2.1.

OCT Measurement and Signal Processing

Measurements were performed using a fiber-based time-domain OCT setup with a 1300-nm super luminescent diode (SLD). Light from the SLD was split into two arms by a 50:50 fiber splitter. Longitudinal scanning was obtained by using a tilting mirror $\left(200\mathrm{Hz}\right)$ in the rapid scanning optical delay (RSOD) line in the reference arm.¹¹ Transversal scanning was obtained by scanning the sample with the object beam. Signals at the detector were digitized and processed by a computer to form 2-D tomographic images ( $512\times 1024\mathrm{pixels}$ , $8\times 1.4\mathrm{mm}$ , axial resolution: $\approx 12\mathrm{\mu }\mathrm{m}$ , lateral resolution $\approx 20\mathrm{\mu }\mathrm{m}$ ).

For the results presented here, 32 consecutive axial samples were used to calculate a frequency spectrum using a Fourier transform algorithm (see Fig. 1 ). A Gaussian waveform was fitted to the spectrum to obtain a center frequency, amplitude, and spectrum width. The Fourier window was moved over the whole dataset (2-D area). A shift of the center frequency relates to flow or other causes of movement of the sampled volume. To correct for low-frequency motion artifacts, caused by breathing and beating of the heart, the average Doppler shift in a specific depth scan (a-scan) was calculated and subtracted from all values in that a-scan. Noise in the resulting dataset was further reduced by applying a moving average window ( $5\times 5$ pixels) and a threshold before the Doppler shift image was superimposed on the amplitude picture. The value of the applied threshold, obtained with control measurements described ahead, determined the minimal detectable flow.

Fig. 1

Measurement scheme. A window of consecutive 32 samples is used to calculate the frequency shift, which is then assigned to the first (top) pixel of that window. Then the window is shifted one pixel down or to the next a-scan and the calculation is repeated. After completion of the last a-scan, the average frequency shift was calculated from all the values in one a-scan. This average value, containing the (low) frequency shift caused by breathing and beating of the heart, was subtracted from all individual values in the a-scan. Then a moving average window and a threshold is applied on the result.

Because of the averaging procedures and the unknown angle of the blood vessels with respect to the probing beam, no attempt was made to calculate real flow velocities. The direction of the flow was not taken into account. From the resulting images, the relative perfused area (RPA), which we define as the ratio of the number of pixels indicating perfusion over the total number of pixels in a region of interest, was determined. The values reported here are ratios of the RPA before and after the treatment and therefore represent relative changes in perfusion (RCP).

Where possible, three areas were identified in the images: skin, connective tissue, and tumor periphery. Figure 2 shows an OCT image in relation to a typical histology image of untreated mouse skin and subcutaneous tumor. A grayscale is used to visualize the differences in back scattering; a brighter layer indicates higher back scattering. The upper layers of skin (stratum corneum, epidermis, and dermis) are highly reflecting (area A in Fig. 2). Between the skin and the tumor or muscle, a low reflecting layer can be observed, which consists of loose connective tissue containing blood vessels (area B). Area C is a combination of a thin, highly vascularized muscle layer that is usually distinguishable below the connective tissue layer forming the capsule of the tumor and the upper part of the tumor.

Fig. 2

Histology image of untreated mouse skin with a subcutaneous tumor (left image), a ruler with a total length of $1\mathrm{mm}$ , and a representative OCT image (different location) is shown on the right image. Three different tissue layers in depth are indicated: A (skin), B (connective tissue), and C (tumor periphery consisting of the capsule and the upper part of the tumor).

To compare the PDT effects quantitatively, the slope of the backscattered light as a function of depth in the tissue was determined. The attenuation coefficient changes if the viability or configuration of the cells changes.² As described by van Leeuwen, this slope is a function of both the point spread function (PSF) of the OCT system and the optical properties (attenuation coefficient) of the tissue.¹³ We have not determined the PSF of our system but fixed the focus at the air tissue boundary, thereby keeping the influence of the PSF constant throughout the measurements. Furthermore, a low NA lens was used to focus the beam on the tissue, which reduces the influence of the PSF on the signals. The reported values are therefore proportional to the attenuation coefficient and can consequently be used to follow relative temporal changes in tissue morphology.

2.2.

