2.1.

Animals

The study was performed on 13 female Wistar rats (five rats with healthy brains and eight with 101.8 rat glioblastoma), with weights of $240\pm 18\mathrm{g}$. All animals were delivered from the Laboratory of Neuromorphology of the Research Institute of Human Morphology (Moscow, Russian Federation). The animals were maintained, two per cage, in a temperature-, humidity-, and light-controlled room, with free access to water and food. Homogenized 101.8 glioblastoma ($\sim {10}^6$ tumor cells) had been inoculated into the ventricle of the right hemisphere of the brain of each of the eight relevant rats using a standard procedure.³¹

Glioblastoma 101.8 is a primary chemically induced brain tumor (grade IV), but after cultivation, it takes on morphological properties similar to human glioblastoma (high cellular density, cell polymorphism, presence of necrotic and hemorrhage areas, and infiltrative growth). CP OCT imaging was performed in four rats on days 7 to 9 after the glioma injection, when the tumor had matured but it did not have multiple necrosis; four rats were studied between 12 to 15 days, when extensive hemorrhages and necrosis appeared in the tumor. The study was approved by the Ethics Committee of the Nizhny Novgorod State Medical Academy (protocol no. 14 from December 10, 2013).

2.2.

CP OCT Device

A spectral domain CP OCT device (Institute of Applied Physics of the Russian Academy of Sciences, Nizhny Novgorod, Russia) with cross-polarization detection was used in the study.^32,33 The central wavelength of this device is 1310 nm with a radiation power of 20 mW. The axial resolution is $10\mu \mathrm{m}$ and the lateral resolution is $15\mu \mathrm{m}$ in air. The probing beam uses circular polarization. The device has a 20,000 A-scans/s scanning rate and performs 2-D lateral scanning within an area of $2.4\times 2.4{\mathrm{mm}}^2$ to obtain the 3-D distribution of backscattered light in the polarization with the same and reversed rotations of the electric-field vector.³¹ Scanning was performed in contactless mode. An advantage of the CP OCT device used is its real-time image formation. Structural 2-D (cross-sections and en-face images) in co- and crosspolarization and a 3-D image in copolarization are displayed on the personal computer monitor in the process of scanning, thereby facilitating assessment of the image quality.

2.3.

Study Design

This study describes an approach for the calculation of CP OCT optical coefficients from different brain areas including the cortex, white matter, and glioma tissue and the construction of color-coded maps based on them.

3-D CP OCT images of the brain were acquired ex vivo immediately after brain excision and crosscutting in the coronal section. The plane of the brain crosscutting for obtaining the CP OCT scans of various brain structures was fixed at a distance of $2.6\pm 3.0\mathrm{mm}$ behind the bregma, allowing comparison of analogous structures in healthy brains with those in the brains with tumors. Sectioning was performed at two sites in this plane: (1) through the center of the tumor to visualize tumor margins with white matter and cerebral cortex and (2) closer to the edge of the tumor for detecting infiltration of the white matter (corpus callosum) by the tumor in greater detail.

A sequential scanning of the entire brain surface was performed row by row with overlapping images (from the cerebral cortex, corpus callosum, and hippocampus to the amygdala and hypothalamus) with further digital processing: building individual color-coded maps—an en face distribution of the optical coefficients values following their reconstruction into a single, complete picture of the rat brain. Between 32 and 46 CP OCT data sets were collected from each rat brain. Each imaging data set consisted of $256\times 256$ A-scans forming the OCT imaging volume. For each collected A-scan set, the optical coefficients were calculated according to the method described in the following section (see Sec. 2.4). After CP OCT scanning, histological evaluations of all the samples were performed.

To estimate whether the optical coefficients would increase the OCT contrast, the color-coded maps were compared with initial structural en-face OCT images in co- and crosspolarization and with the histological data. Unprocessed en-face OCT images were saved from the 3-D volume at a depth of $110\mu \mathrm{m}$ below the tissue surface to avoid image artifacts like glare from the surface (especially from the white matter), while preserving the contrast between the tissues, which rapidly decreases with depth. A-scan realignment was performed to ensure coefficient calculation from the same depth range.

2.4.

Calculation of Optical Coefficients

Given that brain tissue is optically homogeneous, the OCT signal in both polarization channels is attenuated exponentially with depth:

Note that $u$, ($u-uD$) are the attenuation coefficients in the corresponding channels and $z$ is the depth coordinate. Cochannel attenuation coefficient $u$ and interchannel attenuation difference $uD$ were used. The choice of the interchannel attenuation difference instead of crosschannel attenuation was made to highlight possible situations in which the crosschannel signal decays more rapidly than the cochannel signal.

