2.1.

Preparation and Imaging of Animal Model

Female CD1 WT mice, 6–14 weeks old, were anesthetized with an intraperitoneal (IP) injection of ketamine and xylazine ($10\mathrm{mg}/\mathrm{kg}$ xylazine, $80\mathrm{mg}/\mathrm{kg}$ ketamine) and prepared with a thinned-skull cortical window, approximately $4\times 4{\mathrm{mm}}^2$ wide, centered over the right cerebral cortex. The thinned skull allowed for increased light penetration depth while preserving the natural physiological conditions of the brain.¹¹ The mouse was then secured in a stereotactic frame for stable OCT imaging over a select region of interest within the cortical window.

An acute water intoxication model of cytotoxic cerebral edema, in which water is driven from the vascular system into the brain by an osmotic gradient, was used.^12,13 Twenty minutes of baseline OCT imaging was performed after which an IP injection of water was given (dosage 40% of the animal’s body weight). OCT imaging continued for 60 min. One OCT volume was collected every 3 min and spanned a volume of $4\times 4\times 2{\mathrm{mm}}^3$. Three water intoxication experiments and three control experiments (including thin-skull preparation without IP injection) were performed. All experimental procedures and protocols were approved by the University of California Riverside Institutional Animal Care and Use Committee (protocol 2010-0018).

2.2.

Brain Water Content Analysis

The presence of cerebral edema was verified by analyzing BWC via the wet-dry methods. After OCT data collection was completed, animals were sacrificed and the brains were dissected out post mortem and cut into four pieces (right-left cerebral hemispheres, right-left cerebellum). Brains were weighed to obtain the wet weight and dried for 48 h to determine the dry weight. The percent of water content was calculated with the following formula: $\mathrm{BWC}=[(\text{wet weight}-\text{dry weight})/\text{wet weight}]\times 100$.

2.3.

Optical Coherence Tomography Image Acquisition

A spectral-domain OCT system centered at 1300 nm was used to image the in vivo mouse brain [Fig. 1(a)]. The light source consisted of two superluminescent diodes with a combined center wavelength of 1298 nm and a 120-nm full width at half maximum bandwidth. The source light was sent through a fiber optic circulator and an $80/20$ splitter, which divided the light between the sample and reference arms. The light that reflected from a mirror in the reference arm and from the brain tissue in the sample arm was then recombined at the splitter, dispersed by a grating, and detected on the line scan camera. Each spectrum acquired on the camera contained depth information from a single location in the sample (known as an A-line or depth profile). 3-D imaging was achieved by raster scanning the optical beam across the sample using two galvanometer mounted mirrors. The system has an axial imaging range of approximately 2 mm in biological tissue, an axial resolution of $8\mu \mathrm{m}$, and a lateral resolution of $20\mu \mathrm{m}$.

Fig. 1

(a) Schematic of spectral domain optical coherence tomography (OCT) setup. Red arrows depict the optical beam scan pattern for three-dimensional 3-D imaging of sample. m: mirror, gm: galvanometer mounted mirror, gr: grating, lsc: line scan camera. (b) Two-dimensional OCT sagittal image of in vivo mouse brain. S: skull, CTX: cerebral cortex, CC: corpus callosum. $\text{Scale bar}=0.5\mathrm{mm}$. (c) 3-D volume of in vivo mouse brain rendered from OCT volumetric scan.

Each image contained 2048 A-lines, acquired at a rate of 15 kHz, and 250 images were used to render a single volume [Figs. 1(b) and 1(c). Real-time visualization of the sample was achieved using in-house developed software, which utilized a graphics processing unit (GPU) to process and display the OCT structure and blood flow images in real time.¹⁴ The real-time display of OCT images allowed structural markers within the imaging location to be rapidly identified and ensured that the same region of the brain was consistently imaged during and across all experiments. Depth-resolved structural information was retrieved from each A-line by applying a Fourier transform to the spectrum collected on the camera after linearly remapping to $k$-space. The images were displayed on an inverse log scale [Fig. 1(b)].

