1.

Introduction

Near-infrared spectroscopy (NIRS) was introduced more than $25\text{years}$ ago¹ and this technique has been used in many physiological and clinical studies. Applications were found in monitoring of oxygenation in muscle,² and adult^{3, 4, 5} and neonatal^{6, 7} brain. A further application is the estimation of cerebral blood volume (CBV). Wyatt ⁸ combined NIRS assessment of absolute changes of concentration of oxy- and deoxyhemoglobin ( $\Delta c{\mathrm{O}}_2\mathrm{Hb}$ and $\Delta c\mathrm{H}\mathrm{Hb}$ , respectively) with measurement of arterial oxygen saturation $\left(\mathrm{Sa}{\mathrm{O}}_2\right)$ to calculate the CBV in newborn infants. Cerebral blood flow (CBF) can be measured with NIRS using a bolus of $c{\mathrm{O}}_2\mathrm{Hb}$ as a tracer.^{9, 10} Other groups performed NIRS measurement of absorption changes after intravenous injection of a bolus of indocyanine green (ICG) to obtain CBV¹¹ and CBF.¹²

The question has been raised to what extent superficial-cerebral measurement of cerebral Hb-concentrations is influenced by absorption and scattering in external tissue layers, i.e., skin, skull, dura, and cerebro spinal fluid.

Hiraoka ¹³ performed a simulation study with a layered model of the head. They concluded that for piglets, the contribution of superficial (extra cerebral) layers to the effective path length ranged from 60 to 25% when changing the optode separation angle from $30\text{to}75\mathrm{deg}$ .

Okada, Firbank, and Delpy¹⁴ concluded from a simulation study that in the adult head, NIRS measures mainly oxygenation changes in the superficial brain layers, i.e., within the gray matter, at optode separations between 30 and $90\mathrm{deg}$ . The base for this conclusion was the similar $\mu$ values (transport scattering coefficient) in scalp and skull as well as in gray matter.

There is clear evidence of an “optical channeling” effect by extra cerebral layers in the adult human.¹⁵ These authors found a significant decrease of light passage, at optode separation of $4\mathrm{cm}$ , if skin, skull, and dura/cerebrospinal fluid (CSF) were removed. The light attenuation is defined as the negative natural logarithm of the light transmission: $-\ln \left(T\right)$ . They found an apparent light attenuation of 0.30, 1.35, and 1.51 OD, respectively, in successive removal of these layers for one optode. These results might indicate that only 3% of the optical path passes through the cerebral cortex and 97% through the superficial layers. This conclusion is supported by the simulation studies of Okada ^{16, 17}

A somewhat different approach for assessing the influence of superficial layers on NIRS measurement was followed by Smielewski ^{18, 19} These authors made a simultaneous measurement of the blood flow velocity in the middle cerebral artery with ultrasound transcranial Doppler (TCD), and of the skin blood flow velocity with laser Doppler flowmetry (LDF). They concluded that NIRS-monitored cerebral oxygenation changes during $\mathrm{C}{\mathrm{O}}_2$ reactivity tests were statistically not significantly correlated with skin perfusion. On the other hand, high correlations were observed between NIRS $c{\mathrm{O}}_2\mathrm{Hb}$ concentration changes and end-tidal $\mathrm{C}{\mathrm{O}}_2$ and arterial blood flow velocity. Finally, by compression of the superficial temporal artery, it could be concluded that the skin blood flow contribution to the $c{\mathrm{O}}_2\mathrm{Hb}$ reactivity measured with NIRS was approximately 16%.

Although this latter result might lower the concerns about the influence of skin on NIRS measurements of Hb concentration changes in the cerebral cortex, it is still not clear how far the large channeling effects are producing a bias, and whether these effects are wavelength dependent or not. A further question we need to answer is: what contribution do these effects have on the bias and reproducibility of NIRS measurements of Hb concentration changes? A preliminary report about the present study, in which a correction for the skin on concentration changes was investigated, has been published.²⁰

The aim of this animal study was to investigate the influence of the skin on regional cerebral oxygenation measurements. Furthermore, the inter- and intra-individual variability of the measurements was investigated.

