2.1.

Melanoma Model and Imaging Preparations

B16F10 pigmented murine melanoma cells were purchased from American Type Culture Collection and cultured in Dulbecco modified Eagle’s medium supplemented with 10% fetal bovine serum and antibiotics at 37°C and 5% ${\mathrm{CO}}_2$. Cutaneous melanoma was induced by intradermal injection of ${10}^5$ cells in the flank of six- to seven-week-old nude mice (J:NU homozygous, Jackson Labs) using a 30G needle. The mice were kept in microisolator cages with access to food and water ad libitum. All procedures were approved by the Institutional Animal Care Committee (University Health Network, Toronto, Canada: AUP protocol 4401.1).

Diffuse reflectance and OCT imaging were performed four to seven days after tumor cell injection, during the vertical growth phase when the tumors were $\sim 4\mathrm{mm}$ in diameter and 1 to 2 mm in depth. The mice were anesthetized using 5% isoflurane and maintained using 2% isoflurane delivered through a mask. The animal was placed, on a heating pad to maintain body temperature, on a three-axis micropositioning stage. The paw on the tumor-bearing side was immobilized to the stage to reduce movement due to respiration. Immediately prior to imaging and OCA application, the stratum corneum of the skin overlying the normal skin or tumor was removed by tape stripping in order to improve the transdermal penetration of the clearing agent. The OCA comprised a 19:1 mixture of polyethylene glycol (PEG-400) and 1,2-propanediol⁹ and was applied topically prior to the start of imaging. The skin was then covered with gauze soaked with clearing agent until the next imaging session. Although the tumor cells were injected into the dermis, some tumors developed more superficially than others, so that the OCA diffusion time throughout the tumor varied between animals. A typical example is shown in Fig. 1: in this mouse, part of the tumor was located in the dermis and part was subcutaneous.

Fig. 1

White-light photographs of the melanoma mouse model in vivo (a) before and (b) after 250 min of optical clearing.

2.2.

Diffuse Reflectance Measurements

As described above, OCA was applied over the normal skin or melanoma after tape stripping. The tissue was then massaged for 1 min to enhance diffusion.⁹ The first diffuse reflectance measurement was acquired after OCA application but before tissue massage, and then every 15 min for up to 90 min. Diffuse reflectance measurements were acquired using a fiber optic reflectance probe with $520\mu \mathrm{m}$ source–collector fiber separation;⁸ although somewhat dependent on tissue optical properties, the typical sampling depth at this source–detector separation is quite superficial, typically $<200\mu \mathrm{m}$. The probe tip (1.5 mm diameter) was placed in gentle contact with the tissue surface and the full reflectance spectrum between 450 and 650 nm was collected in $<1\mathrm{s}$. DRS and OCT imaging were performed in different experimental setups to avoid changing the probe and OCT position when acquiring the data. In both setups, the tumor was fixed to minimize bulk tissue motion due to, for example, breathing. Hence, different groups of mice were used to avoid double application of OCA in the same animal.

2.3.

Optical Coherence Tomography Imaging

The swept-source OCT system has been described previously.¹⁰ Briefly, images were acquired with a 36 kHz Fourier domain OCT system utilizing a short-cavity swept laser source based on a tunable polygon filter, with a sweep range of 110 nm centered at 1310 nm. The axial and lateral resolutions in tissue were 8 and $13\mu \mathrm{m}$, respectively. OCT imaging was performed immediately following tape stripping, OCA application, and 1 min of massage, and then every 30 min during optical clearing. Each imaging session took 3 min for collecting structural images for speckle variance (SV) analysis to highlight the microvasculature; the SV technique has been described elsewhere¹¹ and is based on decorrelation between successive B-scans due to movement of blood cells. SV images were then processed to obtain depth-encoded microvasculature maps of the tumor and surrounding tissues.

Since there was little prior experience in OCT imaging of melanoma, we deliberately varied some of the experimental parameters, such as using glass or plastic slides in contact with the tissue surface during the image acquisition, to test for the consistency of the speckle variance optical coherence tomography (svOCT) results. We noted some biological variability between animals in the growth pattern, and hence depth, of the tumors due to unavoidable differences in the tumor cell microinjection and animal-to-animal variations in biological response. Data from a total of six mice are reported in this initial proof-of-principle study.

