Video-rate in vivo fluorescence imaging with a line-scanned dual-axis confocal microscope

Abstract. Video-rate optical-sectioning microscopy of living organisms would allow for the investigation of dynamic biological processes and would also reduce motion artifacts, especially for in vivo imaging applications. Previous feasibility studies, with a slow stage-scanned line-scanned dual-axis confocal (LS-DAC) microscope, have demonstrated that LS-DAC microscopy is capable of imaging tissues with subcellular resolution and high contrast at moderate depths of up to several hundred microns. However, the sensitivity and performance of a video-rate LS-DAC imaging system, with low-numerical aperture optics, have yet to be demonstrated. Here, we report on the construction and validation of a video-rate LS-DAC system that possesses sufficient sensitivity to visualize fluorescent contrast agents that are topically applied or systemically delivered in animal and human tissues. We present images of murine oral mucosa that are topically stained with methylene blue, and images of protoporphyrin IX-expressing brain tumor from glioma patients that have been administered 5-aminolevulinic acid prior to surgery. In addition, we demonstrate in vivo fluorescence imaging of red blood cells trafficking within the capillaries of a mouse ear, at frame rates of up to 30 fps. These results can serve as a benchmark for miniature in vivo microscopy devices under development.


Introduction
While microscopy of ex vivo tissues provides valuable information for biological investigations and clinical diagnoses, key insights into dynamic processes can only be obtained under in vivo settings. Video-rate microscopy (defined here as ≥15 fps) has shown utility for visualizing cellular-level dynamics, such as blood cells trafficking within capillary networks. [1][2][3][4] In addition, high-frame-rate microscopy is useful for mitigating motion artifacts due to respiration or mechanical jitter when utilizing an intravital, handheld, or endoscopic microscope. [5][6][7][8][9] In recent years, considerable efforts have been undertaken to develop miniature optical-sectioning microscopes for applications such as in vivo microendoscopy and point-of-care pathology. 5,[10][11][12][13][14][15][16][17][18][19][20][21][22] Regardless of the specific modality-e.g., confocal microscopy, optical coherence tomography, photoacoustic tomography, or multiphoton microscopy-a general goal has been to maximize the imaging depth, resolution, contrast, field of view (FOV), and speed of imaging. In practice, various trade-offs are necessary to optimize a device for a particular application. Among various methods, a number of miniature microscopes have been developed that utilize a point-scanned dual-axis confocal configuration (PS-DAC). [23][24][25] A DAC microscope is a unique confocal architecture that utilizes low-numerical aperture (NA) off-axis illumination and collection beams that intersect precisely at their foci. The DAC architecture decreases the undesired background from multiplying scattered and out-of-focus photons, which in turn improves imaging contrast and depth compared to a conventional single-axis confocal microscope. 26 Point-scanned confocal microscopes achieve optical sectioning through spatial filtering with a pinhole detector, in which an image is obtained in a point-by-point fashion by mechanically scanning the sample or by scanning the focal volume within the sample. However, with frame rates that are typically lower than 10 Hz, the images obtained by a point-scanned confocal microscope are often susceptible to motion artifacts. 27 In addition, such slow frame rates (<10 Hz) are lower than the frequency response of the human visual system (∼15 Hz) when perceiving moving objects. 28 In order to address the drawback of limited imaging speed, modifications were made to a DAC microscope to accommodate line scanning instead of point scanning. Even though confocality is lost in one dimension for a line-scanned dual-axis confocal (LS-DAC) microscope, which limits its ability to image deeper within tissues, previous feasibility studies have shown that a slow stage-scanned LS-DAC microscope (with 2-fps imaging rate and 1-ms line-acquisition rate) can maintain the resolution and contrast of a PS-DAC microscope at moderate depths of up to several hundred microns. 29 In addition, previous Monte-Carlo simulations suggest that a LS-DAC configuration provides superior contrast (signal to background ratio, SBR) in comparison to a line-scanned single-axis confocal microscope. 30 Recently, we have also introduced a variation on the LS-DAC technique called sheet-scanned dual-axis confocal (SS-DAC) microscopy, 31 in which a scientific complementary metal-oxide-semiconductor (sCMOS) camera enables an oblique light-sheet to be imaged at each scanned position. However, both the previous LS-DAC microscope and the recently developed SS-DAC microscope were slow stage-scanned prototypes (frame rate at 2 fps) that did not utilize a fast-scanning galvo (up to 30 Hz), high-speed camera acquisition (10-kHz line acquisition rate), and speed-optimized software. On the other hand, a major challenge for many high-frame-rate fluorescence optical-sectioning devices (with short pixel dwell times) is achieving adequate sensitivity. This is especially true for DAC microscopy, in which low-NA optics are utilized. Here, we demonstrate for the first time that an LS-DAC microscope is capable of achieving video-rate imaging of fluorescently labeled tissues with sufficient sensitivity for a number of potential biological and clinical applications.
The results of this study demonstrate that video-rate LS-DAC microscopy is capable of sensitive in vivo video-rate imaging of blood cells trafficking within the capillaries of a mouse ear at 15 fps (FOV ∼500 × 500 μm) or 30 fps (FOV ∼250 × 500 μm). In addition, the 100-μs camera exposure time (line-acquisition rate) used for video-rate in vivo imaging experiments provides sufficient sensitivity to image excised mouse tongue tissues, topically stained with methylene blue, at subcellular resolution. To further demonstrate the versatility of video-rate LS-DAC microscopy for clinical applications, images are obtained from surgically excised human brain tumor specimens (glioma) expressing 5-aminolevulinic acid (5-ALA)-induced protoporphyrin IX (PpIX) fluorescence.