2.

Near-Infrared Spectroscopy

The principle of near-infrared spectroscopy (NIRS) was first described in 1977.¹ NIRS in tissue is based on measurement of light attenuation caused by absorption and scattering. Tissues are transparent only in a small wavelength range $\left(700\text{to}1000\mathrm{nm}\right)$ . In this near-infrared region, penetration depths up to $8\mathrm{cm}$ can be achieved.²¹

The absorption mechanism can be separated in two parts: one due to chromophores with a constant tissue concentration, and one due to chromophores with a variable concentration. In cerebral tissue, the main varying components are the concentrations of oxy- and deoxyhemoglobin ( $c{\mathrm{O}}_2\mathrm{Hb}$ and $c\mathrm{H}\mathrm{Hb}$ ) hemoglobin. The $c{\mathrm{O}}_2\mathrm{Hb}$ and $c\mathrm{H}\mathrm{Hb}$ in cerebral tissue depend on the hemoglobin oxygen saturation and the cerebral blood volume (CBV). Other absorbers are assumed to be independent of hemoglobin oxygen saturation.

When there is no scattering, the absorption can be described by the Lambert-Beer law.

Scattering effects were incorporated in Eq. 1 by introducing an extra pathlength factor, which results in the modified Lambert-Beer equation.^{22, 23, 24, 25} This pathlength factor incorporates the scattering contribution to the overall pathlength of the photons traveling from source to detector. Furthermore, an extra term $R$ was introduced to incorporate oxygenation independent absorption and scattering losses. In matrix notation, Eq. 1 now becomes:

Several research groups have measured molar extinction coefficients. In this work, the modified Keele matrix^{22, 26} was used. The photons arriving at the receiving optode have mainly traveled through a banana-shaped volume spanning from the transmitting to the receiving optode. The size and depth range of this volume can be manipulated by the positioning of the optodes.^{27, 28, 29}

When comparing two different conditions of the medium, the change of attenuation (i.e., $\Delta \overline{A}$ ) can be related to changes in the concentration of a chromophore (or in the extinction coefficient):

In this article, we calculate the concentration change $\left(å\right)$ with respect to the $\mathrm{Sa}{\mathrm{O}}_2=100\%$ level:

The effect of the skin removal was calculated by taking the difference $\left(d\right)$ between these concentration changes measured without and with skin (skin condition 2: area 1 minus area 2). The difference was calculated only for the simultaneous measurements to avoid errors caused by small differences in the physiological condition of the animals. In this case, the DPF of both conditions might be different [Eq. 5].

Since the authors do not have any data about the effect of skin removal on the DPF, it was necessary to take a fixed value of DPF (4.2) into account. For this reason, any DPF changes become apparent as changes of the attenuation and are translated into apparent changes of concentration.

3.

Materials and Methods

3.2.

Design of the Experiment

During the experiments, the arterial oxygen saturation $\left(\mathrm{Sa}{\mathrm{O}}_2\right)$ was changed from normoxia $(\mathrm{Sa}{\mathrm{O}}_2=100\%)$ , to mild hypoxia $(\mathrm{Sa}{\mathrm{O}}_2=80\%)$ , to severe hypoxia $(\mathrm{Sa}{\mathrm{O}}_2=60\%)$ , and back to normoxia. This was achieved by lowering the $\mathrm{Fi}{\mathrm{O}}_2$ by adding ${\mathrm{N}}_2$ to the inspired gasses. The hypoxia conditions were maintained for at least $10\min$ to reach a stable condition. At the end of each stabilizing period an arterial blood sample was taken.