OCT structural images were acquired in SV mode with 760 B-scans taken over a $6\mathrm{mm}\times 6\mathrm{mm}$ surface area (interscan separation of $\sim 8\mu \mathrm{m}$); to enable interframe comparison required for svOCT analysis,¹¹ eight repeat B-scans were acquired at each position. The svOCT algorithm then calculates the interframe intensity variance from the same spatial location,¹¹ with the contrast arising from differences in time-varying speckle properties. Thus, we define

In Eq. (1), if $N$ B-scans are acquired faster than the stationary solid-tissue decorrelation time, then the value of ($I_{ijk}-\overline{I_{Jk}}$) for these pixels approaches zero, thereby suppressing the tissue signal in the resulting SV image. Here, the B-scan acquisition time was set to 25 ms: this was fast enough that signals from stationary tissues did not decorrelate between frames (thus, $\sim 0{\mathrm{SV}}_{\text{signal}}$), while being sufficiently slow to ensure complete interframe decorrelation for pixels representing vascular blood (thus, high ${\mathrm{SV}}_{\text{signal}}$).¹²

Several postprocessing procedures were applied to prepare the microvascular 3-D data sets for depth encoding and for presentation as en-face microvascular images. These included

In addition to comparing the resulting depth-encoded svOCT microvascular maps, the clearing effect was quantified using spatial texture analysis of OCT image speckles.¹³ Specifically, regions of interest (ROIs) within the overlying skin and underlying tumor regions were selected within same cross-sectional OCT images at different time points after application of the clearing agent. The ROI for skin started at the top pixel below the tissue surface. Tumor ROIs were selected manually for each mouse, depending on the tumor depth location. All ROIs were $30\times 100\text{pixels}$, corresponding to a physical size of $125\mu \mathrm{m}$ in $\text{depth}\times 765\mu \mathrm{m}$ laterally. Histograms of signal intensity distributions within each ROI were then fitted using a least-squares method with the gamma distribution function.¹⁴

3.1.

Diffuse Reflectance Measurements

The first measurement (0 min) for both normal skin and melanoma was taken immediately after the application of the OCA in order to minimize variations in optical coupling of the probe and the tissue surface. As seen in Fig. 2(a), normal skin showed around $\pm 20\%$ variation in diffuse reflectance across this spectral range over time after OCA application. The reflectance decreased in the first 15 min and then increased at 30 min, followed by decrease up to 90 min. These changes in the signal may be associated with tissue dehydration/rehydration. Figure 2(c) shows the corresponding diffuse reflectance measurements in experimental melanoma. At 15 and 30 min into the clearing, the diffuse reflectance decreased overall by $\sim 75\%$ compared to the initial value. The clearing reached its maximum effect at 45 min, at which time the reflectance had decreased by $\sim 90\%$. The reflectance then increased at the 60 min time point, possibly due to a physiological response to reestablish hydration, after which the reflectance decreased again at 75 and 90 min [as also seen in normal skin in Fig. 2(b), although to a lesser extent].

Fig. 2

Diffuse reflectance of [(a) and (b)] normal skin and [(c) and (d)] cutaneous melanoma during optical clearing, normalized by the initial value in each case. (a) through (c) Normalized diffuse reflectance spectra, and (b) through (d) changes in reflectance at several wavelengths as a function of the clearing time.

In terms of spectral changes in the diffuse reflectance with clearing, it can be seen that in normal skin [Fig. 2(b)], the reflectance change was nearly uniform across this 200 nm spectral range, except for a slightly greater effect at shorter wavelengths. Since there is a gradient of concentration of OCA into the tissue with topical application, the clearing particularly affects the short wavelengths that have smaller penetration in tissue. Longer application times are required to clear deeper tissue layers. The same behavior was observed for melanoma in the first 45 min. In this particular example, at around 60 min clearing, the characteristic spectral signature of oxyhemoglobin (the double dip at around 540 and 575 nm corresponding to peaks in the ${\mathrm{HbO}}_2$ absorption spectrum) becomes apparent in the tumor, indicating that the probe is detecting light that has propagated through a well-vascularized region of tissue. This effect is washed out at the longer times, probably due to small changes in the probe-sampling position during the measurement. Quantitative interpretation of these data is also limited by the fact that the source-to-detector fiber separation in the diffuse reflectance probe is only $\sim 0.5\mathrm{mm}$, which is not enough to sample the full tumor thickness of 1 to 2 mm. Nevertheless, the diffuse reflectance data do suggest that topical application of OCA does clear melanoma tissue in vivo.