Animals

Female nude BALB/c mice were used for all experiments. HNXOE cells (kindly provided by G. van Dongen, Vrije Universiteit, Amsterdam, The Netherlands) originated from a human head and neck squamous cell carcinoma metastasis in the oral cavity. HNXOE cells were cultured in Dulbecco modified Eagle medium (Gibco, Rockville, MD) supplemented with 5% fetal calf serum (Gibco) and antibiotics and grown at $37°\mathrm{C}$ in a 5% ${\mathrm{CO}}_2$ incubator. To propagate donor tumors in vivo, ${10}^6$ HNXOE cells, in a volume of $0.1\mathrm{ml}$ PBS, were injected subcutaneously in the lower dorsum of mice. When these donor tumors had reached a mean diameter of $8–10\mathrm{mm}$ , tumor fragments $(\pm 1{\mathrm{mm}}^3)$ were transplanted into the lower dorsum of recipient experimental mice. This two-step procedure was used to create well-localized spherical subcutaneous tumors with predictable growth velocity. The time required for the tumor to get to 5–7-mm diameter was about 27 days. During the imaging procedures, the mice were anesthetized by i.p. administration of a mixture of hypnorm and dormicum ( $12.5\mathrm{mg}\bullet {\mathrm{kg}}^{-1}$ fluanisol $+0.4\mathrm{mg}\bullet {\mathrm{kg}}^{-1}$ fentanyl citrate and $6.25\mathrm{mg}\bullet {\mathrm{kg}}^{-1}$ midazolam hydrochloride). To maintain body temperature, a temperature-controlled underlay $(37°\mathrm{C})$ was used.

2.3.

Control Groups

Four control groups were used, each comprising two animals. In the first control group, the blood flow in the skin, connective tissue, and tumor was completely blocked by fixing a metal clamp around the tumor area (see Fig. 3 ). The effects on flow before, during, and after these interventions were measured by OCT. In groups 2 and 3, vasodilation was induced by administration of $5\mathrm{mg}∕\mathrm{kg}$ nicotinic acid ( $0.75\mathrm{mg}∕\mathrm{ml}$ solution) or $300\mathrm{mg}∕\mathrm{kg}$ hydralazine ( $60\mathrm{mg}∕\mathrm{ml}$ solution), respectively. Both drugs are vasoactive but have different working mechanisms. A vasodilation in the cutaneous blood vessels, resulting in an increased blood flow, was expected for both drugs. Nicotinic acid was also expected to lead to an improvement in tumor perfusion by preventing temporary shutdown of tumor vessels. ^{14, 15, 16, 17} Hydralazine is known to reduce the peripheral vascular resistance by directly relaxing arteriolar smooth muscle. This causes a relative decrease in perfusion of some tumor types, as the blood flow is preferentially diverted to critical normal tissues; the so-called steal effect.^{18, 19} In the fourth group, two animals received the photosensitizer but were not illuminated to assess the effect of the drug alone. In one animal of this last group, we imaged a needle that was inserted in the tumor and under the overlying skin. The OCT image was compared with the corresponding histology slide to identify the different layers in our measurement volume (data not shown).

Fig. 3

Mouse under the scanning head of the OCT device, a metal clamp was fixed around the tumor area.

2.4.

PDT Group

mTHPC (Quanta Nova LTD, Scotland, supplied as a solution, $4\mathrm{mg}∕\mathrm{ml}$ drug in a mixture of 1:1:2 propylene glycol:ethanol:water) was administered via the tail vein at a dose of $0.3\mathrm{mg}∕\mathrm{kg}$ . At either $3\mathrm{h}$ or $48\mathrm{h}$ after administration, the mice were anesthetized, after which the skin of the flank and lower dorsum overlying the tumor was observed by OCT. After these pre-treatment measurements, the animals were illuminated with $100\mathrm{mW}∕{\mathrm{cm}}^2$ $652\mathrm{nm}$ light. A total energy of $30\mathrm{J}∕{\mathrm{cm}}^2$ was delivered using a bare fiber. The illumination spot was adjusted to cover the tumor with 0.5-cm margins. Repeated OCT measurements were made for up to 4 hours after illumination. The measurements were done in triplicate (for each time point, on each animal, three independent measurements) and three mice were used in each group. After the last measurement, the animals were sacrificed and tissue samples were collected from both treated and nontreated areas and processed for histology.

2.5.

Ethics

The study protocol was approved by the local animal welfare committee and conformed to national and European regulations for animal experimentation.

3.1.

Control Groups

OCT measurements made before and 10 minutes after clamping of the tumor and overlying skin (group 1) clearly showed a dramatic drop in RPA after clamping (Fig. 4 ). Flow immediately-stalled in all three layers. After removal of the clamp, perfusion remained absent in both skin and tumor during the observation period (10 minutes). In the connective tissue layer, however, Doppler shifts above threshold were seen after removal of the clamp.

