Ultrathin visible-light OCT endomicroscopy for in vivo ultrahigh-resolution neuroimaging in deep brain

A neuroimaging technique that offers high spatial resolution and adequate imaging depth is essential for brain mapping, disease evaluation, and intervention guidance in the deep brain. Nevertheless, striking a trade-off between the imaging resolution and depth remains challenging. Magnetic resonance imaging (MRI), a non-invasive technique, is widely employed in clinical settings, particularly in aiding neurosurgical procedures, such as tumor dissection.¹ However, its spatial resolution of millimeters is insufficient to resolve the small and/or early tumors.² While ultrasound imaging can reach several-centimeter-scale depth in the brain, its resolution of hundreds of micrometers is still suboptimal for visualizing fine brain structures.³ As for optical imaging, for example, two-photon/multi-photon microscopy has made significant advancement in deep-brain imaging, offering sub-micrometer resolution and over 1 mm of imaging depth. However, further extending the imaging depth is impeded by substantial optical aberrations and light scattering, which are a consequence of the brain tissue’s heterogeneous refractive index distribution.^4,5 While two-photon/multi-photon endoscopy excels in functional imaging of deep tissue with cellular and subcellular resolution, it faces significant challenges related to signal intensity, field of view, photodamage, complexity, and integration into clinical settings.^6–8 Recently, minimally invasive endoscopes based on multi-mode fibers have been developed for high-resolution deep-brain imaging. However, these tools are limited by a narrow field of view and comparatively slow imaging speed.^9–11 In contrast, endoscopic optical coherence tomography (OCT) offers an appealing solution, providing depth-resolved, high-speed, high-resolution diagnostic images in real time and in vivo.^12–15 

As an emerging technology, visible-light OCT (vis-OCT) endoscopy offers a significant advancement over its 1300 and 800 nm counterparts,^16–20 affording an enhanced resolution of up to 1 μm in tissue, albeit at the expense of diminished imaging depth.^21–24 Furthermore, vis-OCT enables capillary angiography and assessment of vessel oxygenation, providing comprehensive vascular and functional imaging capabilities.^25–27 These features enhance the potential of vis-OCT for detailed structural and physiological assessments in biomedical applications.²⁷ To overcome the limitation of imaging depth in OCT and also reduce the risk of bleeding and trauma, ultracompact OCT endoscopes of submillimeter diameter have been developed to facilitate minimally invasive deep-tissue imaging in solid organs such as the brain.¹⁵ As for the vis-OCT endoscope, current designs that employ achromatic lenses and distal DC motors present challenges due to their bulkiness and potential safety risks associated with electrical leakage. For example, a 15 mm diameter vis-OCT colposcopic probe has been developed for imaging macaque vaginal tracts,²³ and a vis-OCT probe of 6 mm in diameter has been demonstrated for spectroscopic imaging of tissues.²⁴ 

When it comes to fabricating ultrathin vis-OCT endoscopes of high performance, conventional methods often encounter limitations. In particular, the traditional gradient-index lens-based probe design introduces strong chromatic aberration. This aberration is particularly noticeable in OCT endoscopes operating within the visible light spectrum, and consequently, it leads to a significant deterioration of the axial resolution.²⁸ On the other hand, the fiber-melting method has been demonstrated to fabricate OCT probes of small form-factor, such as those with an outer diameter (OD) of 520 μm, and achieve achromatic performance near 800 nm.¹⁴ However, this method has its own limitations, namely, it lacks the flexibility to customize the ball-lens and necessitates a time-consuming angle-polishing procedure. Consequently, these drawbacks limit its applications and reduce the yielding rate. More recently, a two-photon polymerization method has been adopted to print a freeform microlens on the fiber tip, allowing for customization of the imaging performance of a compact OCT probe, with an OD as small as 457 μm.²⁹ However, this method raises concerns regarding its scalability and high production cost. In addition, the layer-by-layer printing mechanism inherent in this approach can introduce undesired optical surface roughness, ranging from 10 to 200 nm.^29,30 This surface irregularity can, in turn, lead to strong unwanted scattering and significant optical loss. Hence, there remains a pressing need for a novel approach to empower the production of aberration-corrected, compact-sized vis-OCT probes for high-resolution imaging.

