2.1.

Extended-Focus Visible/Near-Infrared Optical Coherence Microscopy System

The extended-focus visible/near-infrared optical coherence microscopy (xf-visOCM) system presented here is based on previous xfOCM implementations^16,22,23 and is optimized for high-resolution in vivo cortical imaging. Briefly, light from a supercontinuum source (Koheras SuperK Extreme, NKT Photonics) is first filtered to obtain an illumination spectrum from the visible to the near-infrared wavelength range, passes through a polarizing beam splitter (PBS) and is then coupled into the microscope. As shown in Fig. 1, at the microscope’s core lies a Mach–Zehnder interferometer to permit splitting the Bessel illumination and Gaussian detection modes. The first beam splitter (BS1) of the interferometric configuration divides the incoming light into the reference and the illumination arm, where an axicon (176 deg, Asphericon) transforms the initial Gaussian beam into a Bessel beam. A series of telescopes guides the illumination beam through a pair of galvanometric scanners and through the sample arm where it is finally focused on the sample (i.e., mouse cortex) through a $10\times$ objective ($\mathrm{NA}=0.3$, UPlanFl, Olympus). The backscattered light from the sample is collected by the objective, descanned, and then guided to the Gaussian detection path by means of the second beam splitter of the Mach–Zehnder configuration (BS2), and finally coupled into the detection fiber and sent to the spectrometer. The reference arm comprises a set of prisms of different glasses (BK7, SF6, and UVFS) to balance the dispersion present in the illumination and sample arms and the residual dispersion is precisely balanced through a fine-compensation unit.¹⁶ The interference between the reference and sample light is recorded by a custom spectrometer, comprising a grating ($600\text{lines}/\mathrm{mm}$, Wasatch Photonics), a custom objective ($\mathrm{EFL}=100\mathrm{mm}$), and a CMOS line camera (Basler, spL2048-140km). All 2048 pixels of the camera were used to record the spectral interferograms, spanning from 500 to 850 nm in wavelength. As shown in Fig. 1(a), a pair of masks are placed after the axicon lens and before the detection coupler to filter the spurious beams generated by the tip of the axicon and specular reflections originating from the sample, respectively. Moreover, the polarization state of the illumination was fixed using a PBS and a linear polarizer (P) placed at the entrance of the microscope. In practice, we observed that without this configuration, the visibility of the fringes of the interferogram was low and not constant throughout the spectrum. These effects can be attributed to the spectral and polarization-dependent transmission of certain optical elements, such as non polarising beamsplitters. By controlling the polarization of the illumination, the fringe visibility could ultimately be maximized.

Fig. 1

xf-visOCM system and characterization: (a) schematic of the microscope based on a Mach–Zehnder interferometer comprised of two beam-splitters (BS1 and BS2) to split the Bessel illumination and the Gaussian detection paths. A fine-dispersion compensation unit in the reference arm enables precisely compensating the dispersion of the interferometer and a $10\times$ air objective focuses the light in the cerebral cortex of a mouse. BBDM, broadband dielectric mirrors; PBS, polarized beam splitter; P, linear polarizer; BS, beam splitter. (b) The ultrabroadband visible/near-infrared spectrum of the xf-visOCM system provides an axial resolution in air of $\sim 1.1\mu \mathrm{m}$ over $\sim 200\mu \mathrm{m}$ in depth before deteriorating to $\sim 5\mu \mathrm{m}$ in air. (c) The FWHM of the lateral PSF of the xf-visOCM system is $\sim 1.4\mu \mathrm{m}$ in water and is maintained over $150\mu \mathrm{m}$ in depth through the extended-focus configuration, as shown by both the heatmaps and their corresponding profiles (in linear scales). Scalebar: $5\mu \mathrm{m}$.

2.2.

Mouse Preparation

All the experiments were carried out in accordance to the Swiss legislation on animal experimentation (LPA and OPAn). The protocols (VD 3048) were approved by the cantonal veterinary authority of the canton de Vaud, Switzerland (SCAV, Département de la sécurité et de l’environnement, Service de la consommation et des affaires vétérinaires) based on the recommendations issued by the regional ethical committee (i.e., the state committee for animal experiments of canton de Vaud) and are in-line with the 3Rs and follow the ARRIVE guidelines. The capabilities of the xf-visOCM system were assessed by imaging the superficial cortex of mice in vivo. To this intent, the cortex of $n=3$ c57bl/6 mice (3 to 4 months old, Charles River) was made optically accessible through an open-skull craniotomy. Mice were first anesthetized using a solution of ketamine (80 to $100\mathrm{mg}/\mathrm{kg}$) and xylazine ($14\mathrm{mg}/\mathrm{kg}$) injected intraperitoneally. Once asleep, their skull was exposed through an incision in the skin of their heads. A circular region of skull of 4 mm in diameter above the somatosensory cortex was drilled, excised, and then sealed using a glass coverslip and dental cement. Finally, a fixation ring was placed on the sealed cranial window to minimize motion artifacts and facilitate the localization of the same cortical region over time. After the surgery, mice were allowed to recover for a period of a week. Prior to imaging, mice were anesthetized using the same ketamine/xylazine mixture as mentioned above and once asleep, were placed on a custom head fixation system for imaging. The mice’s eyes were kept moist using eye ointment, and their body temperature was maintained at 36°C throughout the surgical procedure and the imaging sessions. Postoperative care consisted in daily subcutaneous injections of dexamethasone ($0.2\mathrm{mg}/\mathrm{kg}$), carprofen ($5\mathrm{mg}/\mathrm{kg}$), and buprenorphine ($0.1\mathrm{mg}/\mathrm{kg}$) for 72 h from the day of the surgery.

