We present confocal spectroscopic imaging measurements applied to in-vivo studies to determine the depth dependent hydration profiles of human skin. The observed spectroscopic signal covers the spectral range from 810 nm to 2100 nm allowing to probe relevant absorption signals that can be associated with e.g. lipid and water-absorption bands. We employ a spectrally sensitive autofocus mechanism that allows an ultrafast focusing of the measurement spot on the skin and subsequently probes the evolution of the absorption bands as a function of depth. We determine the change of the water concentration in m%. The water concentration follows a sigmoidal behavior with an increase of the water content of about 70% within 5 μm in a depth of about 14 μm. We have applied our technique to study the hydration dynamics of skin before and after treatment with different concentrations of glycerol indicating that an increase of the glycerol concentration leads to an enhanced water concentration in the stratum corneum. Moreover, in contrast to traditional corneometry we have found that the application of Aluminium Chlorohydrate has no impact to the hydration of skin.

The chemical properties of living tissue are very important to reveal the functional and biological status of tissue cells. The human skin is a unique barrier affording epithelium that constantly renews its outmost layer, the epidermis. The barrier function has been shown to reside in its outmost layer, the stratum corneum, which largely consists of proteins and lipids, which are generated during the catabolism of the epidermal cells, the keratinocytes.1 An important physical barrier is the barrier between the dry environment outside the human body and a much higher and tissue depending varying water concentration inside the human skin.1,2 External environmental factors as well as endogenous disorders impair the body’s barrier against excessive water loss and allows the invasion of exogenous noxae like chemicals or pathogenic microorganisms.2 

Irrespective of the biophysical relevance to study the water concentration in human tissues and especially human skin, there are only very few measurement techniques that allow to study the nature of the water profile in skin.3–5 Techniques such as Corneometry have no spatial resolution nor do they have any chemical sensitivity, but they have an apparent sensitivity to the change in water content.5 Confocal Raman scattering is an interesting technique, however, in the typically used spectral range it does not have an apparent band that can be unambiguously associated to water absorption.4,5 Moreover, the laser power density used in the measurement is relatively high > 300 kW/cm2 due to the small Raman cross section and the desired high lateral and spatial resolution.4 Furthermore, due to the focusing conditions of the measurements they have to be done through a glass cover plate, which can manipulate the tissue and its water barrier. Studies of the near- and mid infrared absorption bands of water allow a quantitative analysis of the water content, but have not been done as a function of depth.6 

Here, we present in-vivo confocal spectroscopic measurements in the NIR to MIR spectral region that are obtained without obscuration of the skin using an ultrafast autofocusing method under visual control and with chemical sensitivity to water and lipids.

The measurements are performed by using the KOSIM (© 4DOS GmbH) System, which consists of a real time imaging branch and a synchronized spectroscopic unit that can be outfitted to cover different spectral ranges. In the KOSIM-IR (© 4DOS GmbH) configuration the spectral range covers between 810 nm and 2100 nm. The focal size (FWHM) varies around 10 um with an optical depth resolution of 5 μm. The scanning resolution was set to 2.5 μm to oversample the FWHM optical resolution. The instrument also delivers live images from the measurement plane, which serves as additional visual control to prevent an accidental focusing on e.g. hairs and pores. The instrument exhibits several autofocusing options such as iterative and Levenberg-Marquardt methods7 to determine the focus on the tissue. Even though the coupling objective is reflective the internal optics exhibit a remaining chromatic aberration that needs to be taken into account. Furthermore, the tissue itself generates a chromatic aberration. Therefore, the autofocus can be set to the wavelength region of interest to optimize the focus. The autofocus procedure is automatically initiated at the beginning of each depth scan and can be embedded in further lateral scanning modes.

In our studies on human skin we took spectra at depth of 0 μm, 2.5 μm, 5 μm, 7.5 μm, 10 μm,15 μm, 20 μm, 30 μm, 50 μm, and 100 μm, respectively. Here, 0 μm represents the skin surface after the autofocusing algorithm and 100 μm represents the deepest measurement, where the focal point was shifted 100 μm deep into the skin. All measurements on skin were taken with an integration time of 50 ms. Based on the obtained signals and by increasing the integration time, measurements in depths of several mm would be possible. The depth profile was being repeated within 9 measurement points for any area of interest. Measurement points were 2 mm apart. Each volunteer has up to 4 different areas of interest on its fore arms and was investigated at two different time points. At time point t0 all areas were untreated. The double-blinded study design required that the different areas are encoded by a study protocol and analyzed after the measurements were taken. Typically all studies feature besides an untreated reference area, an area where a vehicle substance was used, and one or two areas where the vehicle substance with different amounts of active substance (verum) or different actives are applied. Subsequently all measurements where analyzed and the results were averaged for each area.