The NIRS instrument used (OYMON) in the experiments was developed at the Instrumentation Department of the Radboud University Medical Centre Nijmegen (Nijmegen, The Netherlands).³² The equipment contains laser diodes operated in pulsed mode (pulse length $100\mathrm{ns}$ ) at three wavelengths (767, 850, and $905\mathrm{nm}$ ) and a receiver equipped with an avalanche photodiode. Pulse repetition rate was $1\mathrm{kHz}$ per wavelength and the processed signal consisted of the consecutive means over 100 pulses, resulting in an effective $10\text{-}\mathrm{Hz}$ sampling rate for each wavelength. All the NIRS data were continuously stored on a PC for off-line analysis. The estimated optical densities (OD) correspond to the light absorption between the transmitting and receiving fibers. After changing the $\mathrm{Sa}{\mathrm{O}}_2$ levels, we waited until a new steady state was established and one minute periods of stable OD signals were selected for further analysis.

After the measured ODs were averaged over a period of one minute, the difference $\left(\Delta \right)$ between the concentrations at $\mathrm{Sa}{\mathrm{O}}_2$ levels 80 and 60% and those obtained at $\mathrm{Sa}{\mathrm{O}}_2$ of 100% were calculated by using Eq. 4.

The optodes were attached to specially constructed holders that were firmly glued with collodion, a glue commonly used to attach EEG electrodes, directly on the skin or skull of the head of the piglet.³³ Receiving optodes were placed over the left $(\text{area}=1)$ and right $(\text{area}=2)$ hemisphere ( ${\mathrm{C}}_3$ , ${\mathrm{C}}_4$ EEG placement code) and the transmitting optode on ${\mathrm{C}}_z$ position (Fig. 1 ). In this optode configuration, we always have two simultaneous measurements, i.e., over the left and right hemisphere. Optical densities (OD) were measured during stable normo-, mild-, and severe hypoxemia ( $\mathrm{Sa}{\mathrm{O}}_2=100$ , 80, and 60%) and the average ODs were calculated over a one-minute period, which was interactively selected during off-line analysis of the NIRS data. To study the reproducibility, the measurements were repeated after the receiving optodes were switched from position ( ${\mathrm{C}}_3{\mathrm{C}}_4$ to ${\mathrm{C}}_4{\mathrm{C}}_3$ ), and after they were placed back at the starting position. For the three $\mathrm{Sa}{\mathrm{O}}_2$ levels these measurements were repeated, and after returning to normoxia, two control measurements were done. This whole procedure was repeated for each of the three skin conditions (see also Fig. 2 ): all optodes with skin (condition 1, skin-skin); one receiving optode directly on the skull, second receiving optode and transmitting optode on the skin (condition 2); transmitting and receiving optodes on the skull (condition 3, skull-skull). A total of 196 measurements were performed (3 $\text{piglets}\times 3$ skin $\text{conditions}\times 3$ $\mathrm{Sa}{\mathrm{O}}_2$ $\text{levels}\times 3$ repeated $\text{measurements}\times 2$ simultaneous $\text{areas}+3$ $\text{piglets}\times 3$ skin $\text{conditions}\times 2$ repeated $\text{measurements}\times 2$ simultaneous areas).

Fig. 1

Scheme of the optode positions on the piglet head. The measurements were done over two areas: area $1={\mathrm{C}}_z-{\mathrm{C}}_3$ and area $2={\mathrm{C}}_z-{\mathrm{C}}_4$ . In the present experiments, the skin was only removed directly under the fiber connectors ( $\sim 1\mathrm{cm}$ diam) and the skull remained intact.

Fig. 2

Experimental design. Condition 1: all optodes on skin; condition 2: receiver 1 on skull, transmitting optode on skin; and condition 3: all optodes on skull. In all three conditions, the $\mathrm{Sa}{\mathrm{O}}_2$ was changed from 100 to 80 to 60 and back to 100%. The measurements were repeated three times while the receiving optodes were simultaneously exchanged.