Fig. 4

Effect of placement of a clamp on the perfusion of the skin, connective tissue, and tumor. The results are normalized on the preclamping values. The results are means $\pm \mathrm{SD}$ (3 measurements per mouse, 2 mice per data point).

Perfusion of peripheral areas of the HNXOE xenograft, overlying connective tissue, and skin was also measured repeatedly after administration of nicotinic acid or hydralazine (Fig. 5 ). Both drugs induced an increase in perfusion in the skin immediately after administration, but changes induced in skin were much more pronounced for nicotinic acid (Fig. 5a). As expected, an increase in the tumor perfusion was observed after administration of nicotinic acid and a decrease after administration of hydralazine (Fig. 5b). Baseline values of perfusion were restored at 50 minutes and 120 minutes after administration of nicotinic acid and hydralazine, respectively. The same figures (Figs. 5a and 5b) show the average values of the RPS measured in skin of control group 4 (photosensitizer, no light). No significant changes were observed, which is also true for the RPS in the other layers.

Fig. 5

Perfusion in the mouse skin, connective tissue, and tumor after administration of nicotinic acid [left graph (A)] and hydralazine [right graph (B)]. The * indicates significant differences from baseline values [(t-test, paired, $p<0.05$ )]. Legend: $-◆-$ : skin; $-∎-$ : connective tissue; $\mid -▴-$ : tumor; —: control. The results are means $\pm \mathrm{SD}$ (3 measurements per mouse, 2 mice per data point).

3.2.

PDT Group

Figure 6 shows the OCT images and corresponding histology before (Fig. 6a) and $138\min$ after PDT (Fig. 6b) using a 3 hr D-L interval. The two graphs on the right side show the intensity of the backscattered light as a function of depth.

Fig. 6

Histology and OCT image of mouse skin with underlying subcutatious tumor, before treatment (top row, A) and $138\min$ after treatment (bottom row, B, $3\mathrm{hr}$ D-L interval) The graphs on the right show the backscattering intensity that was used to calculate the attenuation (see Fig. 6). Depth 0 represents the tissue surface. The arrows show the area of increased backscattering by edema. This causes an increased signal level and a stronger decrease of the signal in depth.

The most striking histological difference in tissue structure after PDT was the large amount of edema that accumulated in the connective tissue layer (compare Figs. 6a and 6b). This caused an increase in scattering (see also Fig. 7 ) and a thickening of the connective tissue layer (Fig. 6b), which hampered imaging of the underlying tumor.

Fig. 7

Optical attenuation (base e) in the different layers at different times after PDT for short $\left(3\mathrm{hr}\right)$ D-L interval (5A) and $48\mathrm{hr}$ D-L interval (5B). $∎\mid$ skin; $⧈\mid$ connective tissue and $◻$ : tumor. The results are mean $\pm$ SD (3 measurements per mouse, 2 mice per data point).

Changes in light attenuation after PDT for the 3 hr D-L (A) and 48h D-L (B) interval groups are shown in Fig. 7. All layers showed a trend for increased attenuation after PDT. This was greatest in the peripheral tumor layer for the 3hr D-L interval and greatest in the skin for the $48\mathrm{hr}$ D-L interval, but none of the individual measurements was significantly different from baseline. The large error bars reflect biological variability, inhomogeneities in the layer, and the high sensitivity of the fitting procedure for the length of the segment over which the fit is done.

Changes in RCP with time after PDT are shown in Fig. 8 . After PDT with a short D-L interval $\left(3\mathrm{h}\right)$ perfusion of skin and tumor periphery decreased to values of 50% (skin) and 30% (tumor) after 160 minutes. Perfusion in the connective tissue initially increased, followed by a decrease in perfusion to a value that was approximately half the baseline value at 160 to 200 minutes. After PDT using longer D-L interval $\left(48\mathrm{h}\right)$ , no significant changes in RCP were seen.

Fig. 8

Perfusion effects in skin, tumor, and connective tissue after PDT for 3- and 48-hr D-L intervals (Figs. A and B, respectively. Error bars indicate SD). The * indicates a significant (t-test, paired, $p<0.05$ ) difference from baseline values. Legend: $\overline{-◆-}$ : skin; $\overline{-∎-}$ : connective tissue; $\mid -▴-\mid$ : tumor. The results are mean $\pm$ SD (3 measurements per mouse, 3 mice per data point).