The co- and crosschannel attenuation coefficients can be obtained by linear fitting of the logarithms of the corresponding signals and used for quantitative analyses of the CP OCT data. For each collected A-scan set, the optical coefficients were calculated in the same predefined depth range starting from $\sim 70\mu \mathrm{m}$ below the tissue surface (pixel #10) to a depth of $\sim 350\mu \mathrm{m}$ below the tissue surface (pixel #50). In addition to the conventionally used cochannel attenuation coefficient of the samples, the difference in attenuation between the two orthogonal polarization channels was also investigated as a potential predictor of the state of the brain tissue. This coefficient was referenced in this study as the interchannel attenuation difference ($uD$). The A-scans of a white matter sample in co- and crosspolarization on a logarithmic scale and with corresponding linear fits are shown in Fig. 1. En-face color-coded maps of different tissue types were created based on the calculated optical coefficients (cochannel attenuation and interchannel attenuation difference).

Fig. 1

A-scan examples and corresponding values of cochannel attenuation ($u$) and interchannel attenuation difference ($uD$) for various types of brain tissues. (a) Cortex with negative interchannel attenuation difference; (b) cortex with positive interchannel attenuation difference; (c) white matter; and (d) glioblastoma 101.8. Dashed vertical line represents the tissue surface. Coefficients’ calculations were made in the range of 70 to $350\mu \mathrm{m}$.

Optical coefficients were recorded for the 3-D CP OCT images in four groups: group 1, normal gray matter ($n=30$); group 2, normal white matter ($n=30$); group 3, glioblastoma 101.8 without necrosis ($n=16$), and group 4, glioblastoma 101.8 with necrosis ($n=16$). 3-D CP OCT images of normal tissues were obtained from the five rats with healthy brains (two sets of CP OCT images from the right and left hemispheres were scanned from each of two brain subjects after brain incision and coronal cutting). To quantify the tumor tissue, CP OCT data sets containing only tumor tissue were taken (images including tumor margins with gray or white matter were excluded). These images were obtained from the eight rats with inoculated glioblastoma 101.8 (two sets of CP OCT images from the right hemisphere were scanned from each of two brain subjects after brain incision and coronal cutting).

2.5.

Histology

After the OCT scanning, brain samples were fixed in 10% formalin for 48 h and a series of histological sections were made. Thus, the en-face plane of the CP OCT images coincided with the plane of the histological sections. The histological sections were stained with hematoxylin and eosin (H&E) and Luxol fast blue stain with crezyl violet (to identify both myelinated fibers and the neuronal tissue structures) and evaluated by two independent histopathologists using light microscopy (Leica DM2500, Germany). Agreement between the specialists was revealed in 98% of cases.

2.6.

Statistical Analysis

The median value among all values of the optical coefficients calculated for each A-scan of a 3-D CP OCT image was used. The results are expressed as the Me [Q1; Q3] where Me is the median of the coefficient and [Q1; Q3] are the 25th and 75th percentile values, respectively. To compare the tissue type using each coefficient, we used multiple comparisons with the Bonferroni correction. The statistical data processing was done in MS Excel 2010 with a public domain software plugin for statistical analysis STATISTICA 10 (StatSoft, Inc., Tulsa, Oklahoma) and the Prism v6 software (GraphPad Software, La Jolla, California).

3.1.

Unprocessed CP OCT Images versus Color-Coded Maps of Normal Brain

The first part of the study was aimed at illustrating the different sensitivities of the optical coefficients in the visualization of healthy gray and white matter. Due to the structural complexity of the brain, a sequential scanning of the whole brain slice was performed to identify the various brain structures in comparison with the histological data and a stereotaxic rat atlas.³⁴ This provided high-precision identification of the brain areas. Then, to reveal the most sensitive/specific coefficients for distinguishing different brain tissue types, several regions of interest were chosen.

Representative wide-field CP OCT images of a healthy rat brain are shown in Fig. 2. Unprocessed (native) OCT images in co- [Fig. 2(b)] and crosspolarization [Fig. 2(d)] show evident optical differences between white matter structures (such as the corpus callosum, internal capsule, striatum, and optic tract) and gray matter (such as the cerebral cortex, hippocampus, and thalamus). The color-coded maps [Figs. 2(c) and 2(e)] provide much better visualization and differentiation of the white and gray matter that are schematically shown in Fig. 2(a). The cochannel attenuation coefficient [Fig. 2(c)] is seen to be more sensitive to the density of the scattering structures, which is why the highest values (red color) correspond to the white matter tracts (densely packed myelinated fibers); medium values (green and yellow colors) correspond to areas of accumulation of myelinated fibers or to individual, spreading ones and to gray matter with high neuronal density; the remaining structures (bodies of neurons and glia) have low values of cochannel attenuation coefficient (dark blue color). The interchannel attenuation difference [Fig. 2(e)] is more sensitive to elongated scattering structures, providing excellent visualization of white matter areas.