From the ODs at the three wavelengths (767, 850, and $905\mathrm{nm}$ ), the absolute concentration changes $\Delta c{\mathrm{O}}_2\mathrm{Hb}$ , $\Delta c\mathrm{H}\mathrm{Hb}$ , and $\Delta c\mathrm{t}\mathrm{Hb}$ $(\mu \mathrm{mol}∕\mathrm{L})$ were calculated using the modified Lambert-Beer law [Eq. 3].

In our experiments, the distance between the transmitter and each receiver optode was $1.8\mathrm{cm}$ . An estimate of the effective optical pathlength was incorporated by using an average pathlength factor of 4.2, which was estimated for infants.^{24, 34}

4.

Results

The average optical density measured in our experiments was 6 to 7 OD. The reproducibility of repeated density measurements (intraindividual spread) was approximately 0.01 OD (standard deviation, SD), whereas the interindividual spread (between areas, between animals) was around 0.3 OD (SD). When expressed as variability of total hemoglobin concentration changes, these data are $1.5\mu \mathrm{mol}∕\mathrm{L}$ intra- and $4.2\mu \mathrm{mol}∕\mathrm{L}$ interindividual spread.

The optical density change for all three wavelengths between the conditions without and with skin when averaged over the piglets appeared for all arterial oxygen saturation levels nearly equal: removal of one skin part (i.e., for skin condition 2 versus 1) gave an increased optical density change of 0.51 OD ( $\mathrm{SD}=0.28$ , SEM: 0.03, $N=93$ ). Two measurements were excluded because the OXYMON ran out of range.

The mean values of the concentration changes relative to the $\mathrm{Sa}{\mathrm{O}}_2=100\%$ level for the three skin conditions are presented together with the standard error of the mean (SEM) in Table 1 . When the skin was removed (skin conditions 2 and 3), an increased change in $c{\mathrm{O}}_2\mathrm{Hb}$ and $c\mathrm{H}\mathrm{Hb}$ was found compared to skin condition 1; the effect on the change in $c\mathrm{t}\mathrm{Hb}$ was nearly negligible.

Table 1

Mean (SEM, standard error of the mean) concentration changes caused by arterial oxygen saturation changes from SaO2=100 to 80% and from 100 to 60% level. Skin conditions are noted here as receiver (R) and transmitter (T) on the skin or the skull. The concentration differences (Δc) were averaged over the repeated measurements at the different SaO2 levels and over the different animals (n=3) .

The effect of skin removal on the concentration changes is shown in Table 2 . Here, the difference between the concentration changes relative to $\mathrm{Sa}{\mathrm{O}}_2=100\%$ was taken under skin condition 2 for area 1 minus area 2. Only simultaneous measurements are taken to avoid errors caused by differences in physiological condition. To eliminate the difference in area, the same was done for skin condition 1 (both with skin) and the averages for each arterial oxygen saturation level and for each animal were subtracted from the result under skin condition 2. It appears that the observed increase in optical density following skin removal results in an apparent increase of both the $c{\mathrm{O}}_2\mathrm{Hb}$ and $c\mathrm{H}\mathrm{Hb}$ changes of approximately 23%.

Table 2

The Mean (SEM, n=9 ) concentration changes caused by removal of the skin. First, the change in concentration with respect to the SaO2=100% level was calculated, then the difference between the simultaneous measurements at area=1 (receiver on skull) and area=2 (receiver on skin) under skin condition 2 was calculated. The data were corrected for the area and averaged over the repeated measurements (n=3) at the different SaO2 levels and over the different animals (n=3) .

The effect of skin removal on the concentration changes is further illustrated in Fig. 3 . In this picture, the concentration changes of $c{\mathrm{O}}_2\mathrm{Hb}$ and $c\mathrm{H}\mathrm{Hb}$ at $\mathrm{Sa}{\mathrm{O}}_2$ levels of 60 and 80% with respect to those obtained at the $\mathrm{Sa}{\mathrm{O}}_2=100\%$ level obtained for area 1 (C3 to $\mathrm{C}z$ ) are plotted versus those of area 2 (C4 to $\mathrm{C}z$ ). We compared the different areas with both optodes on the skin [Fig. 3(a)] and both optodes on the skull [Fig. 3(b)]. The relation between concentrations as obtained for the left and right hemisphere is well described by a linear fit ( $R^2=0.99$ and 0.94) with a slope of nearly one (1.03 and 0.99). This shows that the observations of the changes in $c{\mathrm{O}}_2\mathrm{Hb}$ and $c\mathrm{H}\mathrm{Hb}$ over both areas are nearly identical.

