Fabrication and characterization of large numerical aperture, high-resolution optical fiber bundles based on high-contrast pairs of soft glasses for fluorescence imaging

: Fabrication and characterization of flexible optical fiber bundles (FBs) with in-house synthesized high-index and low-index thermally matched glasses are presented. The FBs composed of around 15000 single-core fibers with pixel sizes between 1.1 and 10 μ m are fabricated using the stack-and-draw technique from sets of thermally matched zirconium-silicate ZR3, borosilicate SK222, sodium-silicate K209, and F2 glasses. With high refractive index contrast pair of glasses ZR3/SK222 and K209/F2, FBs with numerical apertures (NAs) of 0.53 and 0.59 are obtained, respectively. Among the studied glass materials, ZR3, SK222, and K209 are in-house synthesized, while F2 is commercially acquired. Seven different FBs with varying pixel sizes and bundle diameters are characterized. Brightfield imaging of a micro-ruler and a Convallaria majalis sample and fluorescence imaging of a dye-stained paper tissue and a cirrhotic mice liver tissue are demonstrated using these FBs, demonstrating their good potential for microendoscopic imaging. Brightfield and fluorescence imaging performance of the studied FBs are compared. For both sets of glass compositions, good imaging performance is observed for FBs, with core diameter and core-to-core distance values larger than 1.6 μ m and 2.3 μ m, respectively. FBs fabricated with K209/F2 glass pairs revealed better performance in fluorescence imaging due to their higher NA of 0.59. the good potential of the using for

thermally matched soft glasses with refractive index difference of 0.37 were used for FB development [47].
In this work, we report the fabrication and characterization of flexible, lensless and highresolution imaging FBs with a variety of pixel size and field of view using novel in-house synthesized zirconium, sodium and borosilicate soft glasses. Stack-and-draw technique was used to fabricate FBs with approximately 15000 pixels ordered in a hexagonal lattice based on in-house synthesized glasses. Use of in-house synthesized soft glasses allows us to obtain relatively high refractive index contrast between the fiber core and cladding compared to commercially available FBs made with silica glasses [42,48,49]. Our FBs had NA values of 0.53 or 0.59. These high NA FBs bring together increased fluorescence collection efficiencies and reduced mode overlaps between neighboring fiber cores, i.e. reduced optical crosstalk. These imply better imaging contrast especially for fluorescence applications. We performed numerical analysis which verify the effects of the NA and FB geometry to fluorescence collection efficiency and mode overlap. Increasing the NA of the FBs also allows for reducing the core-to-core distances without a significant reduction in image contrast, resulting in less pixelation artifacts [50]. Critical FB parameters such as core size, pitch between the individual fibers, and NA of fibers were verified so as to optimize the imaging quality. The brightfield and fluorescence imaging properties of these bundles were characterized. First, wide-field imaging experiments on both transmission and fluorescence of a micro-ruler sample, Convallaria majalis sample and dye stained paper tissue sample were carried out and the pixel size effects on resolution were compared. Afterwards, ex-vivo fluorescence imaging of a cirrhotic mice liver tissue that is cleared using the CLARITY technique [51] was performed, and the resulting image contrasts were compared. We also performed fluorescence imaging experiments with FBs in presence of several bending radii as small as 32 mm and observed good contrast for all cases. Our results show significant improvements on FB development with novel material systems for non-invasive, fast and high-resolution fluorescence imaging. Such flexible, high NA FBs can enhance the contrast of the collected fluorescence images thanks to their relatively large collection efficiencies and small mode overlaps.

Development of imaging fiber bundles
Optical FBs studied in this work are fabricated using two different pairs of glass materials [47][48][49]. The first set includes in-house synthesized high-index zirconium-silicate glass labelled ZR3 ( . Each pair of glasses are thermally matched which ensures that they can be drawn together on the optical tower. The basis optical and thermo-physical parameters of the glasses are presented on Table 1.
To develop the FB elements, the modified standard technique for fabricating the stepindex fibers is used [52]. The technique consists of several steps shown in Fig. 1. First, a circular rod, approximately 10 mm in diameter, made from high-index glass and a tube with an internal diameter of 10 mm and an outer diameter of about 20 mm made from low-index glass are prepared. Next, after inserting the rod into the tube, both glasses are drawn on the optical fiber drawing tower and scaled down to obtain a rod with a diameter of about 400 µm. Then, approximately 15000 rods produced in this way are inserted into a thin-walled capillary made of low-index glass to form final preform. The rods form a hexagonal structure where each rod performs the role of a single 'pixel' in the optical bundle. Finally, the FB preform is drawn on the optical fiber drawing tower and scaled down to obtain the final optical bundle. This approach enables obtaining FBs with varying diameters and pitch between each 'pixel' from the same preform. Furthermore, it ensures that the individual 'pixels' in a particular FB have precisely the same optical properties.