Clearing agents match the refractive index in the tissue, decreasing the light scattering. Most of the agents studied have a refractive index around 1.4, which is optimal for soft tissue. However, melanin granules present in melanoma have a refractive index around 1.7,¹⁶ so that the clearing matches the soft tissue but not the melanin granules. Thus, OCA improves the light penetration in melanoma but does not completely eliminate the scattering interfaces.

3.2.

Optical Coherence Tomography Images

B-mode structural images acquired immediately and at 250 min after OCA application are shown in Fig. 3, where it can be seen that immediately after OCA application, only the superficial region of the tumor was visible, down to a depth of $\sim 400\mu \mathrm{m}$. At the later clearing times, the subcutaneous tumor became visible down to $\sim 750\mu \mathrm{m}$ [Fig. 3(b)]. The average intensity gradient with depth before and after OCA application showed the same behavior. The tissue surface has high initial signal and, after OCA application, becomes more transparent and a higher signal is obtained from deeper tissue layers [Fig. 3(c)].

Fig. 3

Example of B-mode OCT imaging of optical clearing in melanoma in vivo. (a) Photograph of the tumor at the skin surface. (b) Structural B-mode images at 0 (upper) and 250 min (lower) of topical application of OCA, taken through the dark spot seen in (a), which is indicated on the OCT images by the blue arrows. The increased OCT imaging depth (green arrow) due to the clearing is apparent. (c) Average intensity gradient with depth of an area in B-scan (dot lines) before and after OCA application.

SV analysis provided tissue microvasculature images immediately after and at 120 and 250 min following OCA application, as exemplified in Figs. 4 and 5. Prior to clearing, svOCT showed the vascular network in the superficial skin and tumor layers, as also observed in the structural B-mode images. After 120 min of clearing, larger and deeper vessels became more visible; this improved further by 250 min. For example, as seen in the bottom-right panel, the microvascular network down to the $\sim 750\mu \mathrm{m}$ depth is beginning to emerge at $t=250\min$. The disorganization of the tumor neovasculature is also apparent (Fig. 4).

Fig. 4

En-face depth-encoded svOCT projections at 0, 120, and 250 min after topical application of OCA for melanoma. Melanoma stimulates angiogenesis, so that more vessels are seen, even prior to clearing. With clearing, the top layers of the tumor become more transparent, allowing visualization of the disorganized deeper tumor microvasculature, to a depth of $\sim 500$ to $600\mu \mathrm{m}$ at 120 min and $\sim 600$ to $750\mu \mathrm{m}$ at 250 min.

Fig. 5

Normal skin en-face depth-encoded svOCT projections at 0, 120, and 250 min after topical application of OCA. In the normal skin, small vessels in a regular pattern are seen before clearing. Larger vessels start to appear with clearing, probably localized under the skin $300\mu \mathrm{m}$ in depth.

Comparing normal skin and melanoma, it is possible to observe larger vessels in the tumor due to angiogenesis. The vascular network is also related to the age of the mice. In this study, the mice were six to seven weeks old, suitable for tumor development but only presenting small and regular vessels in normal skin (Fig. 5). In addition, diffusion of OCA may differ between normal skin and melanoma: due to its rapid growth, melanoma tissue is poorly organized, so likely the diffusion of the OCA within it is greater. To better quantify the effects of optical clearing on imaging the deeper features of the melanoma microvasculature, Fig. 6 shows examples of histograms of the svOCT pixel intensity in ROIs of the overlying skin and the underlying tumor tissue, together with the corresponding $\alpha /\beta$ ratios obtained from fitting Eq. (2) to the OCT structural image data. Note the shift in the signal histograms toward smaller-intensity values with optical clearing in the skin, indicating reduced backscattering of the light (as more photons now reach deeper tissue layers). Conversely, in the underlying tumor, the corresponding signals become stronger, again indicating that more light is now able to penetrate to the deeper tissues.

