2.1.

OCT System

In this work, we used a portable spectral domain OCT system with a central wavelength of 840 nm, a bandwidth of 80 nm, and longitudinal and transverse resolution of 7.8 and 30 μm, respectively. It is made by Shenzhen MOPTIM Imaging Technique Co., Ltd., China. Figure 1 shows the diagram of the OCT system used in this study, and a photograph of the experimental setup is shown in Fig. 2.

Fig. 1

Diagram of the OCT system.

Fig. 2

Photograph of the experimental setup.

Lateral and longitudinal scanning are at 3 and 1.8 mm, respectively. Single lateral scans can be accomplished in $1/15\mathrm{s}$; in other words, we can obtain 15 images every second. The OCT system operation is controlled automatically by a portable personal computer (PC), and a 2-D OCT image is selected every half minute. Altogether, there are more than 300 2-D OCT images obtained in each experiment and stored in the PC for further processing.²⁴

2.2.

Materials

The measurements were carried out on 12 volunteers (7 men and 5 women) aged 22 years or older (mean age 28 years). They were in good health, took no medication, and were selected from the same racial groups randomly. A signed informed consent was obtained from all subjects.

The in vivo window preparation had a 1-cm diameter circle in the skin of dorsal forearms, and OCT images were taken. Before each individual experiment, the detected area would be shaved to eliminate influences to the experimental results by the fine hair on the arm.³⁴ During the measurements, the volunteers were asked to remain still to minimize motion artifacts, and food and drinks were not permitted. Three subjects were excluded from the study because they were not able to complete the experiments.

A standard oral glucose tolerance test (OGTT) using 75 g glucose was performed in all volunteers starting at 0900 (time 0) after an overnight fast. At the same time, OCA was applied to the surface of the skin (approximately 0.1 mL) of the volunteers’ forearms where the OCT images were taken. Synchronously, BGC was monitored serially at about 10-min intervals using one touch ultra easy glucose analyzers. Whole-blood samples were drawn from the subjects’ fingertips.^31,33 The duration of each individual experiment was about 160 min. Before being treated with the OCAs, the detected regions of subjects were measured for about 10 min to establish a baseline. Continuous monitoring of tissue optical properties upon application of different OCAs was performed for up to 150 min. The room temperature was maintained at 25°C throughout the experiment, so as to eliminate influences caused by temperature fluctuation.^32,45

Four kinds of OCA were used in this study: $100\%\mathrm{v}/\mathrm{v}$ glycerol (100% G), $50\%\mathrm{v}/\mathrm{v}$ glycerol (50% G), $16\%\mathrm{w}/\mathrm{v}$ mannitol (16% M), and $16\%\mathrm{w}/\mathrm{v}$ mannitol (10% M) solutions. The $50\%\mathrm{v}/\mathrm{v}$ glycerol solution was prepared through mixing the agents with distilled water, and mannitol solutions were prepared by dissolving 16 or 10 g of mannitol in 100 mL purified and distilled water. Mannitol and glycerol (with 99.9% concentration) were purchased from Tianjhi Damao Chemical Reagent Factory, China.

2.3.

Methods

In this study, OCT images obtained from human skin were processed using an original program developed in Microsoft Visual C++ 6.0 to obtain the OCT signal and calculate the OCT signal slope (OCTSS)⁴⁵ Over approximately 1 mm, which is sufficient for speckle noise suppression, the 2-D images were averaged in the lateral ($x$ axis) direction into a single curve to obtain an OCT signal that represents 1-D distribution of light in depth.²⁶ To reduce the effect of surface curvature, the most flat regions of the 2-D images were used for averaging.⁴⁵ The 1-D distribution was plotted in logarithmic scale as a function of depth for further analysis.