Characterization of fiber bundles
FBs were prepared and characterized in two different groups consisting of three and four FB samples respectively, all having approximately the same number of individual fibers (~15000). The lead-free Zr3/SK222 glasses are used as core and cladding materials in the fabrication of the first group with pixel sizes of 1.1 μm, 1.6 μm, and 10 μm. In the second group of FBs, in-house synthesized K209 and commercially available lead-silicate F2 glass are used as core and cladding materials. Second group consist of FBs with pixel sizes of 1.9 μm, 2.4 μm, 2.5 μm, and 2.9 μm. Scanning electron microscope (SEM) images of the cross sections of these bundles are presented in Fig. 2, and main specifications of fabricated FBs are given in Table 2. The missing pixels in some of the imaging bundles presented in Fig. 2, are the result of defects in the hexagonal lattice assembly that occurred at the stage of fabrication of the image bundles. However, the number of defects is very small and does not exceed 0.5% for all FBs. Therefore, their influence on image quality is very limited.  First group of FBs (FB 1-3) has core refractive index (n f ) and cladding refractive index (n c ) of 1.609 and 1.520, respectively measured for sodium D-line. For the second group of FBs (FB 4-7) n f and n c are measured to be 1.619 and 1.508, respectively. As we discuss in the following, due to higher refractive index contrast (n f /n c ), the optical cross talk between adjacent fiber cores is lower for the second group, which results in better overall imaging performance.
For each FB, the total diameter which corresponds to the imaged field-of-view, and the measured numerical aperture (NA) values are also given in Table 2. NA of an individual fiber inside a FB describes its light-gathering ability. For NA measurements, both ends of FBs were first cleaved with a 20 mm length. After that, the laser beam (488 nm wavelength) was focused at the proximal end of a FB using a microscope objective (Nikon L Plan SLWD, 50x, NA = 0.45), and transmitted laser beam widths at known distances from the FB's other end were measured by a CMOS camera (Thorlabs, DCC1545M). The use of a high magnification microscope objective enables coupling the laser beam to only a few fibers in a FB, which is required for precise NA measurement of individual fibers. As an example, full width half maximum (FWHM) of the beam as a function of distance from the bundle tip measured relative to the first measurement point is given in Fig. 3 for the FB 7. From the linear fit, the NA of the FB 7 is calculated as 0.59 which is in good agreement with the theoretical predictions. The NA measurements were performed for all FBs in the same way, and the measured NA values are given in Table 2. Measured values are consistent for each group of FBs and they are in general agreement with the theoretical predictions given by 2 2 f c NA n n = − that reveals 0.53 and 0.59 for FBs 1-3 and 4-7, respectively. Slight variations between the measured and calculated NA values are attributed to uncertainties caused by coupling of the laser beam to multiple fibers during the measurements. For each FB, the minimum bending radius was also characterized. Approximate minimum bending radii were determined by measuring the radius of curvatures of the FBs after the FB was angled 90° without excessive bending. These measured values are also given in Table 2.  A numerical analysis was also carried out to study how the efficiency of light collection depends on the NA of the presented fibers in comparison with fibers with a smaller NA [53]. The collection efficiency, η, for a single core, that represents a single pixel in the FB located in a medium with refractive index n 0 and collecting light from planar fluorescent source with area A s , is given as [54,55]: where d denotes the core diameter, NA is the numerical aperture, and A s is a function of the distance z between the fluorescent plane and the fiber surface: Figure 4(a) presents η values calculated for FBs with various d and NA values. This relationship shows that, for a constant core diameter, the collection efficiency increases with increasing NA. Therefore, in order to achieve the highest possible resolution and at the same time the highest possible collection efficiency in fluorescence imaging, it is necessary to increase the NA. , where E 1 and E 2 represent the complex electric field functions of the guided modes in the first and second core, respectively. We have used the finite element method (COMSOL Multiphysics) to calculate the overlap integral for FBs with fixed core diameter of 2 µm and different NA values as a function of the distance between the centers of the cores. We show that, for every particular distance between the cores, the mode overlap significantly decreases with an increase in NA (Fig. 4(b)). Hence, we conclude that FBs with high NA values provide low optical crosstalk even for relatively small distances between the neighbouring cores. High NA is obtained when refractive index difference between the core and cladding is increased. Simultaneously it allows for better confinement of the mode in the core area and reduction of the distance between neighbouring cores in the FB, resulting in less pixelation artifacts.

