The cerebral cortex consists of six cortical layers characterized by their histological, neurochemical, and neurophysiologic differences. Such differences lead to heterogeneous neurovascular responses of different cortical layers to various brain stimuli, which have been attracting tremendous interest in brain functional studies [1]. However, to understand the hemodynamic response of each cortical layer to a brain stimulation, there is demand for new neuroimage modalities that can provide not only high spatiotemporal resolution imaging quantification of cerebral blood flow velocity (CBFv) in capillaries, but also a large field of view (FOV), especially sufficient image depth (e.g., $>800\mathrm{\mu m}$), to track the CBFv network dynamics. f-MRI has been widely used for whole-brain functional studies [2], but a higher resolution is required in more and more applications to characterize cerebral hemodynamic changes in individual vascular compartments elicited by neuronal activations. Two-photon microscopy (TPM) has been widely applied for high-resolution 3D imaging of cerebral microvasculature, and a recent study demonstrated extended penetration depth to $>1\mathrm{mm}$ in the mouse cortex with newly developed 1280 nm TPM [3]. Yet, TPM requires fluorescence tracking which may complicate pharmacological studies, and quantitative CBFv imaging is usually limited by a small FOV (e.g., single or few micro vessels). Photoacoustic microscopy was recently reported to enable high-resolution cerebral microvasculature and blood oxygenation imaging, which is an important brain functional study [4].

Recent advances in optical coherence-domain techniques, such as optical microangiography (OA) and optical coherence Doppler tomography (ODT), have permitted tracker-free 3D visualization of cerebrovasculature and quantitative CBFv imaging of vessels of different caliber in the mouse cortex [5,6]. For instance, ultrahigh-resolution ODT (μODT) has been shown to image microcirculatory CBFv network dynamics and identify laser disruption of local microvasculature, and, more importantly, the spreading of spontaneous microischemia elicited by repeated cocaine administration, which requires both high flow detection sensitivity and large FOV [7]. It is noteworthy, however, that because of limited image depth, the insightful heterogeneous CBFv network dynamics in different cortical layers in response to various brain activations is not yet fully understood.

To tackle the challenge, we developed a new μODT system at 1310 nm where the reduced light scattering of biological tissue may lead to increased penetration depth for flow imaging. An ultra-broadband light source ($\lambda =1310\mathrm{nm}$, ${\lambda }_{\mathrm{FWHM}}=220\mathrm{nm}$) was used to illuminate a spectral-domain OCT engine consisting of a $2\times 2$ broadband fiberoptic Michelson interferometer. Light exiting the reference arm was collimated, propagated through a pair of wedge prisms for dispersion compensation, and focused on a retroreflective mirror to maximize the bandwidth ($\mathrm{\Delta }{\lambda }_{\mathrm{cs}}\approx 200\mathrm{nm}$) of the cross-spectrum (i.e., $S_{\mathrm{cs}}(\lambda )\equiv {[S_s(\lambda )·S_r(\lambda )]}^{1/2}$, where $S_s(\lambda )$, $S_r(\lambda )$ were the sample and reference power spectra), so that a high axial resolution of 2.5 μm in brain tissue was reached, which is defined by the transform-limited coherence length, $L_c=2(\ln 2)/\pi ·{\lambda }_0^2/\mathrm{\Delta }{\lambda }_{\mathrm{cs}}$. In the sample arm, the collimated light was transversely scanned by a fast servo mirror and focused on the mouse cortex through a cranial window with a $f16\mathrm{mm}/\mathrm{NA}0.25$ NIR objective, yielding a lateral resolution of 3.2 μm. The backscattered light from the mouse cortex was recombined with the reference light in the detection fiber connected to a custom high-resolution spectrometer where the interference fringes spectrally encoding the depth profile were detected by a high-speed linescan InGaAs camera (2048 pixels, 145 k lines/s; GL2048, Sensors Unlimited) synchronized with sequential transverse scans for 2D/3D μOCT acquisition. Focus tracking was implemented to enable full cortex imaging by incrementing a focal plane 200–300 μm in $z$ (depth) direction. A cross-sectional μODT image ${\nu }_{z,x}$ was reconstructed from the measured phase difference $\mathrm{\Delta }{\varphi }_{z,x}$ between adjacent A-scans by a phase subtraction method (PSM: ${\nu }_{z,x}=\lambda \mathrm{\Delta }\phi /(4\pi n\tau \cos \theta )$, where $\theta$ is the incident angle with flow, $n$ is the refractive index of the brain, and $\tau$ is the duration between two A-scans) or by a phase-intensity method (PIM) [7,8]. Since this process was computationally intensive, a graphic processing unit (GPU) with custom GUI programming was implemented to boost FFT and phase detection, which allowed for real-time rendering and display of maximum intensity projection (MIP) of μODT images, e.g., as fast as 473 fps for a B-scan containing $1k\times 2k$ pixels. Therefore, the dynamic features of the CBFv network were readily visualized during imaging [9,10].