The analysis is shown for a t0 reference measurement in Fig. 1. Fig. 1(a) shows raw intensity spectra as a function of depth. A monotonic decrease of intensity that is amplified in certain regions around 1400 nm- 1500 nm and 1700 nm – 2000 nm due to the presence of water absorption can be identified.8 A signal from a MgF2 coated Al mirror is being used as reference. This mirror allows to correct for the transmission through the optical system and absorption of e.g. water in air. This enables us to analyze the spectra in terms of their absorption by normalizing the data with the reference mirror signal and by multiplying with the relative integration times between skin (50 ms integration time) and mirror measurements (2 ms integration time). The reflected light is suppressed in case a particular absorption occurs at specific wavelengths. Thus, an absorption peak shows up when one plots the corrected 1-R spectra as shown in Fig. 1 (b). One can clearly observe the basic water absorptions that would be expected in this wavelength range, which also change as function of depth. Most pronounced are the two water absorption bands close to 1950 nm and 1450 nm.8,9 The bands at 1750 nm and 1150 nm can be attributed to lipids (CH absorption).10 Some measurements where automatically excluded, such as non-monotonic absorption profiles that violate Lambert-Beers Law of absorption. Also measurements with no signs of absorption were neglected. These results are typically induced by a movement of the volunteer during the measurement or by very rough skin generating problems in the autofocusing procedure due to enhanced stray-light. Moreover, to track the focusing parallel live imaging of the measurement can be used as well.

FIG. 1.

Typical measurement data at t0 of skin in-vivo. In (a) we show the raw data as a function of depth (10 depth points) from 0 μm (surface of the skin - bold red curve) to 100 μm depth (bold blue curve). Also shown in black is a reference signal taken from an Al-mirror for calibration purposes. In (b) we show the absorption calculated as 1-R. One can observe the evolution of 4 different bands as a function of depth (color coded as in (a)) related to lipids and water. (c) Fitted spectra (Fit: black line) as described in the text. The inset shows the evolution of the intensity of the peak at 0.865 eV as function of depth. For clarity we do not show the evolution of the band at 0.635 eV as they show essentially identical results. (d) Calculated changes of the water concentration in m%. The fit (black line) corresponds to the model of the skin water barrier as described in the text.

FIG. 1.

Typical measurement data at t0 of skin in-vivo. In (a) we show the raw data as a function of depth (10 depth points) from 0 μm (surface of the skin - bold red curve) to 100 μm depth (bold blue curve). Also shown in black is a reference signal taken from an Al-mirror for calibration purposes. In (b) we show the absorption calculated as 1-R. One can observe the evolution of 4 different bands as a function of depth (color coded as in (a)) related to lipids and water. (c) Fitted spectra (Fit: black line) as described in the text. The inset shows the evolution of the intensity of the peak at 0.865 eV as function of depth. For clarity we do not show the evolution of the band at 0.635 eV as they show essentially identical results. (d) Calculated changes of the water concentration in m%. The fit (black line) corresponds to the model of the skin water barrier as described in the text.

Close modal

In order to model the observed absorption bands a superposition of Lorentz Oscillators is used. We can define the fit function as

I(z)=y0+iIiωΓi  z2ωi22+ω2Γi2,
(1)

ωi, Γi and Ii correspond to the respective frequency of the absorption band, its full width half maximum, and its strength. There are two important observations when fitting the spectra as shown in Fig. 1(c). First, the absorption band at 0.634 eV consists out of two bands centered at 0.6315 ± 0.00135 eV and 0.6544 ± 0.00138 eV. Secondly, the parameters of all used frequencies and widths are independent on the depth. Thus, they can be fixed and only the intensities are varied as a function of depth. A summary of the relevant fixed parameters, their errors, and their assignments is given in Table I. The inset in Fig. 1(c) shows the dependence of the intensity of the absorption band as a function of depth. The absorption shows a variation of the gradient between different points that can be related directly to Lambert Beers Law

I(Δz)=I0expηcΔz.
(2)
TABLE I.

Fit parameters of the absorption spectra.