Video-Rate Line-Scanned Dual-Axis Confocal
Microscope The sCMOS camera used as a detector in this study was able to acquire ∼10;000 raw acquisitions (confocal lines) per second at a 100-μs exposure time for a thin rectangular region of interest of 8 × 2048 pixels. The center 3-pixel width of this strip corresponded to an imaging dimension of ∼2.7 μm in the tissue sample (slightly larger than the width of our diffraction-limited focal line) and was binned to create a digital confocal slit.
For each two-dimensional (2-D) image, the scanning mirror was programmed to translate the focal line in the x-direction (see Fig. 1). Serially acquired confocal lines were stitched together, enabling a LS-DAC video imaging rate of 15 fps over a ∼500 × 500 μm FOV, or a 30 fps video imaging rate over a smaller FOV of ∼250 × 500 μm. A consistent sampling pitch of ∼0.75 μm per pixel (at 100-μs line-acquisition rate) was utilized in the x-direction. According to diffraction theory, the FWHM width of the focal line in tissues is 1.4 μm. 32 Thus, we were sampling the x-direction at approximately the Nyquist frequency.
The sampling density in the y-direction was determined by the pixel pitch in the sCMOS camera, which is 6.5 μm. Taking into account the magnification of the SIL sample holder (n ¼ 1.45), as well as the magnification of the collection-side optics (5×), the sampling pitch within tissue was approximately 0.9 μm per pixel. The diffraction-limited FWHM resolution in the y-direction is 1.2 μm. 32 Therefore, we were slightly undersampling the y-direction per Nyquist.

System Characterization with Reflective Targets
The axial response of the video-rate LS-DAC microscope was measured with a flat reflective mirror in a homogeneous scattering phantom, 5% Intralipid (μ s ∼ 11 mm −1 ), or a nonscattering water sample. To obtain the axial response as a function of defocus (mirror distance from the focal plane), a chrome mirror was translated axially away from the microscope's focal plane (z ¼ 0) using a computer-controlled actuator (Newport Corporation, CMA-12CCCL) at a velocity of 0.1 μm∕s.
Here, the imaging depth is expressed as a nondimensional quantity, "the perpendicular optical length" (L P ), which is the total number of scattering mean free paths on a round-trip path between the tissue surface and the mirror target, L P ¼ 2μ s d, where d is the physical depth of the mirror in the imaging media, and μ s is the scattering coefficient of the media. The axial resolution is defined as the FWHM of the axial-response curve (approximately Gaussian in shape).
The lateral resolution of the video-rate LS-DAC microscope was assessed by imaging a high-resolution reflective 1951 USAF bar target (Edmund Optics, 58-198).