To address this need, we have previously proposed a liquid shaping technique. This technique greatly facilitates the customization of distal microlens on the wettability-modified substrate and fabrication of achromatic OCT probe of about 0.6 mm in OD, which can work at 800 nm.³¹ It directly fabricates a custom microlens on the fiber tip for side focusing, thereby eliminating the need for angle polishing (see Fig. S1). Moreover, the liquid-shaped microlens affords a sub-nanometer surface roughness, which reduces the spectacular scattering and insertion loss on the lens surface. Our method offers significant fabrication flexibility, enabling precise customization of imaging parameters of OCT endoscopes, such as longitudinal focal shift, focused spot size, and working distance. This enhanced capability holds great potential for fabricating small vis-OCT endomicroscopes.

In this work, we introduce an ultrathin vis-OCT endomicroscope measuring ∼0.4 mm in diameter and providing an ultrahigh resolution of 1.4 μm × 4.5 μm (in axial and transverse directions in air). Fabricated using our proprietary liquid shaping technique, this endomicroscope allows precise control over the microlens size and shape, enabling achromatic and anastigmatic OCT imaging. Its compact size facilitates minimally invasive deep-brain imaging in mice, reaching depths of up to 7.2 mm. OCT images clearly delineate the mouse brain structures, including isocortex, corpus callosum, and caudate putamen. We further examined myeloarchitectures and cytoarchitectures in isocortex and demonstrated that the vis-OCT endomicroscope outperforms the 800 nm version in imaging myelinated axon fibers and nerve fiber bundles. Our vis-OCT endomicroscope overcomes the trade-off between imaging resolution and depth, providing a new instrument for minimally invasive, ultrahigh-resolution neuroimaging in deep brain in vivo.

Figure 1(a) shows a vis-OCT endomicroscope and its distal optics. The distal optics consists of a single-mode fiber, a non-core fiber (NCF), and a tilted semi-ellipsoidal microlens with an incident angle θ of 52.5°. Previous studies have demonstrated that a semi-ellipsoidal microlens can effectively correct the endoscope’s astigmatism [defined as the ratio of spot diameter in the x axis to that in the y axis on the focal plane; see Fig. 1(a)] induced by the negative cylindrical effect of the transparent protective sheath.³¹ The shape and size of lens are defined by the contact angles in the x-z and y-z planes, i.e., CAxz and CAyz, as well as the elliptical half major and minor lengths, i.e., EHM and EHN.

We utilized the ray-tracing method to calculate the imaging parameters of vis-OCT endomicroscope, including longitudinal focal shift (in the spectrum ranging from 493 to 705 nm), focused spot size, and working distance (defined as the distance between the outer surface of glass tube and the focal plane). They can be flexibly tailored by adjusting the design parameters of a semi-ellipsoidal microlens, such as the NCF length and the EHM length [see Figs. 2(a)–2(c)]. By setting a threshold for each imaging parameter, such as a longitudinal focal shift of ≤10 μm, a focused spot size of ≤10 μm, and a working distance between 150 and 250 μm, we could identify a desired region of interest for optimal probe design [see Fig. 2(d)]. For each given EHM length, we can derive its corresponding EHN length to correct astigmatism by optimizing the focal spot’s eccentricity ratio to 1. As a prime example, if we select a probe design of a 550 μm NCF length and a 140 μm EHM length, the EHN length for anastigmatic performance is ∼138.5 μm, corresponding to a contact angle of about 88.5° and 90° in the x-z and y-z planes, respectively. This design affords an achromatic and anastigmatic vis-OCT endomicroscope with a longitudinal focal shift of about 6.4 μm, an astigmatism of 1.02, a focused spot size of ∼4.4 μm, and a working distance of about 185 μm.