In parallel, to assess the image-depth improvement of the new 1310 nm μODT for deep flow detection, a prior 800 nm μODT system was included for comparison, whose system parameters were an axial resolution of 1.8 μm in tissue (illuminated with an 8 fs $\mathrm{Ti}\text{:}{\mathrm{Al}}_2{\mathrm{O}}_3$ laser with ${\lambda }_0=800\mathrm{nm}$, $\mathrm{\Delta }{\lambda }_{\mathrm{cs}}=154\mathrm{nm}$), transverse resolution of 3.0 μm, and an A-scan rate of up to 27 kHz [7]. To avoid phase saturation and low sensitivity for capillary flow imaging, A-line rates for 1310 nm μODT were set to 6 and 3 kHz for imaging the upper (0–200 μm) and lower ($>200\mathrm{\mu m}$) cortex, respectively; the equivalent A-line rates for 800 nm μODT were utilized based on ${\lambda }_{800\mathrm{nm}}·f_{800\mathrm{nm}}={\lambda }_{1310\mathrm{nm}}·f_{1310\mathrm{nm}}$ to match phase shift detected by the two systems.

Eight-week-old CD-1 mice were used for comparative study during which they were anesthetized with inhalational 2% isoflurane and mounted onto a custom stereotaxic frame. A $3\mathrm{mm}\times 3\mathrm{mm}$ cranial window was surgically created on the mouse sensorimotor cortex with the dura remained intact [11]; the exposed cortex was filled with 1% agarose and sealed by a glass coverslip to minimize motion-induced noise and artifacts.

Figure 1 compares 3D CBFv images of the mouse sensorimotor cortex acquired by 800 vs 1310 nm μODT. Although MIP images show almost no difference, except slightly higher capillary density in Fig. 1(b), likely because of more detectable and deeper capillary flows, their 3D renderings in Figs. 1(a) and 1(b) clearly demonstrate the increase in CBFv image depth from $\sim 300\mathrm{\mu m}$ for 800 nm μODT to $\sim 1.4\mathrm{mm}$ for 1310 nm μODT, namely, over a four-time improvement. This is crucial because the new technique enables quantitative imaging of the CBFv networks across the entire cortex (I-VI layers), thus potentially allowing us to quantify their dynamic heterogeneity in response to functional brain activations. Detailed image analyses presented in Figs. 1(c)–1(f) clearly indicate heterogeneous CBFv distribution along the depth in the cortex, e.g., 0–0.2 mm, predominantly occupied by large, fast pial flows (c, layers I-II); 0.2–0.4 mm, occupied mostly by densely populated short capillary flows likely oriented perpendicularly (d, layers III- top IV); 0.4–0.6 mm, slightly less populated but longer capillary flows likely oriented more toward lateral directions (e, layers IV-V); and 0.6–1.4 mm, continuously less populated, slower capillary flows (layer VI, corpus callosum). Figure 1(g) plots the statistical figure of the capillary flow networks in different cortical layers, showing that the capillary CBFv continued to increase from $0.14\pm 0.05\mathrm{mm}/\mathrm{s}$ on top to $0.20\pm 0.04\mathrm{mm}/\mathrm{s}$ at $\sim 600\mathrm{\mu m}$, but then dramatically decreased to $0.08\pm 0.04\mathrm{mm}/\mathrm{s}$ within the bottom cortical layers. This rapid decrease of capillary CBFv in deep cortex (e.g., 0.6–1.4 mm) could be caused by either the system sensitivity loss or the inherent CBFv decrease in the microcirculatory networks, or by their confounds. Thus, further study may be needed to clarify this.

 

Fig. 1. In vivo 3D CBFv networks on the mouse sensorimotor cortex ($2.4\mathrm{mm}\times 2.0\mathrm{mm}$) imaged by 800 nm μODT vs 1310 nm μODT. (a), (b) 3D rendering to illustrate dramatically enhanced image depth by 1310 nm μODT. (${\mathrm{a}}^{\prime }$) and (${\mathrm{b}}^{\prime }$) Corresponding MIP images. (c)–(f) En face CBFv images of sub-stack cortex at different depths from 0 to 1.4 mm. (g) Capillary CBFv at different depths. CTX, cortex; CC, corpus callosum. Low-dose intralipid ($0.5\mathrm{mg}/\mathrm{kg}/\mathrm{h}$) was given to the animal intravenously as an optical contrast agent to enhance capillary flow sensitivity for panels (a) and (b).