Fig. 2

Wide-field unprocessed CP OCT images and color-coded maps of healthy brain (coronal cross section). (a) Schematic view of the rat brain showing the relationship between gray matter (white color) and white matter (gray color).³⁴ The dark gray color marks brain structures with densely packed myelinated fibers and tracts, while the bright gray color indicates areas with thin myelinated fibers or single bundles. (b)–(e) Wide-field brain images reconstructed from unprocessed images in (b) co-, and (d) crosspolarization; (c), (e) the corresponding color-coded maps: (c) cochannel attenuation coefficient, (e) interchannel attenuation difference. Color-coded maps based on the calculated optical coefficients provide better contrast for clear visualization of the different brain structures. Black dotted boxes (1, 2, 3) indicate regions of interest for detailed analysis (see comments in the text).

We chose three different regions of interest in Fig. 2 to demonstrate the best correspondence of the interchannel attenuation difference (red-color) for visualizing the white matter in comparison with the unprocessed images and even the cochannel attenuation maps. Box #1 shows the margin between the cerebral cortex (stratified gray matter) and, located directly below it, white matter (corpus callosum) with strongly oriented fibers. This region is of vital neurosurgical interest because the predominant localization of glial tumors in humans is subcortical with a tendency to extensive growth toward the cerebral cortex and germination through it. Variations in the CP OCT color-coded maps of normal cortex and white matter are shown in Figs. 5(a) and 5(c), upper row. Results of the quantitative analysis of these tissue types are shown in Figs. 5(b) and 5(d) and in Table 1. The cortex was well differentiated from the white matter using both cochannel attenuation and the interchannel attenuation difference ($u_{\mathrm{WM}}=9.39$ $[8.77;9.85]{\mathrm{mm}}^{-1}$ and $u{D}_{\mathrm{WM}}=4.20$ $[3.91;4.97]{\mathrm{mm}}^{-1}$ versus $u_{\mathrm{CTX}}=2.25$ $[2.03;2.51]{\mathrm{mm}}^{-1}$ and $u{D}_{\mathrm{CTX}}=0.51$ $[0.24;0.63]{\mathrm{mm}}^{-1}$, $p<0.001$).

Table 1

Optical coefficients’ difference between tumorous and nontumorous brain tissue types.

Two other regions of interest (boxes #2 and #3) were chosen to stress the high level of contrast between structures that contain only gray matter or only white matter and that these can be clearly identified using the optical coefficients. In box #2, the hippocampus, consisting of gray matter (low values of optical coefficients, blue color), is clearly delineated because of being surrounded by white matter having higher values of the optical coefficients (yellow-green and red colors). Box #3 demonstrates that using the cross-scattering coefficient, isolated white matter bundles (left and right crus of the fornix) passing through the gray matter can be detected.

In summary, comparison of the color-coded maps and unprocessed CP OCT images is indicative of the benefit of using optical coefficient calculation (postprocessing) for higher-contrast imaging of the different brain structures. The interchannel attenuation difference has high specificity for the visualization of white matter containing myelin fibers when compared to the cochannel attenuation coefficient.

3.2.

Unprocessed CP OCT Images Versus Color-Coded Maps of the Brain in the Glioblastoma Model

The next step was to determine the differences in optical properties between nontumorous and tumorous brain tissue using CP OCT color-coded maps. For this purpose, the tumor core tissue and its margin with infiltration into surrounding white and gray matter were analyzed. Identification of glioma infiltration is quite challenging, but is a very important task during glioma surgery.

Wide-field structural CP OCT images and color-coded maps of the brain samples with 101.8 glioblastoma were reconstructed and compared carefully with histological data. Several regions of interest, including the tumor core, tumor margin, and surrounding brain tissue, were identified by a pathologist and then analyzed in detail in the CP OCT data.

Figure 3 shows the potential of CP OCT for the tumor core detection and its margin with the cortex. Morphologically, tumor tissue has a heterogeneous structure mainly due to the presence of multiple small or large areas of hemorrhages and necrosis [Figs. 3(e) and 3(f)]. In unprocessed CP OCT images, the tumor tissue appears homogeneous and darker compared to the gray matter [Figs. 3(a) and 3(b)]. By contrast, variegated OCT color-coded maps clearly demonstrate tumor heterogeneity [Figs. 3(b) and 3(d), bright-green spots on dark blue background] compared with the more or less homogeneous OCT signal from normal gray matter [Figs. 2(c) and 2(e)], Fig. 5(a), upper row, dark blue color] and white matter [Figs. 2(c) and 2(e)], Fig. 5(a), upper row, bright stripe with colors from green to red]. It should be noted that by visual assessment of the color-coded maps, the cochannel attenuation coefficient is more sensitive for the detection of tumor cell clusters, blood, and necrotic areas than is the interchannel attenuation difference. Therefore, cochannel attenuation maps appear to be better for the visualization of tumor tissue.