3.1.

xf-visOCM Resolution Characterization

The system’s axial resolution was assessed using a mirror placed on the sample holder and by changing the length of the reference arm. As shown in Fig. 1(b), an axial resolution of $\sim 1.1\mu \mathrm{m}$ in air ($\sim 0.85\mu \mathrm{m}$ in water) is preserved over the first $\sim 200\mu \mathrm{m}$ in air ($\sim 150\mu \mathrm{m}$ in water) before significantly increasing and reaching $>5\mu \mathrm{m}$ by $\sim 400\mu \mathrm{m}$ in air. The axial resolution was measured as the full-width at half-maximum (FWHM) of the axial point spread function (PSF) in linear scale. The lateral PSF of the system was interrogated by imaging a sample of 30 nm gold nanoparticles embedded in PDMS. From the acquired data, an average PSF for seven different depths was obtained by extracting and averaging over five individual PSFs at each plane. The resulting PSFs over depth are displayed in Fig. 1(c), where the FWHM of the central lobe is maintained at $\sim 1.4\mu \mathrm{m}$ over $150\mu \mathrm{m}$ in depth. At larger depths, the PSF slightly broadens due to the presence of additional sidelobes and by a decrease in the signal intensity (observable here through an increase in the background noise). This loss in intensity is caused by a combination of the shape of the axial profile of the Bessel beam and by the Gaussian apodization²⁴ and by the system’s roll-off.

3.2.

In Vivo Imaging of Cortical Structures

With the system characterized, the imaging performance of the xf-visOCM platform was assessed by imaging the superficial cortex of mice. All images presented in this paper were acquired at a line rate of 20 kHz, with an incident power on the sample of 0.5 mW and were averaged over five volumetric acquisitions, i.e., C-scans. The imaging time for a single volumetric acquisition was from 2 to 4 min. Figures 2(a) and 2(b) show a B-scan and en face views at different depths of the superficial cortex, respectively. Similarly to previous observations of cortical structures using OCT,^{3,6,15–17,20,25,26} dura mater (highlighted by green arrows) and myelinated axons (pointed by yellow arrows) are characterized by a stronger backscattering than the neuropil and appear as bright layers and fibers, respectively. The extended-focus capabilities of the system are illustrated in the en face views presented in tile (b) showing the ability to resolve the fine bright myelinated axons throughout a depth of $80\mu \mathrm{m}$. Myelinated axons can also be visualized at a depth of $30\mu \mathrm{m}$ from the cortical surface in another animal, as shown in Fig. 2(c). In addition, through their increased backscattering, these structures could be segmented from the parenchyma using a simple thresholding operation, similarly to Merkle et al.¹⁵ By adapting the threshold for attenuation in depth, the axons were segmented at different depths to obtain a depth-encoded maximum intensity projection (MIP) as shown in Fig. 2(d).

Fig. 2

xf-visOCM in vivo imaging: the imaging capabilities of the xf-visOCM system were assessed by imaging the superficial cortex of mice. (a) A B-scan across the first $\sim 100\mu \mathrm{m}$ in depth of the cortex highlights the different structures observable, such as bright, myelinated axons (pointed by yellow arrows), and dura mater (pointed by green arrows). (b) The en face views at different depths present the same structures as in (a) and show the preservation of the lateral resolution in depth. (c) In another animal, myelinated axons can also be visualized in the en face view at a depth of $30\mu \mathrm{m}$ from the cortical surface. (d) A depth-encoded view of segmented myelinated axons exposes the dense distribution of these fibers within the first layer of the brain. Scalebars: $100\mu \mathrm{m}$.