We fabricated the vis-OCT endomicroscope with the previously reported liquid shaping technique, which enables the creation of a microlens with a custom size and shape.^31,32 The volume of curable optical liquid (NOA 81, Norland Product Inc.) required for the microlens was ∼6.2 nl after considering a shrinkage ratio of 7.69% in polymerization.³¹ Moreover, by utilizing an elliptical substrate boundary, we could achieve a semi-ellipsoidal microlens with a desired conic coefficient. Microlens fabrication requires several key procedures, including substrate processing, liquid generation, and lens polymerization. The microlens can be polymerized within 10 min under the ultraviolet light exposure. Detailed fabrication details can be seen in the previous work.³¹ The resultant microlens provided an optical surface of sub-nanometer roughness, reducing unwanted scattering in the endomicroscope. Furthermore, a multi-layer anti-reflection coating of 17.75 nm Ta₂O₅, 33.3 nm SiO₂, 142.45 nm Ta₂O₅, and 91.72 nm SiO₂ was applied to the curved surface of the microlens to reduce the back-reflection by ∼1.8 dB. As for the endomicroscope packaging, we used standard fibers, including a single mode fiber (460HP, Thorlabs) and a short piece of non-core fiber (FG125LA, Thorlabs). The fibers underwent standard fiber processing procedures such as cleaving and splicing with a glass processor (GPX3800, Thorlabs) to build a fiber probe. During the coupling procedure, the fiber probe was aligned in the y-z plane of the semi-ellipsoidal microlens [see Fig. 1(a)]. Finally, the imaging probe was encased in a 30-gauge metal enclosure and a transparent glass capillary tube (Molex LLC, with an OD of about 399 µm) to construct a liquid-shaped endomicroscope measuring ∼0.4 mm in diameter and about 215 mm in length [see Fig. 1(b)].

A vis-OCT system was constructed in our laboratory by utilizing a supercontinuum laser (SuperK EXTREME EXU-6 OCT, NKT photonics) to provide broadband visible light with a central wavelength of 620 nm and 3 dB bandwidth of 118 nm, a 50:50 fiber coupler (TW560R5A2, Thorlabs Inc.) to form an interferometer, and a fast spectrometer (Cobra-S visible, Wasatch Photonics Inc.) to detect OCT signals. We used the single-mode fiber 460HP (Thorlabs Inc.) for laser transmission throughout the imaging system. It was found in vis-OCT that even a tiny high-order dispersion mismatch between the sample and reference arms could severely degrade the axial resolution. Therefore, it is crucial to match the materials and their amounts in both arms. To this end, we used a prism pair made of highly flint N-SF11 in the reference arm to match the optics used in the fiber optics rotary joint (MJPP-FAPB-532-46-FA, Princetel Inc.) in the sample arm (see Fig. S2). It should be noted that the microlens on the OCT endomicroscope did not cause strong dispersion (see Fig. S3). We also utilized the polarization controllers (FPC030, Thorlabs Inc.) to mitigate the effect of polarization mismatch in the two arms. Furthermore, a home-developed phase-based digital dispersion compensation method was adopted to further optimize the axial resolution of the vis-OCT system.³³ Briefly, we put a mirror in front of the endomicroscope to measure a raw interferometric signal I(λ). After spectral linearization in the wavenumber space, we utilized Hilbert transform and Fourier transform of the linearized signal I(k) to obtain the complex interferometric signal $Ĩ(k)$⁠, phase delay ϕ, and optical path difference z_OPD, respectively. Since the linearization coefficients k of the spectrometer are pre-calibrated by the manufacturer, we could calculate the dispersion phase ϕ_D = ϕ − 2kz_OPD, apply this phase to compensate the residual system dispersion mismatch, and obtain the dispersion compensated signal $Icomp=Ĩ(k)exp(−iϕD)$⁠. A working flow chart of this digital dispersion compensation method is also provided in the supplementary material (see Fig. S4). For our imaging system, the detection sensitivity was about 93.7 dB at 62.5 kHz A-scans/s, with an average incident power of ∼14 mW on the sample.