Images of the cerebral vascular network were generated by utilizing the phase information contained in the OCT signal, a method known as Doppler OCT (DOCT).^9,15,16 This technique enables high resolution imaging of motion within the sample by calculating the Doppler frequency shift caused from reflection off of a moving sample. The depth resolved phase difference between consecutive A-lines was determined and this phase difference is directly proportional to the axial component of the velocity of the scattering particle. Static tissue introduces minimal phase shifts and is represented by a low intensity in DOCT images while blood flow induces a large phase difference, represented by a high intensity.

2.4.

Analysis of Attenuation Coefficient and Cerebral Blood Flow

The attenuation coefficient of a sample describes how strongly the sample scatters and absorbs light. In this study, change in the tissue’s average attenuation coefficient was used to quantify the effects of cerebral edema in the cerebral cortex. The attenuation coefficient was determined on a pixel-by-pixel basis, using a recently developed model for depth-resolved attenuation coefficient determination in OCT data.¹⁷ This method makes it possible to create an attenuation image, in which contrast is indicated by the local attenuation coefficient from an OCT intensity image with a known pixel size (Fig. 2). The tissue was assumed to have an index of refraction of 1.4.

Fig. 2

(a) Sagittal optical coherence tomography (OCT) intensity image of in vivo mouse brain and (b) corresponding attenuation image. $\text{Scale bar}=0.5\mathrm{mm}$.

The average attenuation coefficient of the cortex was determined over time as edema progressed in the animal model. All pixels contained in the regions of interest (ROI) and within 0.7 mm of the surface (gray matter region) were included in the average calculation. The average attenuation coefficient during baseline was determined from all data acquired during the first 20 min and the percent change in the average attenuation coefficient was calculated at all points in subsequent volumes.

Cerebral blood flow was analyzed by calculating the percent change in flow density using the DOCT image information. A threshold was set to distinguish the pixels that represent flow from background pixels in the DOCT data, within the same ROIs that were used for attenuation analysis. The total DOCT intensity was calculated as the sum of all intensities in pixels identified as flow regions and divided by the number of pixels in the ROI to determine the flow density. The percent change in flow density was then calculated for each time point in order to determine how the cerebral blood flow was changing over time. Maximum intensity projection (MIP) images of the cerebral blood flow were displayed in an en face orientation and filtered using a Gabor filter.¹⁸

3.1.

Optical Changes from In Vivo Cerebral Edema Model

The percent change in attenuation coefficient was determined from the OCT intensity data within a $3\times 3{\mathrm{mm}}^2$ ROI over the cerebral cortex. The areas close to the cortical window boundaries were excluded from the calculation because of uneven bone thicknesses in these regions. The average value of the attenuation coefficient in the cerebral cortex remained steady during baseline and then decreased soon after the IP injection was administered at 20 min (Fig. 3). The decrease continued approximately linearly in time as cerebral edema progressed. A maximum decrease of 8% was observed by the end of the experiment (at which time standard error of the mean was 1.93). Over the same amount of time, the average attenuation coefficient from the control experiments remained relatively constant (Fig. 3). The BWC measurements confirmed that the cerebral edema group exhibited a higher BWC compared to the control group (Fig. 4).

Fig. 3

Percent change of average attenuation coefficient in the cerebral cortex over time from in vivo cerebral edema mouse model. Arrow indicates IP water injection for water intoxication group. Error bars represent standard error for $n=3$.

Fig. 4

BWC measurements of right cerebral cortex quadrant. Error bars represent standard error.

The attenuation coefficient within the cerebral cortex exhibited some variation as a function of depth. As an example, an average depth profile from an area of approximately $200\times 200\mu \mathrm{m}$ within the cerebral cortex is shown in Fig. 5. During baseline, the attenuation coefficient had a median value of $1.2{\mathrm{mm}}^{-1}$. After water intoxication, this value decreased to approximately $1.1{\mathrm{mm}}^{-1}$. The percent difference between these two profiles as a function of depth is shown of the right $y$-axis and resulted in an average percent difference of $-12\%$.

