the inside of the arms right after using portable NIR systems. In the region of 700-2200 nm, NIR reflectance spectra were collected using fiber-optics. To enhance the spectral features, derivative techniques were used. We used log 1/R, first, and second derivatives for the calibration; first derivatives showed the better calibration results. Figure 8b shows the first derivative spectra of log 1/R. There is no considerable difference in the spectral features between the NIR spectra of hairless mouse skin and human skin. As previously mentioned in the hairless mouse section, the region below 1150 nm was not used. Four spectral regions, 1150-2200, 1150-1650, 1350-1500, and 1850-1950 nm, were investigated for finding the spectral regions for the best model. The calibration results are listed in Table 3. The same spectral region used in the portable NIR system was implemented to compare the performance of the developed portable NIR system, and the calibration using the scanning-type spectrometer could not be a better result. In the comparison of the 1350-1500nm and 1850-1950-nm ranges, the calibration results were not

significantly different, showing