In Vivo Video-Rate Fluorescence Imaging
This study was performed in accordance with an animal use protocol approved by the Institutional Animal Care and Use Committees at Stony Brook University. For in vivo imaging, fluorescein isothiocyanate (FITC)-conjugated dextran (Sigma Aldrich, FD2000S, 2000 kDa, 10 mg∕mL) was injected retro-orbitally into an anesthetized mouse to highlight its vasculature. The mouse was anesthetized with Avertin (Sigma Aldrich, 2,2,2-tribromoethanol, 20 mg∕mL) via intraperitoneal injection at 250 mg∕kg body mass. The animal was placed on a custom platform that allowed for the imaging of the vasculature in its ears. The animal remained under anesthesia during the imaging experiments and was immediately euthanized upon the completion of the experiments.
In order to image blood cells trafficking within the capillaries of mouse at various depths beneath the ear skin surface, a computer-controlled actuator (Newport Corporation, CMA-12CCCL) was used to translate the sample stage along the z-axis in 25 μm∕s steps. Three-dimensional images of mouseear vasculature were constructed by acquiring a stack of horizontal image sections [see Figs. 2(a) and 2(b)]. The laser power for in vivo fluorescence imaging of red blood cells trafficking within the capillaries of a mouse ear was ∼1.8 mW.

Ex Vivo Video-Rate Fluorescence Imaging
To show that the camera exposure times (100 μs) used for these video-rate in vivo imaging experiments could also provide sufficient sensitivity to image both topically applied and systemically delivered fluorescent contrast agents in animal and human tissues, we imaged excised mouse tongue tissues topically stained with methylene blue, as well as PpIX-expressing brain tumors from glioma patients that had been administered 5-ALA prior to surgery. The laser power for ex vivo imaging of both mouse tongue and PpIX-expressing human brain tumor was ∼1.3 mW.
Excised mouse tongue tissues were topically stained with 0.5 mL of 1% methylene blue (MB) for 10 min and then rinsed thoroughly with 1× PBS. Images were obtained from the top surface of the tongue.
Thick (unsectioned) human brain tumor specimens (lowgrade glioma) were obtained during surgery and fixed in 4% paraformaldehyde before being imaged. Prior to surgery,  II) for image-guided resection of malignant brain tumor, was orally administered to glioma patients at the Barrow Neurological Institute in Phoenix, Arizona, under an IRB-approved protocol and IND approval for the use of 5-ALA. 5-ALA is intracellularly metabolized by the mitochondria to form PpIX, which has been shown to preferentially accumulate in various tumor cells (including glioma cells). 33 Low-grade gliomas typically do not generate enough PpIX to enable detection using wide-field lowresolution fluorescence imaging, but optical-sectioning confocal microscopy has the resolution and sensitivity to visualize the sparse generation and punctate PpIX expression in these tissues. 34 Here, thick low-grade glioma tissues were imaged with the video-rate LS-DAC microscope to visualize the subcellular PpIX expression (405-nm excitation and 625-nm collection).

Results
Our results indicate that video-rate LS-DAC microscopy is capable of in vivo fluorescence imaging of blood cells trafficking within the capillaries of a mouse ear at a frame rate of 15 fps (FOV ∼500 × 500 μm) or 30 fps (FOV ∼250 × 500 μm). Furthermore, we demonstrate that this device possesses sufficient sensitivity to visualize fluorescent contrast agents topically applied or systemically delivered in animal and human tissues. No significant photobleaching was noted during the in vivo or ex vivo experiments.