In addition to imaging neural fibers, the high resolution of the system permits discriminating fine features such as capillaries and the boundary of large vessels from the surrounding background. A fine dark region can be discerned within large vessels, below their membranes [blue arrow in Fig. 3(a)], as previously reported by Merkle et al.¹⁵ and identified as a cell-free layer within the lumen of large caliber vessels. Similarly to our previous work in ex-vivo slices at higher lateral resolution,¹⁶ the lumen of capillaries can be distinguished from the neuropil as dark elongated structures, as shown by the red arrows in Fig. 3(a) and by the red arrows in the minimum intensity projections (MinIP) shown in Figs. 3(b) and 3(c) and 4(c). In contrast to slices where the reduced intensity was linked to the absence of scatterers in the lumen of the vessels, here the reduced backscattering seems to be caused by the dynamic nature of the scattering of red blood cells (RBC). The images displayed Figs. 3 and 4 were obtained by averaging 8 to 16 B-scans acquired at the same lateral position (in addition to the $5\times$ averaging over C-scans) and represent the low-pass filtered component of the backscattering (i.e., its static component). Therefore, due to the motion and discrete nature of RBCs flowing in capillaries,²⁷ the static contribution of their backscattering is weaker than that of the surrounding parenchyma and appears darker than the tissue. The final pixel size of every en face image shown in Figs. 2Fig. 3–4 after filtering is $512\times 512$.

Fig. 3

xf-visOCM in vivo vascular imaging: owing to the high resolution of the xf-visOCM system, the microvasculature can be resolved both in the backscattering tomograms and in the angiograms. (a) A B-scan covering the first $\sim 100\mu \mathrm{m}$ in depth of the cortex reveals large vessels at the surface and capillaries that can be resolved as dark elongated structures, as pointed by the red arrows. Moreover, a close-up on the large caliber vessel reveals a thin dark cell-free layer below the vessels membrane (blue arrow). Vascular structures can be visualized by either performing a minimum intensity projection on the static backscattering, in (b) and (c), or an MIP on the angiogram, (d) and (e). Similar features are highlighted between the two visualizations by red arrows in (b) and (c) and pink arrows in (d) and (e). Scalebars: $100\mu \mathrm{m}$.

Fig. 4

xf-visOCM in vivo imaging of neural cells: neural cells can be identified in (a) the backscattering B-scan and (b) en face views as dark spheroids surrounded by the neuropil. These cells can be identified up to a depth of $\sim 135\mu \mathrm{m}$. (c) Fine features can be further visualized through minimal intensity projections, highlighting the presence of mainly vascular structures from 35 to $85\mu \mathrm{m}$ in depth, and (d) neural cells from $85\mu \mathrm{m}$ and beyond. (c) Capillaries can be identified in the MinIP by their hollow lumen, as highlighted by the red arrows in tile. (e) The spatial distribution of these cell bodies can be appreciated by a depth color coded minimum intensity projection. Scalebars: $100\mu \mathrm{m}$.

Conversely, an angiogram, as shown in Figs. 3(d) and 3(e), can be obtained by filtering out the static component of the backscattering^28,29 (over 8 and 16 B-scan repetitions, respectively). The high-pass filtering operation was performed through a pointwise complex subtraction of each B-scan.²⁸ The depth-encoded MIPs shown in Figs. 3(d) and 3(e) were obtained by normalizing the angiogram with the structure as described by Srinivasan et al.^2,19 to account for the reduced intensity in depth and alleviate the impact of bright structures in the shadows of large vessels [here mainly bright myelinated axons in the shadows of vessels, visible as bright dots in Fig. 3(e)]. These angiograms highlight the presence of large pial vessels at the surface and the beginning of the cortical microvascular meshwork in deeper regions. Moreover, the location of capillaries in the angiogram coincides with the dark elongated structures reported in the MinIP Figs. 3(b) and 3(c) of the same figure (highlighted by the red and pink arrows on Figs. 3(d) and 3(e), respectively), reinforcing our observation that capillaries can be distinguished as darker structures in the backscattering images.

In contrast to OCT systems operating at longer wavelengths, the penetration depth of visible light is reduced to $\sim 150\mu \mathrm{m}$ in tissue.^9,30 Nevertheless, as shown in Fig. 4, cell bodies located at $\sim 100\mu \mathrm{m}$ in depth can be resolved in the B-scan and in the en face views presented in tiles (a) and (b) (temporally averaged frames). Similarly to other observations of cell bodies in the visible range and using other illumination spectra,^{4,17,19,20,26} these cell bodies appear as spheroids with a lower intensity than the surrounding parenchyma, and their visualization can be further enhanced through a minimum intensity projection, as shown in Figs. 4(d) and 4(e). Moreover, MinIPs at different depths, as shown in Figs. 4(c)–4(e), hint to the laminar organization of the cortex, where the first $\sim 85\mu \mathrm{m}$ are almost devoid of these cell bodies and display a sparse microvascular network, whereas deeper regions show a dense population of cells. Figures 4(d) and 4(e) were obtained by shifting the focus slightly deeper within the cortex, to counteract the attenuation of light. Overall, the shape, contrast, and depth distribution [highlighted in Fig. 4(e)] of these cells coincide with the presence of neural cells in the first part of layer II/III of the rodent cortex.^4,19,26