Fig. 3

Hemoglobin concentration changes due to changes of $\mathrm{Sa}{\mathrm{O}}_2$ level; simultaneously measured data from left hemisphere plotted versus those of right hemisphere. (a) Optodes of both hemispheres on intact skin (skin condition 1). (b) Optodes of both hemispheres on skull (skin condition 3). (c) Receiving optode over right hemisphere on skull, transmitting and receiving optode on the left hemisphere on skin (skin condition 2). The area correction was applied to skin conditions 2 and 3 by subtracting the left-right difference at skin condition 1 $\left(d\Delta \right)$ at each arterial oxygen saturation level and for each piglet.

In Fig. 3(c) the effects of skin removal at one hemisphere are shown. Again, a high correlation is found $(R^2=0.89)$ but the slope constant equals 0.85 in this case. This implies that skin removal produces an (apparent) increase by 15% of concentration changes induced by changing the $\mathrm{Sa}{\mathrm{O}}_2$ level.

In Fig. 4 the concentration changes due to skin removal are plotted as the difference with respect to the $\mathrm{Sa}{\mathrm{O}}_2$ 100% level versus the $\mathrm{Sa}{\mathrm{O}}_2$ level. A practically linear relation is found between $\mathrm{Sa}{\mathrm{O}}_2$ and $c{\mathrm{O}}_2\mathrm{Hb}$ (decrease) and $c\mathrm{H}\mathrm{Hb}$ (increase). As was observed already in Table 2, the effect of skin removal on $c\mathrm{t}\mathrm{Hb}$ is practically negligible and independent of the $\mathrm{Sa}{\mathrm{O}}_2$ level.

Fig. 4

The difference between receiver without and with skin (skin condition 2) of the concentration changes as a result of arterial oxygen saturation change from 100 to 80 and to 60%. Concentrations were averaged over the repeated measurements and animals at each arterial oxygen saturation level.

5.

Discussion

The observations in Sec. 4 show that the optical density increased for all wavelengths with a mean of 0.51 OD, for single optode skin removal. This corresponds to an increase of the light attenuation by a factor of 1.67. This OD increase is caused by changes in boundary conditions (skin removal and different light path through skull-CSF layer-brain tissue) and by changes in absorption. Many simulation studies about the influence of superficial layers have been done,^{16, 17, 35, 36} but these concentrate on the effect of skull and CSF layer. To our knowledge these authors did not look exclusively at the influence of the skin. A finding of increased absorption similar to the one presented in Sec. 4 was published by Young ¹⁵ These authors found upon skin removal an increase of the overall attenuation by a factor of 0.5, i.e., a density increase of 0.69 OD. They ascribed this apparent “attenuation” increase by the removal of light tunneling effects within the skin and in subcutaneous layers. The data presented in our study (i.e., 0.51 OD) are of the same order of magnitude.

In Table 1 it is shown that removal of the skin (under one or two optodes) yielded an increase in magnitude of the concentration changes of $\Delta c{\mathrm{O}}_2\mathrm{Hb}$ and $\Delta c\mathrm{H}\mathrm{Hb}$ (with respect to the $\mathrm{Sa}{\mathrm{O}}_2=100\%$ level), and the $\Delta c\mathrm{t}\mathrm{Hb}$ remained nearly constant. The fact that the $\Delta c\mathrm{t}\mathrm{Hb}$ did not change shows that the $\mathrm{Sa}{\mathrm{O}}_2$ step from 100 to 60% did not cause a change in the blood content in the examined area.