We conducted detailed characterizations of the microlens. As shown in Fig. 3(a), the slightly elliptical three-dimensional (3D) form of the microlens was accurately gauged using a confocal optical profilometer (MarSurf CM Expert, Mahr Inc.). In Fig. 3(b), we quantified the ellipticity by measuring and fitting two representative two-dimensional (2D) profiles in the x-z and y-z planes. The results demonstrated fitted EHN and EHM lengths of ∼138.4 and 140.5 μm, respectively. Furthermore, we measured the contact angles in the x-z and y-z planes (CAxz and CAyz) of ∼88.6° and 89.8°, respectively. These measurements closely align with the microlens design parameters. In addition, the surface roughness of the liquid-shaped microlens was ∼0.84 nm, which has also been reported in our previous study.³¹ 

We further characterized the vis-OCT endomicroscope using the home-developed vis-OCT system. As shown in Fig. 3(c), we measured the beam diameters in both the x and y directions. Similar measurements in both directions confirm the anastigmatic performance of the endomicroscope, with a low astigmatism ratio of 1.02. The focused beam spot was measured at a working distance of ∼180 μm, illustrating an average spot size of about 4.5 μm calculated by a weighted function from DataRay.³¹ Furthermore, we measured the axial resolution by placing a mirror in front of the OCT endomicroscope, including the protective sheath. By moving the mirror along the light-emitting direction, we were able to measure the axial resolution at various imaging depths. As shown in Fig. 3(d), the axial resolution along the imaging depth of 800 μm in air demonstrated minimal fluctuation of less than 3.6%. At an imaging depth of around 250 μm, a typical point spread function (PSF) exhibited an ultrahigh axial resolution of 1.42 μm (see the PSFs at other depths in Fig. S5). We further confirm the achromatism of the endomicroscope by the nearly unchanged back-reflected spectra along the light-emitting direction [see Figs. 3(e) and S6 for comparable performance with a ball-lens based achromatic design¹⁴]. The spectra were measured by positioning a mirror at different locations relative to the focal planes of the OCT endomicroscope along the imaging depth. These spectra were captured using an optical spectrum analyzer. To demonstrate the high-resolution imaging capability, the vis-OCT endomicroscope was used to image a tape and a human thumb at a rate of 10 f/s [see Fig. 3(f)]. We can easily appreciate the layered structures of the tape in the OCT image. As for human skin, the OCT image clearly delineates the epidermis, the dermis, and the sweat duct. We can even identify the fine luminal structure of the sweat duct, appearing as pale-white areas within the high-intensity duct wall [see the inset of Fig. 3(f), indicated by red arrows].

In vivo minimally invasive high-resolution neuroimaging in deep brain has been demonstrated on nude mice (n = 4), following a protocol approved by the Animal Experimentation Ethics Committee at the Chinese University of Hong Kong. The experimental procedures, including anesthetization, burr hole opening, endoscope deployment, and histology processing, have been reported in our previous studies.^15,31 For OCT deep-brain imaging, the mouse’s head after scalp incision was stabilized using an ear bar to minimize motion artifacts. The OCT endomicroscope was rotated circumferentially by a fiber optics rotary joint, which was mounted on a computer-controlled translational stage to enable pullback imaging [see Figs. S7(a) and S7(b)]. The OCT imaging was conducted at a speed of 10 f/s, with a probe pullback speed of 100 μm/s.

Figure 4(a) shows a 3D OCT image of a 7.2 mm deep mouse brain, which is reconstructed from a series of 2D circumferential B-scans (720 frames) captured along the longitudinal direction (z axis). This 3D visualization distinctly reveals three brain structures, including the isocortex (IC), the corpus callosum (CC), and the caudate putamen (CP). These structures correlate well with the corresponding histology, as shown in Fig. 4(b). Among them, a major blood vessel (BV) can be identified in the isocortex part, and striatopallidal fibers (SF) are clearly displayed in the caudate putamen.

The en face images, derived from the mean-intensity projection of an unwrapped 3D volume in coronal (y-z), sagittal (x-z), and transverse (x-y) planes [see en face projection procedures in Figs. S7(c) and S7(d)] also allow appreciation of mouse brain structures [see Figs. 4(c)–4(e)]. These images visualize the myelinated axon (MA) fibers and their myeloarchitectural trends in the isocortex layer [see Figs. 4(c)–4(e), red dashed curves]. In the sagittal and transverse images, the neuronal cell bodies stand out as high scattering regions against their background (see the enlarged insets d₁ and e₁). A major blood vessel is identifiable by its shadow of decreased intensity [see Fig. 4(d), black arrow], as also seen in the 3D image. Moreover, the corpus callosum's nerve fiber bundles exhibit strong intensity, with the OCT signal quickly attenuating [see Fig. 4(d), bule box]. The enlarged view reveals the microscopic details of the fiber bundles (see the enlarged inset d₃). Finally, the representative striatopallidal fibers in the caudate putamen region are clearly visualized [refer to Figs. 4(c) and 4(d) and the enlarged inset d₄]. Their size diminishes deeper into the brain regions [see Fig. 4(d), green triangles].