Brightfield and fluorescent imaging with fiber bundles
The experimental setup used for brightfield and fluorescence imaging with FBs is presented in Fig. 5. A direct contact between the imaging bundle and imaged sample was employed in all of our experiments except for those in which the Convallaria majalis sample was imaged. Since the Convallaria majalis sample was located beneath a cover glass, the proximal end of the FB could not be placed immediately in contact with that sample. Only for this sample, an imaging doublet pair was employed between the proximal end of the FB and the sample for collecting the brightfield images. In the experimental setup, there are two different illumination modules for brightfield transmission and epifluorescence imaging systems. For transmission, a standard broadband light source was used to illuminate the sample that was placed between the light source and the distal end of the FB. Hence, transmitted light is transferred to the FB directly. For fluorescence imaging, an Ar + ion laser with a central wavelength of 488 nm was used to excite the samples. The incident laser light passed through a spherical plan-convex lens (f = 100 mm) and reflected from a dichroic mirror to a focusing objective with 50x magnification (Nikon L Plan SLWD). Here, the convex lens was used to focus the laser beam into the back-focal plane of the objective and, thus, to establish for widefield illumination of the fiber bundle. The proximal end of the FB was placed in the frontfocal plane of the microscope objective. Hence, excitation light was transferred to the sample through the FB itself. For detection, both brightfield and fluorescence imaging systems used the same imaging module. FB collected the transmitted or emitted fluorescence light and transferred it to the microscope objective. After the microscope objective, the light passed through the dichroic mirror and two longpass filters (Thorlabs, FEL0500 and FGL495) which blocked the excitation wavelength. Afterwards, an achromatic doublet lens (f = 50 mm) that served as the tube lens of the infinity-corrected imaging system transferred the transmitted or emitted light to a CMOS camera (Hayear, HY-3307). A micrometer translation stage was used to control the position of the sample relative to the fiber bundle. Brightfield microscopy experiments were performed with two samples: A micro-ruler sample and a Convallaria majalis (Lily of the Valley) root tissue [57] sample. While, fluorescence imaging experiments employed a lens cleaning paper tissue sample stained with fluorescein dye and cirrhotic liver tissues stained with fluorescein isothiocyanate (FITC). We note that photobleaching did not hamper our fluorescence imaging experiments due to the use of high fluorophore concentrations together with low laser intensities and low laser exposure times.
The fluorescence images shown in this manuscript were recorded from different positions on the samples and employed different laser intensities and camera parameters optimizing the image contrast. Hence, intensities in the reported fluorescence images do not represent a direct measure for the fluorescence collection efficiencies of individual FBs. We also recorded fluorescence images of similar locations in the dye-stained paper tissue sample with FBs 2 and 4 using the same laser excitation power and camera parameters. In agreement with the theoretical predictions, these images showed larger collected fluorescence intensities with FB 4 in contrast to FB 2. We refrained ourselves from making a quantitative comparison between these results and the theoretical predictions shown in Fig. 4(a), mainly because of sample deformations between measurements upon contact with the FB, and FB coupling efficiency variations caused by fiber cleaving.   , 1.6 μm, and 10 μm. The quality of the brightfield image acquired by FB 1 is lower in comparison with the images acquired using FBs 2 and 3. Figures 6(g)-(i) present fluorescence images of the dye stained paper tissue sample acquired by FBs 1-3, respectively. Fluorescence emitted from the fibers of the paper tissue is apparent in all images. For FBs 1 and 2 with small pixel size, the poor quality of the fluorescence images is attributed to optical crosstalk between individual fibers in a FB caused by a relatively low refractive index contrast between fiber core and fiber cladding (i.e. low NA).
The contrast of a brightfield or fluorescence image was quantified by σ/µ where σ is the standard deviation of pixel intensity and µ is the mean pixel intensity of a selected region of interest in a given image. While σ increases with the contrast of an image, division by µ compensates for the changes in the overall brightness of different images. σ/µ values calculated for all brightfield and fluorescence images in this paper are shown in Table 3. In order to investigate imaging quality in the presence of bending, fluorescence images of stained paper tissue were recorded using FB 2 for several bending radii (infinity, 190 mm, 120 mm and 32 mm). The resulting σ/µ values were calculated in the range of 0.25 -0.39 which revealed a good contrast for all cases.