Mode AssignmentFrequency (eV)Width (eV)
Water (Overtone) 0.865 ± 0.0006 0.128 ± 0.0110 
Water 0.631 ± 0.00135 0.0789 ± 0.00676 
Water 0.654 ± 0.00138 0.0138 ± 0.00166 
Lipids 0.729 ± 0.00141 0.152 ± 0.0228 
Lipids 1.07 ± 0.00822 0.253 ± 0.0644 
Mode AssignmentFrequency (eV)Width (eV)
Water (Overtone) 0.865 ± 0.0006 0.128 ± 0.0110 
Water 0.631 ± 0.00135 0.0789 ± 0.00676 
Water 0.654 ± 0.00138 0.0138 ± 0.00166 
Lipids 0.729 ± 0.00141 0.152 ± 0.0228 
Lipids 1.07 ± 0.00822 0.253 ± 0.0644 

η is the absorption coefficient in μm-1 c is the water concentration in m% and Δz is the differential distance between two measurement points of the profile. This implies that the absorption strength is 1 η at 100 m% and 0.5 η at e.g. 50 m%, where m% is water in mass percent. This treatment assumes that water is the dominant absorbing substance in the spectral range of interest. For simplicity we concentrate in the following discussion on the band at 0.865 eV where η is 0.0068 μm-1.9 Furthermore, the effective distance seen by the instrument is also determined by its optical confocal resolution along the z-axis. Since the nominal step width is of the same order of magnitude the effective travel of the photons along the Δz axis is given by:

Δz=zstep2+zoptics2.
(3)

By inverting equation (2) towards the water concentration c and using Δz according to equation (3), one obtains the change in water concentration from the differences of the gradients between the intensity points the change in water concentration. The result is shown in Fig. 1(d). The error bars in the m% values result from the fit errors of the intensities. In Fig. 1(d) we can clearly observe over the first couple of measurement points a nearly constant water concentration. At about 15 to 20 μm depth a sudden step in the water concentration can be seen. This jump occurs exactly in a region with large gradient in the inset of Fig 1(c) as expected from Eqn. (2). Below 20 μm a nearly constant slope of the water concentration within the error bars is being observed. The behavior of the water concentration can be approximated by a sigmoidal function

m%(z)=m%max1+exp(m%slope(m%stepz))+m%offset,
(4)

where m%max is the maximum amplitude of the jump in the water concentration, m%step is the depth position where the jump in the water concentration occurs, i.e. water barrier of human skin is, m%slope controls how steep the jump is, and m%offset represents the remaining water concentration of the outer dead skin cells.

The fit yields for this particular measurement m%step = 14.5 ± 3 μm, m%max = 71.8 ± 16.6, m%slope = 0.197 ± 0.11 μm-1, and m%offset = 4.41 ± 13. This indicates that the water barrier in human skin facilitates an increase of the water concentration of about 70 % at a depth of about 14 μm within 5 μm.

Surely, it is not valid to derive general skin properties from a single measurement. The above mentioned analysis simply illustrates the general approach to measure and determine the hydration dynamics of skin in vivo. In order to test and benchmark our technique we have conducted two different studies that we briefly outline in the following.

In the first study, 33 volunteers applied emulsions with varying Glycerol concentrations to four skin areas on the forearm. At time point t0 all four were untreated and at t1 the test persons had applied 2 weeks twice a day to area 1 – nothing (untreated reference), area 10 - 7% Glycerol, area 20 - 2% Glycerol, and area 30 - 0% (only vehicle) to the skin. Overall the study comprised the analyses of 7128 profiles and 71280 spectra, respectively. The established Corneometer served as benchmark. The Corneometer exhibits no chemical sensitivity but can track changes in the water content.11 The measurement delivers arbitrary units. Nevertheless, there are a number of studies that clearly show an enhanced humidity level of human skin under the protocol outlined above.12 The resulting water profiles were subject to a statistical analysis employing a non-parametric Wilcoxon using Statistica from Statsoft.13,14 The results are shown in Fig. 2.

FIG. 2.

Glycerol study at two time points t0 (without treatment) and t1 (with treatment) – see text for details. (a) Shows the KOISM-IR results at a depth of 7,5 μm. While at t0 all investigated areas show no trend indicating a water concentration of about 5% prior to the treatment, the t1 results show a clear trend. The untreated reference area (1) shows no change whereas the Areas 10 and 20 show a significant increase in water concentration of up to 30% and 15 % respectively. Area 10 corresponds to 7% Glycerol and Area 20 to 2% Glycerol. Area 30 represents the vehicle used to embed the Glycerol for treatment. This situation is nearly identical in (b) at a depth of 10 μm. In (c) we show as reference the established Corneometry measurement. Though the measurement essentially shows a similar trend the gradual changes are clearly more pronounced in the KOSIM-IR data sets.