The local increase of glucose concentration in the chosen region caused the scattering coefficient to decrease, which was clearly reflected in the OCTSS graph. The average effects of OCAs on glucose monitoring by OCT were estimated by analyzing correlation between the OCTSS changes and the actual blood glucose concentration. The correlation coefficients ($R$) and $P$ values were calculated. The OCTSS was calculated by linear least-squares fit of a segment with the length of 75 to 125 µm, where the signal is linear and undergoes minimal alterations. This segment lies in the 400- to 650-µm depth range. This analytical method is shown in Fig. 3. The OCT signal slopes were plotted in the inverse scale, and then slopes of four consecutive OCT signals were averaged, normalized, and plotted as a function of time or of actual blood glucose concentration.³⁰

Fig. 3

OCT signals recorded from the human skin in vivo at different times, indicating OCTSS analysis method.

The reason we decided to monitor BGC in the forearm is for convenience. The epidermis of the forearm appears to be much thinner than that of the hand, finger, or wrist.⁴⁶ During rapid blood glucose changes, however, forearm blood glucose lags behind fingertip blood glucose by 30 min on average. As a result, there is a lag time ($t_{\text{lag}}$) between glucose-induced changes in skin optical properties and variations in BGC.^47–49 We varied $t_{\text{lag}}$ to find the maximal absolute value of $R$ to determine the value of $t_{\text{lag}}$.⁵⁰ Values of $t_{\text{lag}}$ varied from 10 to 50 min with an interval of 1 min, and the value of $R$ could be calculated according to each $t_{\text{lag}T}$. When $R$ reached its maximum value, the lag time of the test was determined.

3.1.

Correlation between OCTSS and BGC

The experiments in which sharp blood glucose changes were realized by glucose uptake can be divided into five groups: group A was a control group (no OCA was used), group B used 50% G, group C used 10% M, group D used 100% G, and group E used 16% M. Every volunteer who completed the experiments took part in all five tests (nine people tested in each group).

In group A, good correlations were found between the values of blood glucose and the OCTSS, which is in agreement with Larin’s findings in 2002.³⁰ In the results of group A, when OCTSS is calculated in the $430\sim 550\text{-}\mu \text{m}$ depth region from the skin surface, $R$ (correlation coefficient between the OCTSS and the BGC) varied between 0.81 and 0.87, and the average value was 0.848; no calculated $P$ value was higher than 5%, and most of the $P$ values were $<1\%$. One representative result is shown in Fig. 4.

Fig. 4

(a) OCTSS (plotted in the inverted scale) and corresponding BGCs for one of the experiments of group A; no OCA was used in this group. (b) OCTSS versus BGC for the data shown in (a); $R=0.83$, $P<5\%$. The slope was calculated at a depth region of 430–510 µm.

Figures 4 and 5, which are derived from the same individual, indicate that with the increase of BGC, the OCTSS decreases. In human skin, the scattering of light depends on the ratio of the refractive indices of the components. Changes in the ratio of the refractive indices of the scattering particles and the solute lead to a variation in the transparency of the tissue. An increase in BGC leads to a subsequent rise in the refractive indices of blood and interstitial fluid, whereas the refractive index of scattering particles in the skin remains unchanged.^51,52 Thus, the refractive index mismatch between the solute and the particles diminishes and the scattering is decreased, which was clearly reflected in the OCT signal slopes.

Fig. 5

OCTSS versus BGC for one of the experiments of group B; 50% glycerol was used in this group; $R=0.92$, $P<1\%$. The slopes were calculated at a depth region of 560–650 µm.

Figures 4 and 5 indicate that when the OCTSS is calculated at a depth region of 560 to 650 µm, the correlation between the OCTSS and the BGC is better than that of the region of 430 to 510 µm. Most likely, this phenomenon is caused by the gradient of glucose concentration from dermal blood microvessels to the epidermis. Therefore, sensitivity and accuracy of the OCT measurements of BGC would be maximal in the lower dermis area.³⁰ This was also demonstrated in the results of groups B and C.