The same region of the mouse brain underwent imaging using both an 800 nm OCT endoscope and a vis-OCT endomicroscope. The 800 nm OCT endoscope, with an outer diameter of ∼0.6 mm delivers an axial resolution of about 2.4 μm and a transverse resolution of about 8.4 μm in air. Figures 5(a) and 5(b) show two representative en face images in the coronal plane, obtained from the vis-OCT and the 800 nm endoscope. Both the images clearly discern the brain regions, including IC, CC, and CP, demonstrating a strong correlation with the histology [see Fig. 5(c)]. Moreover, the vis-OCT image resolves some intricate microstructures more distinctly [see Figs. 5(d) and 5(e)]. For instance, the myelinated axon fibers, hardly visible in the 800 nm OCT image (d₂), are easily identifiable in the vis-OCT image (d₁), as corroborated by the enlarged histology image (d₃). Similarly, the nerve fiber bundles in the vis-OCT image (e₁) are more distinguishable than those in the 800 nm OCT image (e₂), and these microscopic structures align with the histology result (e₃). Furthermore, we include 2D cross sections that highlight the striatopallidal fibers in the caudate putamen for comparison [see Fig. 5(f)]. Upon examination of the 6X enlarged views, the fibers in the vis-OCT image clearly exhibit a more distinct boundary than those in the 800 nm OCT image [refer to Fig. 5(g)]. In conclusion, the evidence presented above substantiates the superior resolution of our vis-OCT endomicroscope compared to the previous 800 nm OCT endoscope.

Current deep-brain neuroimaging techniques grapple with striking a balance between minimal invasiveness, imaging resolution, and imaging depth. To overcome this competing challenge, we introduce the visible-light OCT endomicroscopy approach with an ultrathin form factor of ∼0.4 mm in OD and an ultrahigh resolution of 1.4 × 4.5 μm² in the axial and transverse directions in air. We employed a liquid shaping technique to design and fabricate the vis-OCT endomicroscope. This method not only yields an achromatic and anastigmatic endomicroscope but also maintains its ultracompact size, simply by adjusting the non-core fiber and ellipsoidal microlens. To further optimize the imaging performance, we used an anti-reflection coating to minimize the back-reflection from the microlens’s curved surface to approximately −57.8 dB. In addition, we meticulously designed and constructed a vis-OCT system that effectively mitigated the high-order dispersion mismatch between the sample and reference arms. This is a critical factor in acquiring high-resolution images in the visible light domain.

Compared to conventional fabrication methods of monolithic OCT endoscopes, such as gradient-index fiber-based and fiber melting methods, our approach addresses the yielding problem associated with the time-consuming and labor-intensive fiber end polishing procedure. Our method eliminates the need for this step. We built five visible-light OCT probes using the liquid shaping technique, all of which performed well and exhibited similar characteristics (see Table S1). To improve the success rate of our method, we implemented several measures. First, we constructed the endoscopes in parallel under the inspection of microscopic cameras (see Fig. S1). Second, during the gluing procedure, we used a 30 min UV light exposure to ensure complete polymerization and firm bonding of the lens-fiber coupling area. In addition, we heated the substrate to ∼40 °C, allowing the lens to be safely and easily removed from the substrate due to their different thermal expansion properties. It should be noted that the lens’ working temperature ranges from −150 to 125 °C. Our method does not compromise the robustness of the endomicroscope. For microlens fabrication, the angle precision is about 0.5° and the size precision is about 1 μm, as reported in our previous work.³¹ Therefore, we can precisely control the size and shape of the fabricated microlens [see Figs. 3(a) and 3(b)]. We also conducted a longitudinal study to demonstrate the long-term stability and robust imaging performance of the endomicroscope (see Fig. S8).