When the transmission and fluorescence images shown in Fig. 6 are compared, we observe that image resolution and contrast are comparatively lower for the case of fluorescence imaging with FBs 1 and 2. In case of transmission imaging, created image is only based on scattering and absorption of the illumination light. Therefore, the properties of an image are defined by illumination conditions. In case of fluorescence imaging, image is created as re-emission from fluorescent dyes and properties of the image are defined by the Lambertian emission. As a result, for low NA FBs, it is not possible to collect all the emitted light into respective pixels. Some fluorescence emitted with relatively larger divergence angles leaks into neighboring fibers creating a bright background in the image, hence reducing the image quality [41].   In order to demonstrate the micro-endoscopic imaging capability of the studied FBs, exvivo imaging of cirrhotic mice liver stained with FITC was performed. Before staining, the liver tissue was cleared by using the method of CLARITY following the sample preparation protocol described in Ref [51]. This method helps to stabilize the tissue structure by preserving protein and nucleic acid contents of the sample via acrylamide-based monomers. In order to obtain a transparent sample, thermal lipid removal steps were applied by sodium borate buffer (200 mM, pH 8.5) containing 4% (wt/vol) sodium dodecyl sulfate (SDS) for 2-3 weeks. After this clearing process, optically cleared biological sample was labelled with 1:100 α-smooth muscle actin -FITC antibody (Sigma Aldrich, F3777). We note that this clearing process is not necessary for collecting the fluorescence signal from biological tissues in contact imaging mode with FBs. Such a tissue clearing approach will be especially important for noncontact fluorescence imaging of the samples with FBs. FITC has a peak absorption of 488 nm and peak emission of 515 nm thus, an argon ion laser with 488 nm central wavelength was used for exciting the specimen. With the FB tip placed in contact with the specimen at nodular cirrhotic area, cross-section of a vein and hepatocellular region are visible (especially in Figs. 8(b)-(g)). These structures in the nodular region are typical type of structure at the advanced stages of cirrhosis and liver disease.

Conclusions
In this paper, we demonstrated fluorescence and transmission imaging performance results of FBs developed with two different sets of glass compositions: Lead-free zirconium-silicate and borosilicate glasses (ZR3/SK222), sodium-silicate and lead-silicate glasses (K209/F2). The considered pairs of glasses are thermally matched and offer high difference of refractive indices, 0.089 for ZR3/SK222, and 0.111 for K209/F2. As a result, high NA for both types of FBs are obtained: NA = 0.53 and NA = 0.59, respectively. Using stack-and-draw method, we produced FBs with approximately 15000 pixels ordered in a hexagonal lattice and various diameters, core sizes and distances between the pixels. We tested seven different FBs with core sizes varying between 1.1 and 10 μm. To study the effect of the NA and pixel size on their performance, we recorded brightfield images of a micro-ruler and a Convallaria majalis sample, and fluorescence images of a dye stained paper tissue and ex-vivo cirrhotic liver tissue. We also performed numerical studies to verify the effects of NA, core diameter and core-to-core distance to fluorescence collection efficiency and mode overlap between neighboring fiber cores. Results revealed good image resolution and contrast for all FBs except for FB 1 (Core diameter = 1.1 µm, core-to-core distance = 1.6 µm, and NA = 0.53) in cases of both transmission and fluorescence imaging. Minimum core diameter and core-tocore distance values of 1.6 µm and 2.3 μm were sufficient for eliminating the optical crosstalk between individual cores even in the presence of bending for both sets of glass compositions. FBs 2, 4-7 have pixel sizes similar to the ones previously reported by other groups (NA = 0.4; core diameter = 2.2 µm; core-to-core distance<4 µm in Ref [31], NA = 0.27; core diameter = 1.8 µm; core-to-core distance~3.5 µm in Ref [41]), but offer much larger NA. As a result, they can collect higher fluorescent signal and achieve higher contrast values especially in fluorescence imaging. For FBs with smaller pixel size (FB 4) or lower contrast of refractive index (FB 1, 2) images were relatively degenerated due to the cross-talk between individual pixels. In order to ensure high imaging quality with smaller pixel sizes, refractive index contrast (n f /n c ) can be further increased for an optimum core-to-core distance reducing the cross-talk between neighboring pixels. Our results show the good potential of the studied high NA FBs fabricated using in-house synthesized soft glasses for micro-endoscopic fluorescence imaging applications.