2.1.

Animal Preparation

A standard hamster (Mesocricetus auratus) cheek pouch model was used.²⁷ By application of 0.5% DMBA (9,10 dimethyl-1, 2-benzanthracene) in mineral oil three times per week, mild to severe dysplasia developed in three to six weeks, progressing to squamous cell carcinoma at approximately 10 to 12 weeks. Histological features in this model have been shown to closely correspond with pre-malignancy and malignancy in human oral mucosa.²⁷ This study used nine female animals, 10 to 12 weeks old. The median lining wall of one cheek pouch of each hamster was treated with DMBA carcinogen in mineral oil; the other cheek pouch was treated only with mineral oil and served as control. Previous studies have shown that, over 16 weeks, this carcinogenesis process in one cheek pouch does not affect the other cheek.^{28, 29} Therefore, the untreated cheek pouch can be used as the control. Half of the hamsters were euthanized and their cheek pouches were dissected and pinned onto histological corkboards for immediate OCT and polarimetry imaging. The other half of the hamsters were imaged in vivo using intraperitoneal 4:3 Ketamine HCL (100 mg/ml): Xylazine (20 mg/ml) at a dose of 1.0 cc/kg. These animals were wrapped in mylar during imaging to prevent hypothermia. The cheek pouches were inverted, held in place with a fixation device, kept moist using irrigation with sterile saline, and imaged.

2.2.

Spectral-Domain Optical Coherence Tomography

Spectral-domain optical coherence tomography is a high-speed, high-resolution, noninvasive, cross-sectional imaging technique based on a Michelson interferometer. The experimental setup is shown in Fig. 1. Four arms in the Michelson interferometer are assigned by a broadband light source (1310 nm of center wavelength and 90 nm of a full width at half maximum), a spectrometer (0.13 nm of spectral accuracy and 7.7 kHz of frame rate) as a fringe detector, an immobilized mirror at reference arm, and a two-axis scanner with focusing lens at sample arm. A mathematical description for the principle of the spectral-domain OCT was previously described in detail.³⁰ Here, we summarize how it works. Reflected lights from the immobilized mirror and each scattering particle in a sample make a signal in spectral domain that is detected by the spectrometer. When the path difference between the reference and the sample arms is defined by 2z and the refractive index of the sample by n, the scattering event by each particle at a different depth z is encoded in the frequency 2nz of cosine function and the signal is a sum of the cosine functions with different z’s, where the amplitude of each cosine is proportional to the scattering amplitude. An inverse Fourier transform of the signal in spectral domain gives a complex signal in the z domain. The powers of peaks in the z domain represent the scattering amplitudes and are converted to gray scales to make an array of images along the z direction. While the two-axis scanner scans the xy plane, the array is continuously acquired to make a volume image. The current setup has an 8-μm (in air) depthwise and 13-μm lateral resolutions.

Fig. 1

Schematic of a fiber-based spectral-domain OCT: a 130-nm wide spectrum was sampled by a 1 × 1024 InGaAs detector array at 7.7 kHz. Imaging depth and depth resolution were 3.4 mm and 8 μm in air, respectively. A 2-axis scanner with two galvo mirrors was used. LCL, low-coherence light; CM, collimator; DAQ, data acquisition system; DG, diffraction grating; FL, focusing lens; LSC, line scan camera; GM, galvo mirror; RL, relay lens.

2.3.

Polarimetry

The system, shown in Fig. 2, consists of four voltage controlled optical elements (R1, R2, R3, and R4) (Meadowlark Optics, Frederick, Colorado). The light source is a fiber optic illuminator (Dolan-Jenner Industries, Inc., Lawrence, Massachusetts) with a 30-watt, 20-volt halogen bulb. The light passes through a laser line interference filter (F) with wavelength transmittance from 400 to 700 nm. The light emerging from the filter is collimated by a 38.1 mm focal length convex lens (L) (Newport Corporation, Fountain Valley, California) to provide a broad beam to illuminate the sample. The resulting light then passes through a Glan Thompson 100,000:1 polarizer (P1) (Newport Corporation, Fountain Valley, California), a linear polarizer set at horizontal, and two voltage-controlled variable retarders (R1, R2). These optical components allow for modulation of the input state of polarization. The variable retarders produce all of the necessary states of linearly and circularly polarized light. The operational basis underlying voltage-controlled optics is that voltage supplied in the form of a 2-kHz square wave modulates the degree of rotation of the polarization axis and the amount of retardation by the voltage controlled retarder. Light scattered from the sample toward the camera passes through another series of polarizing optics. The backscattered light from the sample is imaged by the objective lens (10×/0.30 N.A.). Theoretical resolution of this objective (λ/(2*N.A.)) is about 0.7 to 1.2 μm. Before not using the lens, the resolution was about 15 to 20 μm. The calibrated resolution is about 1.5 to 2.0 μm. The output branch contains another polarizer set at vertical (P2), consisting of two voltage-controlled variable retarders (R3, R4) encountered by the light in reverse order from the input branch. The resulting images are captured by a computer controlled 1024×1024 pixels, 14 μm×14 μm per pixel, and 12-bit CCD camera (Kodak, San Diego, California). Extracting the retardance images from the measured Mueller matrix requires performing the polar decomposition and a mathematical description for Mueller matrix retardance imaging from the polarimetry system was described in detail.²⁶

Fig. 2

Schematic of Mueller matrix polarimetry system.

2.4.