Fig. 6

Structural OCT image pixel intensity distributions of melanoma and normal skin for the ROIs shown in the insets ($\text{scale bar}=100\mu \mathrm{m}$): (a) and (b) at time 0, and (c) and (d) after 250 min of clearing with physical massage of the tissue. The blue lines show fits to the gamma distribution function of Eq. (2) and the resulting $\alpha /\beta$ ratios. The centers of the ROIs for skin and melanoma were selected at depths of 65 and $650\mu \mathrm{m}$ below the tissue surface, respectively. Both ROIs were $30\text{pixels}\times 100\text{pixels}$ ($125\mu \mathrm{m}\times 765\mu \mathrm{m}$).

Figure 7 summarizes the $\alpha /\beta$ ratios before and after clearing data for the six mice, where the imaging conditions, periods of OCA application, and depth of the tumors varied between animals, as indicated.

Fig. 7

Summary of the changes with optical clearing in the $\alpha /\beta$ ratio in skin overlying melanoma (green) and in melanoma (red). Note that the ratio before clearing divided by the ratio after clearing is plotted, so that values $>1$ (red horizontal line) indicate a reduction in signal intensity (as in the skin) and values $<1$ represent signal intensity increase (as in tumor). The ROIs for the skin were selected below the tissue surface. Tumor ROIs were selected manually for each mouse depending on the tumor depth, which is indicated above each data point. Both ROIs were $30(\text{depth})\times 100(\text{lateral})$ pixels, corresponding to a physical size of $125\mu \mathrm{m}\times 765\mu \mathrm{m}$. The clearance times and mouse number are indicated along the abscissa axis. Triangles indicate that OCT imaging was done without covering the tissue surface; circles indicate a thin plastic covering and squares indicate that the tissue surface was covered with a glass coverslip. These data serve to demonstrate that the changes with optical clearing are relatively insensitive to the exact OCT imaging conditions.

The overall trends show that the $\alpha /\beta$ ratio in the skin increased consistently with clearing (average $\sim +20\%$), while the ratio consistently decreased in the melanoma (average $\sim -30\%$). As per our conjecture, the surface tissues are becoming more transparent, allowing greater depth of penetration of the light into the underlying tumor tissue. Covering the tissue surface during imaging did not appear to make a substantial difference to these ratios, indicating robustness in the measurements and analysis. There was also no systematic trend apparent with increasing clearing time, likely due to the biological variability in the model. However, systematic time dependence was indeed observed in single mice, as illustrated by the example shown in Fig. 8. In this case, increasing loss of signal intensity with clearing time was apparent in the skin, as was better tissue visualization at the 400 to $525\mu \mathrm{m}$ layer in the tumor, as also seen in the corresponding increased brightness of the microvasculature. Figure 9 shows that the image contrast and resolution at this depth in tumor are also improved by the optical clearing. There was minimal change in the $\alpha /\beta$ ratio at the deeper layer in the tumor, indicating that the OCA did not reach this depth even after 4 h of OCA application.

Fig. 8

Example of the time course of optical clearing in one mouse: in 5 skin (0 to $125\mu \mathrm{m}$—squares) and in tumor at two different depths (ROIs at 400 to 520 circles and 750 to $875\mu \mathrm{m}$ triangles).

Fig. 9

Example of the improvement in svOCT image quality within melanoma as a result of optical clearing. (a) Depth-encoded en-face projection of microvasculature after 250 min clearing. (b) The area within the white dashed rectangle in (a) is shown at three different time points during clearing; the arrows indicate the apparent reduction in its measured width (black line). (c) Corresponding normalized speckle variance intensity across the diameter of this vessel and its cross-section.

The shape of the cross-sectional profile through the blood vessel in Fig. 8(b), with the flat-top appearance and sharp edges at the 250 min time point, suggests that the effect of clearing may be due to altered optical properties of the tissue (thus improving imaging contrast / resolution) and/or physiological response of the vessel to the OCA.