Fig. 5

Attenuation depth profile from a small region of cerebral cortex. The gray line represents the values during baseline and the black line represents the same depth profile after water intoxication. The percent difference between the two profiles is shown by the red dotted line on the right $y$-axis. The average percent difference over the entire depth is $-12\%$.

The rate of change of the attenuation coefficient was calculated for the baseline (defined as the first 20 min of the edema experiment) and for edema (postIP injection) by fitting the data using linear regression (Fig. 6). These rates of change were also compared with the control experiments. A $t$-distribution test was performed to compare the rates of change. Statistical analysis identified that the rate of change during edema was statistically significant ($p<0.05$) from both the control ($p=0.0148$) and the baseline ($p=0.0135$). As expected, the rate of the change in the attenuation coefficient during the control was not statistically different from baseline. These results indicate that the progression of edema can be identified by measuring a rate of change in the attenuation coefficient.

Fig. 6

Rate of change in average attenuation coefficient during baseline, edema (postIP injection) and control experiments. Error bars represent the standard error.

Consistent with previous reports using this mouse model, the severity of cerebral edema monotonically increased over time.⁶ Analysis of the cerebral blood flow confirmed that flow decreased near the end of the experiment before the animal expired (Fig. 7). The flow initially increased quickly after the IP injection was administered, presumably due to the injection itself increasing circulating blood volume, and small capillaries became increasingly visible. After approximately 45 min postIP injection, the flow began to slow; at the end of the experiment minimal flow was visible in the MIP images (Fig. 7). The decrease in flow is consistent with the Monro-Kellie doctrine, which states that the cerebral blood volume decreases to compensate for an increase in brain volume.¹⁹ Using DOCT, we are able to detect when this volume compensation process is occurring.

Fig. 7

(a) MIP en face images of the cerebral vascular network created from DOCT volumetric data. The orientation is specified by the rostral (R) and midline (M) arrows. Time points (in minutes) are labeled in the bottom left corners. The red arrow indicates an area where capillaries became increasingly visible after IP injection compared to baseline. (b) Percent change in flow density as a function of time for water intoxication and control animals.

3.2.

Local Attenuation Coefficient Trends

The spatial specificity afforded by volumetric OCT data allowed for the optical properties of local regions to be interrogated. The imaging volume was divided into smaller ROIs to investigate the attenuation coefficient trends as (1) a function of distance from midline, (2) from rostral to caudal and (3) axial location. The water intoxication model used in these experiments is a global model of cerebral edema, causing the entire brain to become edematous. The attenuation trends within these localized regions confirmed the global nature of this model, as each of the ROIs demonstrated the same decrease in attenuation following water intoxication with no statistical difference at various sagittal, coronal, or axial positions (Fig. 8). The ability to interrogate local regions and assess how specific areas within the brain are affected by edema with OCT illustrates the value of optical detection over standard methods of ICP and BWC measurements, which only offer a global assessment.

Fig. 8

Percent change of average attenuation coefficient in the cerebral cortex from three regions spanning (a) medial-lateral, (b) rostral caudal and (c) axial positions (ROIs shown in inset). Arrow indicates IP injection time point. Error bars represent standard error.

The blood flow in the same local ROIs was analyzed for comparison. The blood flow directly affects the formation of edema because the excess water enters the brain from the vascular network. Throughout all localized ROIs, the blood flow initially increased after the IP injection and then fell below baseline level 1 h after injection (Fig. 9). The similarity in the flow densities as a function of time revealed that blood flow in all areas of the brain changed in the same way and that compression of blood vessels occurred at the same time within the ROIs.

Fig. 9

Percent change in flow density as a function of time as measured in the three local ROI groups. Arrow indicates IP injection. Error bars represent standard error.