FIG. 2.

Glycerol study at two time points t0 (without treatment) and t1 (with treatment) – see text for details. (a) Shows the KOISM-IR results at a depth of 7,5 μm. While at t0 all investigated areas show no trend indicating a water concentration of about 5% prior to the treatment, the t1 results show a clear trend. The untreated reference area (1) shows no change whereas the Areas 10 and 20 show a significant increase in water concentration of up to 30% and 15 % respectively. Area 10 corresponds to 7% Glycerol and Area 20 to 2% Glycerol. Area 30 represents the vehicle used to embed the Glycerol for treatment. This situation is nearly identical in (b) at a depth of 10 μm. In (c) we show as reference the established Corneometry measurement. Though the measurement essentially shows a similar trend the gradual changes are clearly more pronounced in the KOSIM-IR data sets.

Close modal

As expected from Fig. 2 the KOSIM IR measurements track a significant change in the water concentration at a depth of 7 μm and 10 μm, i.e. just at and above the water barrier shown in Fig. 2(c). As shown in Fig. 2(a) and Fig. 2(b) one obtains no statistical difference at t0 between the four different regions before treatment. After treatment, however, at t1 we find again no change in the untreated reference area and the area receiving only the vehicle, but significant changes when Glycerol is applied. With the application of 7% Glycerol the water concentration jumps about 30 % and with 2% Glycerol about 15%. This trend is confirmed also in Fig. 2(c) where the results from the Corneometer measurements are shown. From the comparative measurements it is evident that the depth resolution of the KOSIM IR generates a much clearer gradient between the measurements compared to the Corneometer that averages over an unknown depth region.15 

Furthermore, due to its lacking chemical sensitivity for some ingredients the Corneometer can lead to misleading results. As an example some treatment remains at the skin surface tend to change the Corneometer results. In order to study this topic we have applied the KOSIM IR to a second comparative test.

Aluminium Chlorohydrate (ACH) has been used for decades as key active in cosmetic products such antiperspirants. In a second study 25 persons participated with two skin regions. At t0 no treatment occurred and at t1 one region was subject to 35% ACH and the second region was subject to a 0% ACH i.e. only vehicle. Overall 10800 profiles and 108000 spectra were analyzed for this study. The results are shown in Fig. 3. At t0 the different measurement areas show no difference in neither the KOSIM IR nor the Corneometer. However, at t1 the ACH treated region exhibited a reduction of the Corneometer units, which is then in analogy to the Glycerol study interpreted as a reduced moisture level of the skin. However, ACH generates treatment residues on the skin that can potentially affect this finding. Clarys at al. have shown how different chemicals can influence the Corneometer readings.16 Indeed, when analyzing the water absorption bands in the skin directly as being accomplished by the KOSIM-IR no change in the water concentration neither at 7.5 μm nor at 10 μm was found.

FIG. 3.

ACH Study at two time points t0 (without treatment) and t1 (with treatment) – see text for details. (a) and (b) show the KOISM-IR results at a depth of 7,5 μm and 10 μm. At t0 all investigated areas show no trend indicating a water concentration of about 10 % prior to the treatment, the t1 results shows also no clear trend with a nearly constant water concentration. Area 10 corresponds to 10 % ACH and Area 20 to the vehicle. In (c) we show as reference the established Corneometry measurement. These data seem to suggest a reduction in the water concentration. However, the Corneometer has no chemical sensitivity and, thus, measures the residues on top of the skin surface.

FIG. 3.

ACH Study at two time points t0 (without treatment) and t1 (with treatment) – see text for details. (a) and (b) show the KOISM-IR results at a depth of 7,5 μm and 10 μm. At t0 all investigated areas show no trend indicating a water concentration of about 10 % prior to the treatment, the t1 results shows also no clear trend with a nearly constant water concentration. Area 10 corresponds to 10 % ACH and Area 20 to the vehicle. In (c) we show as reference the established Corneometry measurement. These data seem to suggest a reduction in the water concentration. However, the Corneometer has no chemical sensitivity and, thus, measures the residues on top of the skin surface.