In our proof-of-concept demonstration, we achieved deep-brain imaging in mice using our vis-OCT endomicroscope, reaching a depth of up to 7.2 mm. The OCT images clearly distinguished brain structures such as the isocortex, corpus callosum, and caudate putamen, and these findings correlate well with the histology results. Our vis-OCT endomicroscope also revealed substantial fine details of brain microstructures. For instance, it effectively displayed myelinated axon fibers in the en face projections from various orientations. Furthermore, we identified neuronal cell bodies in the isocortex and nerve fiber bundles in the corpus callosum, both of which were presented as local high-intensity areas against the background. In addition, we observed a trend of decreasing size in striatopallidal fibers within a deep region of the caudate putamen. This ultrahigh imaging capability is critical for deep-brain mapping in clinical studies. For instance, MRI scans, with a spatial resolution of about 1 mm, often struggle to delineate the ventral border of the subthalamic nucleus and visualize the internal medullary lamina lamina,³⁴ Clinical ultrasound imaging also faces significant limitations in resolution, fundamentally constrained by the wavelength of the sound waves, typically resulting in a resolution limit of around 100–200 μm.³⁵ In contrast, the vis-OCT endomicroscope provides microanatomical imaging at a resolution of nearly 1 μm in tissue. This capability holds the potential to enable more precise deep brain stimulation (DBS) target visualization, improved lead localization, and ultimately better treatment outcomes. Moreover, the vis-OCT endomicroscope could potentially pave the way for a deeper understanding of the mechanisms of action of DBS in neuromodulation.

The imaging speed for the current animal study is 10 f/s for B-scan or 62 500 axial scans per second for A-scan, and it is compatible with in vivo imaging, as validated by previous studies. For instance, Yuan et al. reported in vivo mouse colon imaging at a speed of 5 f/s (40 980 axial scans per second).¹⁴ Similarly, Sun et al. demonstrated in vivo endoscopic OCT on a mouse tongue using MEMS scanning mirrors at a speed of 2.5 f/s.³⁶ Mavadia-Shulka et al. achieved in vivo murine colon imaging at ∼70 000 axial scans per second using a distal scanning motor,³⁷ a speed comparable with ours. Current imaging speed of our endoscope needs improvement for future clinical use. For example, intravascular OCT typically operates at speeds of 100 f/s or higher.^38,39 It is important to note that the reported imaging speed for deep brain imaging is an exemplar speed. We have also demonstrated higher imaging speeds in a previous work, such as at 20 f/s.¹⁵ The imaging speed in our current work is primarily limited by the spectrometer, which can operate at speeds of up to 250 000 axial scans per second.

Performing in vivo imaging presents several challenges. First, we need to handle the OCT endomicroscope with care due to its fragile distal end, which is only about 0.4 mm in diameter. To ensure successful and consistent in vivo imaging, we typically prepared at least three workable OCT endomicroscopes with comparable performance to mitigate the risk of probe damage. Second, considerable care should be taken during animal handling to minimize the risk of hemorrhage, which can deteriorate OCT imaging quality. After anesthetizing the animal and incising its scalp, we created burr holes ∼1 mm in diameter in the skull [see Fig. S7(b)], ensuring they avoid major blood vessels. Before inserting the OCT endomicroscope, we used a bare fiber, about 250 μm in diameter, to make a small hole through the pia mater. The OCT endomicroscope was then inserted through the burr hole into the mouse brain at a slow speed of ∼10 μm/s.

To achieve good imaging results, several measures need to be taken. First, designing and building a high-quality imaging system is essential. We optimized the dispersion mismatch using both hardware and software methods. Second, designing and fabricating an effective OCT imaging probe is crucial. Our probe design corrected optical aberrations, such as astigmatism and chromatic aberration. The fabricated probe achieved high one-way transmission (>90%) and low back-reflection (<−56 dB), ensuring a high signal-to-noise ratio for the obtained images. In addition, we used a thin probe (∼0.4 mm in diameter) and implemented careful animal handling measures mentioned above to minimize the impact of blood on OCT imaging.