In this study, low-concentration OCAs performed better than high-concentration ones. In groups B and C, both the sensitivity and accuracy of the measurements were improved significantly. The results of group B were the best ones: when OCTSS was calculated in the $540–660\text{-}\mu \text{m}$ region, $R$ ranged between 0.88 and 0.94, average value 0.908; no calculated $P$ value was higher than 1%. One representative result is shown in Fig. 5.

OCAs, such as glycerol, have been identified that can change the light-scattering properties of turbid biological tissues. The strong affinity of OCAs produces a water flux out of tissue and leads to dehydration in the tissue. There is a difference in the biomembrane (mainly cell membrance) penetrability between water and glycerol. These factors are the main reasons for optical clearing. In addition, the interaction between collagen and OCAs can also lead to a substantial reduction in tissue light scattering.^51–53 Thus, there would be more effective light dose penetrating through the SC and the epidermis into the lower dermis area, where blood in the skin tissue is much more abundant. The probing depth of OCT was increased when the glycerol optical clearing agent was applied topically on the skin.⁴⁰

Mannitol is also a hyperosmotic agent with molecular weight similar to that of glucose. However, the influence of mannitol on biological tissues has been little studied except in the dura mater.⁴³ In our study, 10% mannitol did not perform as well as 50% glycerol. In group C, $R$ varied between 0.85 and 0.93, average value 0.892 (no calculated $P$ value was higher than 1%). For comparison, the results for groups A, B, and C are shown in Table 1.

Table 1

Correlation coefficient between OCTSS and BGC for groups A, B, and C.

In groups D and E, we could hardly see any result better than in group A, most likely because diluted OCAs more easily diffuse into biological tissue. We did not continue investigating the optical clearing effects of these two kinds of high-concentration agents in our trials.

3.2.

No Glucose Uptake

To see the effects of OCAs only, three additional groups of experiments were proposed in which there is no glucose uptake. No OCA was used in group A1, 50% G was used in group B1, and 10% M was used in group C1. In these groups, no good correlation was found between the values of blood glucose and the OCTSS, and no value of $R$ was $>0.78$. Figure 6 gives two representative result of groups A1 and B1. More details about the results are shown in Table 2.

Fig. 6

(a), OCTSS (plotted in the inverted scale) and corresponding BGCs for one of the experiments of group A1; $R=0.63$, $P<5\%$. The slope was calculated at a depth region of 450–540 µm. (b), OCTSS (plotted in the inverted scale) and corresponding BGCs for one of the experiments of group B1; $R=0.65$, $P<5\%$. The slope was calculated at a depth region of 500–600 µm.

Table 2

Correlation coefficient between OCTSS and BGC for groups in which there is no glucose uptake.

The main reason we did not find good correlation may be that when BGC is low and undergoes tiny fluctuations, it leads to little rise in the refractive indices of blood and interstitial fluid, and the refractive index mismatch inside the tissue does not change much. Thus, there is a less effect on the OCT signal.

3.3.

Lag Time between OCTSS and BGC

In this study, the OCT signals presenting the optical properties of the skin tissues were obtained from dorsal forearms, and the glucose values were measured by fingerprick. The glucose levels in interstitial fluid determined the optical properties of the skin tissues, while the BGC values in fingertips closely followed arterial BGC. However, changes of glucose levels in interstitial fluid do not occur at the same time as those in the intravasal compartment; actually they occur with a delay. During rapid blood glucose changes, the forearm lags behind the fingertip by about 30 min on average. As a result, the detection of hyperglycemia as well as hypoglycemia could be delayed. This physiological delay is dangerous.^48,49

The maximum and minimum values of $t_{\text{lag}}$ (the lag time between the OCTSS and the BGC) in our experiments were 15 and 46 min, respectively; the mean value was 33 min. When compared with group A, however, the $t_{\text{lag}}$ in group B was 8 min shorter on average, and 5 min shorter in group C. This is mainly because the OCT signal slopes in groups B and C were calculated in the lower dermis region where the glucose level in interstitial fluid more closely follows that in veins.