Experimental Characterization with Reflective Targets
Axial-response measurements of the microscope, to a flat reflective mirror in a 5% Intralipid scattering phantom (or water), were performed over a range of depths [Figs. 2(a) and 2(b)], where depth is expressed here in terms of a nondimensional perpendicular optical length (L P ) as described in Section 2.2. The contrast (SBR: signal-to-background ratio) of the videorate LS-DAC microscope in scattering media can be approximated as the ratio between the infocus signal (mirror located at z ¼ 0) and the background signal (mirror located 15 to 20 μm away from z ¼ 0  Fig. 2(c). However, this slight curvature is not a significant problem when imaging within thick tissues.  To image an FOV of 500 × 500 μm, the imaging frame rate was decreased to 15 fps [ Fig. 3(b)]. However, the trafficking of red blood cells in capillaries could still be visualized and quantified (typical red blood cells velocity in the mouse ear capillaries is ∼0.1 to 2 mm∕s) 38 [see video 2 and Figs. 3(e) and 3(f)].

Ex Vivo Imaging of Mouse Tongue
Excised mouse tongue stained with methylene blue was fluorescently imaged at 15 fps over an FOV of ∼500 × 500 μm. The ex vivo image shown in Fig. 4(a)-acquired with a 100-μm lineacquisition rate, or 15-fps imaging rate-contains similar detail and contrast compared to the image of the same tissue in which a 1-ms line-acquisition rate (a 10× longer integration time) was utilized along with slow-stage scanning [ Fig. 4(b)]. However, as shown in the example line profiles, there is a slight increase in noise for the high-speed image frame, as is expected due to lower photon counts at a 100-μs line-acquisition rate. The fungiform papillae (red arrow) and the numerous surrounding filiform papillae (blue arrows) are clearly visualized at a depth of 50 μm. Figure 4(c) shows a H&E stained histology section of a mouse tongue exhibiting similar structural details.

Ex Vivo Imaging of Protoporphyrin IX-Expressing Human Brain Tumor
Fluorescence imaging of thick (unsectioned) glioma tissues with LS-DAC microscopy at 15 fps [ Fig. 5(a)] demonstrates adequate sensitivity to visualize subcellular PpIX expression (405-nm excitation and 625-nm collection), with similar structural details in comparison to fluorescence histopathology of PpIX, as shown in red in Fig. 5(b). Previous studies have already indicated the potential benefit of intraoperative confocal microscopy for guiding glioma resections. [39][40][41] However, the frame rates of previous devices have been insufficient to mitigate motion artifacts during handheld intraoperative use. Therefore, a miniature LS-DAC microscope, with a fast frame rate of 15 fps or higher, would be beneficial.

Discussion and Conclusion
In this study, video-rate imaging at 30 fps has been achieved with an LS-DAC microscope with sufficient speed, contrast, and sensitivity to visualize in vivo blood cell dynamics. In addition, high-speed LS-DAC microscopy at 15 fps provides sufficient sensitivity and contrast to visualize cellular details in tissues topically stained with a fluorescent contrast agent, methylene blue, as well as the subcellular expression of fluorescent PpIX in glioma tissues from patients administered with 5-ALA prior to surgery. This is the first demonstration of in vivo video-rate imaging at 15 to 30 fps with a tabletop LS-DAC device, which provides an important benchmark for a miniature handheld LS-DAC microscope under development in our lab. Unlike previous LS-DAC prototypes that utilized slow-stage scanning, we have optimized the hardware, instrumentation, and software to achieve video-rate in vivo imaging. These results are a valuable step toward demonstrating the feasibility of an LS-DAC microscope for clinical applications, especially considering the low-NA collection optics utilized in a DAC microscope, in which the sensitivity for high-speed fluorescence detection has been unclear in the past. The results of this study suggest that LS-DAC microscopy has the potential to provide valuable functional and morphological information for point-of-care cancer diagnostics and intraoperative guidance of tumor-resection procedures.   5 (a) Surgically resected glioma (brain tumor) specimen fluorescently imaged at 15 fps over a FOV of ∼500 × 500 μm. The patient was administered 5-aminolevulinic acid (5-ALA) prior to surgery. (b) A histological section of a similar glioma specimen from a patient administered with 5-ALA prior to surgery. The image shows DAPIstained nuclei (blue), 5-ALA-induced protoporphyrin IX fluorescence (red), and the expression of glial fibrillary acidic protein (green). Scale bar ¼ 100 μm.