Fig. 3

Ex vivo CP OCT identification of glioblastoma with necrotic areas and its margin with the cerebral cortex. Unprocessed OCT images (a) in co- and (c) in crosspolarization. Color-coded maps: (b) the cochannel attenuation coefficient; (d) interchannel attenuation difference. (e), (f) Histological images with Luxol blue and crezyl violet staining. The area of interest in (e) is presented under high magnification in (f). T, tumor; GM, gray matter; H + Nec, hemorrhage and necrosis. For comparison with the corresponding histology, en-face (same orientation as the color-coded maps) has been provided. Scale bars 1 mm.

In the case of extensive infiltration of the normal brain tissue with glioblastoma cells [Fig. 4(e)], the myelin fibers are destroyed [Fig. 4(f)]. In the interchannel attenuation difference maps, the OCT signal of the area involving fragmented white matter is significantly reduced [Fig. 4(d), infiltrated white matter—infWM] compared to the white matter of the peritumoral area and healthy brain tissue [Fig. 2(e), Fig. 5(a), upper row, colors from green to red]. In the cochannel attenuation maps, the OCT signal from the white matter can indicate high values close to those of healthy tissues and the outer contour of the white matter area can be observed [Fig. 4(b), infWM].

Fig. 4

Improving OCT contrast by building color-coded maps based on optical coefficients in the detection of brain tissue infiltration by glioblastoma cells. Unprocessed OCT images (a) in co- and (c) in crosspolarization. Color-coded maps: (b) the cochannel attenuation coefficient; (d) interchannel attenuation difference. (e), (f) Histological images with Luxol blue and crezyl violet staining. The area of interest in (e) is presented under high magnification in (f). Pink arrow indicates the transition point of the most destroyed and preserved myelinated fibers. T, tumor; infGM, infiltrated gray matter; infWM, infiltrated white matter. For comparison with the corresponding histology, en-face (same orientation as the color-coded maps) has been provided. Scale bars 1 mm.

Fig. 5

Variability of color-coded maps of (a) cochannel attenuation and (c) interchannel attenuation difference in four groups of tissues: normal cortex (group 1), normal white matter (group 2), glioblastoma without necrosis (group 3), and glioblastoma with necrosis (group 4). The distribution of (b) cochannel attenuation and (d) interchannel attenuation difference coefficient values between different tissue types. Data are presented as medians with 25th and 75th percentile values. *Significant differences between white matter and other groups ($p<0.001$); #significant differences between the cortex and groups with glioblastoma ($p<0.001$). Scale bars 1 mm.

The infiltration of gray matter leads to significantly increased cellular density and due to this, the scattering properties of the cortex are locally increased. This is manifested in the cochannel attenuation maps by localized increases of the OCT signal values and makes the cortex appear more heterogeneous [Fig. 4(b), infiltrated gray matter, infGM] compared to a healthy equivalent [Fig. 2(c), Fig. 5(a), upper row, dark blue color]. Conversely, in the interchannel attenuation difference maps, the cortex infiltration is not obvious [Fig. 3(d), infGM].

The cochannel attenuation and interchannel attenuation difference values for glioblastoma 101.8 without necrosis were 3.72 $[3.51;4.17]{\mathrm{mm}}^{-1}$ and 1.36 $[1.32;1.58]{\mathrm{mm}}^{-1}$, respectively; for glioblastoma 101.8 with necrosis 6.53 $[5.72;7.10]{\mathrm{mm}}^{-1}$ and 2.01 $[1.91;2.59]{\mathrm{mm}}^{-1}$, respectively (Table 1). The results show that by using both coefficients, tumor tissue can be well distinguished from the white matter ($p<0.001$) and from the cortex ($p<0.001$) [see Figs. 5(b) and 5(d)], but the ratio of the interchannel attenuation difference of the white matter and other tissue groups is higher than for the cochannel attenuation coefficient (Table 1), indicating the higher specificity of this coefficient for white matter in differentiating from other tissue types in comparison with the cochannel attenuation coefficient.

Interchannel attenuation difference detects the same tendency for differences of other tissue groups from the cerebral cortex. Therefore, this coefficient can serve as a very promising tool for intraoperative normal brain tissue detection in glioma surgery, especially if the glioblastoma has extensive necrotic areas.