Download Full Size | PDF

In addition to enhancing the CBFv network image in Fig. 1, the increased image depth of 1300 nm μODT may allow us to analyze microvasculatural heterogeneity in the mouse cortex. Low-dose intralipid ($0.5\mathrm{mg}/\mathrm{kg}/\mathrm{h}$) was given to the animal intravenously for flow contrast enhancement. Thus, cortical microvasculature was imaged by contrast enhanced μODT (c- μODT) which allows for high SNR and fewer shadow effects for quantifying capillary density, especially at a deep cortical area. Figure 2 shows a side view of a 3D image of cerebrovasculature on the mouse sensorimotor cortex (a: $2.0\mathrm{mm}\times 0.5\mathrm{mm}\times 1.4\mathrm{mm}$). Interestingly, large pial vessels aligned along the cortical surface, as well as those penetrating perpendicularly downward are readily seen and highlighted with light blue arrows.

 

Fig. 2. 3D cerebrovasculature of the mouse sensorimotor cortex ($2.0\mathrm{mm}\times 0.5\mathrm{mm}\times 1.4\mathrm{mm}$) acquired by 1310 nm contrast-enhanced μODT. Right panel, statistical distribution of microvasculatural density (fill factor) as a function of depth below the cortical surface. Arrows, penetrating pial vessels.

Download Full Size | PDF

In addition, the capillary beds were distributed highly heterogeneously in different cortical layers. To quantify the density distribution of the microcirculatory vessels, capillary skeletons were extracted automatically based on a distance map (e.g., $\varphi <10\mathrm{\mu m}$) in the MIP image so that the fill factor defined as the pixel #s occupied by the capillaries vs the total pixel #s, excluding those occupied by large vessels, was computed to characterize the capillary density [12]. The right panel plots the result, which shows that the capillary fill factor first increased from 0 to 400 μm (layers I-III), reached a plateau within 400–700 μm, and then gradually became sparsely populated from 700 to 1200 μm as it approached the deeper areas (layers V-VI and below). This type of capillary density patterns is consistent with the TPM measurement previously reported [13]. The combined merits of quantitative flow imaging and large FOV that covers the full-thickness cortical CBFv networks may place the new 1310 nm μODT technique uniquely well suited for various brain functional studies.

To validate the utility of 1310 nm μODT for quantitative monitoring of depth-dependent brain dysfunction, we performed CBFv imaging studies on the sensorimotor cortices of C57BL/6 mice after repeated cocaine exposure ($30\mathrm{mg}/\mathrm{kg}/\mathrm{each}$, intraperitoneally). Different from the procedures described above, two $3\mathrm{mm}\times 3\mathrm{mm}$ cranial windows were created on the sensorimotor cortices of both hemispheres to correlate abnormal behaviors with local cerebral hemodynamic dysfunction of the mice.

Figure 3 compares of the CBFv images of sensorimotor cortices on both left and right hemispheres of a mouse undergoing transient ischemic attacks as a result of repeated cocaine exposures. Interestingly, the upper panels show that CBFv (mostly large pial flows) in the upper left cortex (a) was much higher than that of the upper right cortex (b); however, lower panels were the opposite, i.e., the microvascular CBFv in the deeper left cortex (c) was dramatically lower than that of the deeper right cortex (d).

 

Fig. 3. In vivo CBFv images of left and right sensorimotor cortices of a mouse after repeated cocaine exposures ($30\mathrm{mg}/\mathrm{kg}$, i.p.) acquired by 1310 nm μODT. (a), (b) upper cortices; (c), (d) deeper cortices; (e) statistic comparison of capillary CBFv between left and right cortices.

Download Full Size | PDF

This was confirmed by the statistical analysis in Fig. 3(e) that the mean CBFv in the upper right cortex was 64.4% slower than the upper left cortex; the mean CBFv in the deeper left cortex was 58.9% slower than the deeper right cortex. The unbalanced microcirculatory CBFv between the left and the right sensorimotor cortices was likely associated with the observed transient cerebral palsy (paralyzing left paws). Although more study needs to be done, our previous work [7] on cocaine elicited microischemia tends to suggest that local blood supply deficit (e.g., decreased capillary CBFv) in the mouse sensorimotor cortex elicited by repeated cocaine might likely contribute to cerebral palsy.

In summary, we present 1310 nm μODT to enhance quantitative 3D imaging of cerebral blood flow networks. In vivo mouse brain validation studies demonstrate the utility of 1310 nm μODT for effectively increasing the depth for quantitative 3D imaging (not angiography) of the capillary CBFv networks from $\sim 300\mathrm{\mu m}$ of prior 800 nm μODT to $\sim 1.4\mathrm{mm}$. Such technological capability is crucial for functional brain studies because it enables us to quantitatively characterize the CBFv networks across the entire cortex (I-VI layers) and their dynamic, yet highly heterogeneous changes in response to brain activation, as well as the associated downstream behavior patterns. For instance, the new technique allows us to identify the correlation of repeated cocaine elicited transient cerebral palsy with the flow redistribution across different cortical layers and the resultant imbalance in microcirculatory CBFv networks between the left and right sensorimotor cortices. Future work will be focused on optimizing flow detection sensitivity and spatiotemporal resolution for various brain functional studies [12].