In this study, we used four mice of the same age (4 weeks old) to demonstrate the imaging capability and repeatability of our OCT endomicroscope. By doing so, we also aimed to protect animal welfare by minimizing the number of animals used for this study. All the mice were imaged successfully with good detail, and we provided representative images of the same brain region in four mice (see Fig. S9). Slight image differences may appear due to individual differences among the mice.

Previous research has demonstrated that deep-brain neuroimaging can be accomplished using an OCT galvo-scanning method at a longer wavelength, such as 1.7 μm, to increase imaging depth (1.6 mm in tissue).⁴⁰ However, this approach is considered as a double-edged sword, as the increased imaging depth comes at a cost of compromised imaging resolution (∼10.5 μm in air). In contrast, our method successfully circumvents the depth limitation (imaging depth of 7.2 mm in tissue) by deploying an ultrathin endomicroscope in the deep brain while simultaneously maintaining ultrahigh resolution of ∼1.4 μm (in air). It is important to note that the imaging depth of our method is not inherently limited to 7.2 mm. Instead, it depends on the depth to which the endoscope is deployed within the mouse brain [see Fig. S7(a)] and the size of the brain itself. This effectively underscores the imaging advantages of our vis-OCT deep-brain endoscopy over previous methods.^40–42 Furthermore, our vis-OCT endomicroscope features a compact form factor with an OD of just 0.4 mm, which is significantly smaller than the approximately 1 mm OD of common DBS electrodes.³⁴ The smaller size of the endomicroscope reduces the risk of complications such as intracranial hemorrhage (average rate of 1.90%⁴³) and damage to the brain structures during deep brain surgery.

Currently, our vis-OCT imaging system confronts the challenge of a relatively high insertion loss of the fiber optics rotary joint of about 3.15–5.01 dB. This loss primarily stems from the small mode field diameter of 460HP of about 3.5 μm, resulting in a low and inconsistent coupling efficiency between the optical elements during rotation. The inconsistent loss led to the imaging brightness fluctuation along the adjacent A-scans in the images, and we have implemented a normalization method to compensate such fluctuation (see Note S1). In addition, we observed that the fiber optics rotary joint introduced some spectral loss during rotation, which contributed to the degradation of axial resolution. Variations in polarization states within the system also impacted the axial resolution. In this study, we optimized the polarization controllers to minimize these effects and reduce resolution degradation (see Note S2, Figs. S10 and S11). We will develop new broadband coupling methods or the implementation of an electric-free distal scanning mechanism to replace the high-loss fiber optics rotary joint. In addition, the integration of the endomicroscope with a stereotactic system could potentially enable intraoperatively real-time intracranial navigation and avoidance of major blood vessels in the deep brain, thereby enhancing patient safety during neurosurgery. Vis-OCT offers enhanced contrast for visualizing microvascular structures. In future work, we also plan to optimize the imaging endomicroscope and incorporate spectral contrast OCT angiography techniques to improve blood vessel visualization,²⁶ further leveraging the benefits of vis-OCT.

The supplementary material includes Figs. S1 to S11, Note S1, Note S2, and Table S1.

This work was supported by the Research Grants Council (RGC) of Hong Kong SAR (Grant Nos. ECS24211020, GRF14203821, and GRF14216222), the Innovation and Technology Fund (ITF) of Hong Kong SAR (Grant Nos. ITS/240/21 and ITS/252/23), and the Science, Technology and Innovation Commission (STIC) of Shenzhen Municipality (Grant No. SGDX20220530111005039).

The authors have no conflicts to disclose.

Chao Xu: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Resources (equal); Software (equal); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Tinghua Zhang: Writing – review & editing (supporting). Syeda Aimen Abbasi: Methodology (supporting). Peng Liu: Writing – review & editing (supporting). Bryan P. Yan: Writing – review & editing (equal). Sze Hang Calvin Ng: Writing – review & editing (supporting). Wu Yuan: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Supervision (equal); Writing – review & editing (equal).

All data needed to evaluate the conclusions in this paper are present in this paper and its supplementary material. Additional data related to this paper may be requested from the corresponding author.