Close modal

We present confocal spectroscopic imaging measurements to specifically access the hydration dynamics of human skin in vivo. From the evolution of the absorption bands we derive the change of the water concentration in m%. We have applied our technique to study the hydration dynamics of human skin after treatment with different concentrations of glycerol. We found that an increase of the glycerol concentration leads to proportionally enhanced water concentration gradients in the stratum corneum. The sensitivity of the method allowed the generation of a clear dose response curve for glycerin. Furthermore, we showed that Kosim IR can be employed in areas where the dielectric based readings of the traditional Corneometer reach the technical limits. Thus, we could reject the assumption that the application of ACH to human skin leads to a reduction of the water concentration in the skin. We presume that the spectrally modular design of our technique lends itself to a variety of further applications, which will be explored in the future.

We thank Maren Meyer, Birthe Körbl, and Volker Schreiner for valuable discussions. This study was funded by Beiersdorf AG.

1.
K.-P. W.
Maibach
, “
Transepidermal water loss and barrier function of aging human skin
,” in
Bioengineering of the Skin: Water and the Stratum Corneum
. (
E. B. P.
Elsner
, Editor)
Florida, USA
:
CRC Press
(
1994
).
2.
M. L.
Forslind
,
The Skin as a Barrier
(
M. L.
Maibach
, Editor)
Florida USA
:
CRC Press
.
3.
P. J.
Caspers
,
G. W.
Lucassen
,
R.
Wolthuis
,
H. A.
Brunning
, and
G. J.
Puppels
, “
In vitro and in vivo Raman spectroscopy of human skin
,”
Biospectroscopy
4
,
31
39
(
1998
).
4.
P. J.
Caspers
,
G. W.
Lucassen
,
H. A.
Brunning
, and
G. J.
Puppels
, “
Automated depth scanning confocal Raman microspectrometer for rapid in vivo determination of water concentration profiles in human skin
,”
J. Raman Spectrosc.
31
,
813
818
(
2000
).
5.
W.
Courage
, “
Hardware and measuring principle corneometer
,” in
Bioengineering of the Skin: Water and the Stratum Corneum
. (
E. B. P.
Elsner
, Editor)
Florida, USA
:
CRC Press
(
1994
).
6.
P. A.
Philippe Humbert
,
Measuring the Skin
(
Springer
,
Berlin Heidelberg
,
2004
).
7.
D.
Marquardt
, “
An algorithm for least-squares estimation of nonlinear parameters
,”
Journal of the Society for Industrial and Applied Mathematics
11
,
431
441
(
1963
).
8.
H.
Arimoto
,
M.
Egawa
, and
Y.
Yamada
, “
Depth profile of diffuse reflectance near-infrared spectroscopy for measurement of water content in skin
,”
Skin Research and Technology
11
(
1
),
27
35
(
2006
).
9.
K. P.
Williams
, “
Optical properties of water in the near infrared
,”
Journal of the Optical Society of America
64
,
1107
1110
(
1974
).
10.
G. B.
Altshuler
,
R. R.
Anderson
, and
D.
Manstein
, Patentnr. States Patent, Patent No. 6,605,080 B1. USA (
2003
).
11.
U.
Heinrich
,
U.
Koop
,
M.-C.
Leneveu-Duchemin
,
K.
Osterrieder
,
S.
Bielfeldt
,
C.
Chkarnat
,
J.
Degwert
,
D.
Häntschel
,
S.
Jaspers
,
H.-P.
Nissen
,
M.
Rohr
,
G.
Schneider
, and
H.
Tronnier
, “
Multicentre comparison of skin hydration in terms of physical-, physiological- and product-dependent parameters by the capacitive method (corneometer CM 825)
,”
International Journal of Cosmetic Science
25
,
45
-
53
(
2003
).
12.
J.
Polaskova
,
J.
Pavlackova
, and
P.
Egner
, “
Effect of vehicle in the performance of active moisturizing substances
,”
Skin Research & Technology
21
(
4
),
403
-
412
(
2015
).
13.
L.
Sachs
,
Angewandte Statistik
(
Springer
,
Berlin Heidelberg
,
2002
).
14.
Statsoft
. www.statsoft.de (
2017
).
15.
E.
Berardesca
, “
EEMCO guidance for the assesment of stratum corneum hydration: electrical methods
,”
Skin Research and Technology
3
,
126
132
(
1997
).
16.
P.
Clarys
,
R.
Clijsen
,
J.
Taeymans
, and
A. O.
Barel
, “
Hydration measurement of the stratum corneum: Comparison between the capacitance method and the impedance method
,”
Skin Research and Technology
18
,
316
323
(
2012
).