Co-localization

In order to co-localize OCT, polarimetry, and histology, the blood vessel network in the cheek pouch was used as a marker. After the cheek pouch was held in place with a fixation device, it was imaged in vivo with OCT, polarimetry, and light microscopy. One thousand and twenty four two-dimensional OCT slices (xz plane) were recorded every 5 μm along the y-direction and volume-rendered. En face image (xy plane) of the cheek pouch was taken by polarimetry and light microscopy. Another En face image was also reconstructed from the OCT volume-rendered image and compared to the results of polarimetry and light microscopy. The blood vessel network was clearly imaged for all modalities with a window size of 5.12 mm × 5.12 mm in the xy plane. After the animal was sacrificed, the tissue was harvested and prepared for standard paraffin sectioning and hematoxylin and eosin (H&E) staining. One thousand and five hundreds histological slices (xz plane) were made every 6 μm and imaged using a light microscope with 1, 4, and 10× objective lenses. Every fifteenth histological slice (every 90 μm) was selected to be volume-rendered with 1× of magnification. A reconstructed en face image (xy plane) was made to visualize the blood vessel network and compared to the en face images from the different modalities.

3.1.

Matching En Face Images

Figure 3 shows matching polarimetry image and microscope photo, OCT image, and histology. Figure 3a is the raw polarimetry image with traced blood vessels, Fig. 3b is the original microscope image with blood vessel tracings from the OCT image superimposed and Fig. 3c is the original histology with blood vessel tracings from the microscope image superimposed. Polarimetry, microscopy, OCT and histology image can be co-localized using the blood vessel tracings.

Fig. 3

(a) Raw polarimetry image with traced blood vessels, (b) original microscope image with blood vessel tracings from the OCT image superimposed, and (c) original histology with blood vessel tracings from the microscope image superimposed.

3.2.

Optical Coherence Tomography

Optical coherence tomography (OCT) provided volume rendering images of the tissues. The depth-resolvable feature of high-resolution OCT enabled us to identify every tissue feature, including the blood vessel network, breakdown of the basement membrane, and epithelial as well as subepithelial changes. This was complementary information, which polarization detection and visual examination could not provide. In areas of dysplasia epithelial thickening, down growth of rete pegs, abnormal keratinization, and intact basement membrane are all clearly visible. In areas of malignancy, tissue disorganization, patches of thickened keratinized layer, and/or greatly thickened squamous epithelial layer, as well as breakdown of the basement membrane and invasion of the underlying submucosal connective tissue, are observed. These observations correspond with similar features in the H&E stained histopathological images (Figs. 4 and 5).

Fig. 4

In vivo OCT image and corresponding H&E stained histology of hamster cheek pouch with different stages: (a) healthy region [TeX:] $(\rm SS = 0)$ $(\mathrm{SS}=0)$ ; (b) dysplasia region ( [TeX:] $\rm SS = 1$ $\mathrm{SS}=1$ or 2); (c) squamous cell carcinoma region ( [TeX:] $\rm SS = 6$ $\mathrm{SS}=6$ ). Enlarged images (left for OCT and right for histology) were taken from the white boxes in the original images (center).

Fig. 5

(a) In vivo OCT 3D-image (1, QuickTime, 1.7 MB) and (b) corresponding histology of a hamster cheek pouch with the squamous cell carcinoma (SS = 6) (2, QuickTime, 2.2 MB). Image size: 1.14 mm (W) × 0.52 mm (H) × 0.72 mm (D).

Figure 4 shows OCT images (left side) and matching histology (right side). Figure 4a shows healthy oral mucosa [standard scale (SS) = 0]. Normal epithelial surface appears very bright as the keratinized layer is very reflective. Underneath the epithelial layer, the base membrane is clearly visible with the connective tissue. Figure 4b shows an area of dysplasia (SS = 1 or 2). The epithelial surface is thicker than healthy tissue and the surface keratinized layer is still very bright and completely intact. The basement membrane shows some down growth into the underlying connective tissue but it is clearly still intact. Figure 4c depicts an image of squamous cell carcinoma (SS = 6) that surface keratinized layer is thicker and brightens. The epithelium layer is grown down and the base membrane is broken down.

Figure 5 shows in vivo OCT 3D-image (left side) and corresponding histology of a dysplastic area of the hamster cheek pouch (right side). Surface and subsurface mucosal layers including blood vessels were clearly visible in the 3-D OCT images. Epithelial changes throughout carcinogensis were apparent, including epithelial thickening, loss of basement membrane integrity, and subepithelial invasion.

3.3.

Polarimetry

After two weeks of carcinogenesis using DMBA, the very early changes could not be distinguished with the naked eye. However, tissue retardance was already two to three times greater than that of normal tissue. By four weeks, the dysplastic tissue had retardance four to five times greater than that of normal tissue. Figure 6 shows the normalized retardance images and the gray level indicates the relative magnitude of retardance from baseline. Figure 6a depicts the image after two weeks of DMBA. The green circled sites (A, B, and D) had two to three times greater retardance than healthy sites (SS = 0) and shows an image of dysplasia (SS = 1 or 2). Figure 6b depicts the image after four weeks of DMBA. The circled sites had a four to five times greater retardance than healthy sites and shows an image of squamous cell carcinoma (SS = 6). Green circled sites (A, B, C, and D) in Fig. 6b were also detected using OCT and confirmed with histology, whereas yellow circled sites were confirmed with histology but remained undetected using OCT alone. In these images, we observe a significant retarding structure after the residual retardance is removed. We expect that these patterns are caused by an underlying birefringent structure such as oriented collagen fibers.

Fig. 6

Normalized retardance images from polarimetry based on Mueller matrix. Images after (a) two weeks and (b) four weeks of DMBA (scale bar: 1 mm).

Figure 7a shows how much change of retardance occurred through carcinogenesis and Fig. 7b is a chart of normalized retardance with statistics along the lines in Fig. 7a.

Fig. 7

(a) Shows how much change of retardance occurred through carcinogenesis and (b) chart of normalized retardance with statistics along the lines in (a).