The calculation of the absolute change of hemoglobin concentrations due to manipulation of the arterial oxygen saturation is done for identical optode positions and skin conditions. In this case, $R$ and other oxygenation-independent absorption terms cancel out in the equations [Eq. 3]. Nevertheless, the effect of skin removal on DPF remains in the equation and this is actually being measured [Eq. 4]. However, since a change of DPF cannot be estimated by cw-NIRS, it will be interpreted as an apparent change of concentration.

During hypoxia, the $\mathrm{Pa}\mathrm{C}{\mathrm{O}}_2$ level was measured by arterial blood gas analysis and the variations were small (less than 5%). There was no correlation found between $\mathrm{Pa}\mathrm{C}{\mathrm{O}}_2$ and $\mathrm{Sa}{\mathrm{O}}_2$ . Along the line of thought in the preceding paragraph, the present results are most probably due to an increase of the amount of light passing through the cortex of the brain, i.e., more tissue path containing hemoglobin due to a reduction of the tunneling effect, and therefore a higher absorption. In Eq. 5 a distinction was made between the DPF with and without skin. When the skin is removed, the optode distance stays constant but the effective light path, i.e., the DPF, might have changed. The light probably will go deeper into the brain tissue, or alternatively the tunneling effect might have changed. This will be translated into an apparent change in concentrations. The $d\Delta$ concentration effects might be caused by changes in the absorption and by changes in the optical pathlength, the latter of which is related to an increase in scattering and possible reduction of the tunneling effect. The DPF was in our calculations kept constant; therefore, the calculated concentration changes $d\Delta$ will be overestimated.

In Table 2 the difference between simultaneous measurements with the receiver on the skin and on the skull is taken (skin condition 2, i.e., left brain versus right brain). We see that the $d\Delta c{\mathrm{O}}_2\mathrm{Hb}$ decreases and the $d\Delta c\mathrm{H}\mathrm{Hb}$ increases when the arterial oxygen saturation decreases. This supports the assumption of an increased amount of blood in the increased light path when the skin is removed. The $d\Delta c\mathrm{t}\mathrm{Hb}$ remains nearly zero because $\Delta \mathrm{t}\mathrm{Hb}$ nearly did not change either.

In Table 1 we look at the effects of skin condition on the concentration changes caused by arterial saturation changes. We see that removal of one skin part gives a large difference in concentration changes. But removal of two skin parts makes nearly no additional difference on the concentration changes. This might be explained by the change in tunneling effect. Removing one skin part may stop the tunneling at the skin-skull layer, so the second skin removal will not change the tunneling but it will only change the light path through the tissue.

From Tables 1, 2 we see that removal of the skin gave an apparent increase of the concentration changes of 23%, which is nearly equal to the result of 15% shown in Fig. 3(c). This implies that the oxygenation changes due to changes of (physiological) conditions, which are measured with NIRS, can have 15 to 23% influence of the presence of the skin.

The present study was made with a relatively small distance between transmitting and receiving optodes. Since it is known that enlarging this distance yields less influence of skin on concentration measurements, such an enlargement would be the first choice for improvement of the NIRS measurement of cerebral oxygenation. However, at the same time, precision decreases due to enhanced light attenuation related to the enlarged light path. This induces a need for improvement of the signal-to-noise ratio of the density measurements.

A second possibility would be to use experimental data to correct the concentration changes measured with undisturbed, i.e., intact skin. The magnitude of this correction might, however, be species dependent. Nevertheless, the present study shows that such a correction would surely not be trivial, nor the result of it negligible.

We conclude that the skin influences regional cerebral oxygenation measurements. Moreover, we hypothesize that after removal of the skin, the estimated concentration changes of $c{\mathrm{O}}_2\mathrm{Hb}$ and $c\mathrm{H}\mathrm{Hb}$ due to arterial saturation changes are dominated by oxygenation changes in the brain tissue, while assuming that in a stable condition of the animals, the influence of the CSF layer and the skull is oxygen independent.