2.1.

Specifications for the Achromatic, Ultra-Slim Objective

The design specifications for the achromatic objective are listed in Table 1. The optical requirements are similar to those described in our previous work.¹⁴ However, the current objective has a higher NA and a larger FOV, which were accomplished by relaxing the maximum clear aperture requirement from 0.9 to 1.5 mm. The increased clear aperture means the assembled objective is as wide as a 14-gauge core biopsy needle, rather than the 18-guage objective we previously reported.¹⁵ In addition to providing a smaller diffraction-limited spot size and larger imaging area, the new objective’s higher object space NA and lower magnification cause the image space NA to more closely match the NA0.35 fiber bundle (improving the light collection efficiency).

Table 1

Design specifications of the NA0.55 achromatic, ultra-slim objective.

The object space NA0.55 means the objective will have a point spread function (PSF) size of 1.14 μm at the 515-nm emission peak of proflavine, a fluorescent dye which stains cell nuclei, allowing visualization of the size and the shape of each nucleus. In image space, the PSF size is 2.28 μm, which is still well beneath the 3.8-μm core-to-core spacing of the individual fibers in the fiber bundle. This sampling frequency, combined with the $-2\times$ magnification, means the lateral resolution of the endomicroscope will be approximately 3.8 μm. A 360-μm FOV is matched onto the 720-μm active diameter of the image guide by the $-2\times$ magnification. The short 15-μm working distance allows the objective to image tissue in contact with the first lens. Achromatizing brings the excitation and the emission planes of proflavine to the same focal plane, which can improve the collection efficiency.⁵ Also, a wavelength range from 452 to 623 nm allows the achromatic objective to work with proflavine and cresyl violet, simultaneously.¹⁴ The objective is telecentric in object space, and nearly so in image space. Landau et al.¹⁶ reported that maintaining telecentricity in image space for such narrow objectives would compromise transverse magnification and performance at the edge of the FOV. Varying the principle ray angle from 0 deg, though it leads to some vignetting, helps to maintain higher image quality than if the design was doubly telecentric.

An objective meeting the requirements in Table 1 was designed in ZEMAX (Radiant Zemax, LLC., Redmond, Washington). The layout of the design is displayed in Fig. 1. The objective is composed of six plastic lenses. The detailed analysis for selecting the specific plastics has been thoroughly described elsewhere^14,15,17 and includes low autofluorescence for the design wavelengths, high transmission, ease of fabricating aspheres, and the potential for mass production via injection molding. Zeonex E48R and PMMA are both crown-like plastics, while polystyrene behaves optically like a flint. Zeonex E48R also has very low water absorption, which is why it was selected to be in contact with tissue. Lenses 2 and 3 form an achromatic doublet of PMMA and polystyrene with the optical adhesive NOA61 acting as the cement, and lenses 4 and 5 do the same.

Fig. 1

Custom achromatic objective design.

2.2.

Nominal and Expected Performance of the Achromatic, Ultra-Slim Objective

The nominal performance of the objective is shown in Fig. 2. The modulation transfer function (MTF) in Fig. 2(a) shows that the objective maintains diffraction-limited performance across the FOV. It accomplishes this in part by assuming a curved object; this is permissible because the sag in object space is less than the thickness of a typical breast cell. The chromatic focal shift plotted in Fig. 2(b) confirms that the objective is achromatic. Also, the 6.0-μm diffraction-limited range is larger than the 4.0-μm maximum chromatic focal shift.

Fig. 2

(a) Nominal MTF of the objective. (b) Nominal chromatic focal shift of the objective.

The design was thoroughly toleranced in ZEMAX. The tolerances listed in Table 2 were estimated based on in-house fabrication capabilities of an ultra-precision lathe. The lenses were formed via single-point diamond turning (SPDT). For injection molding, the tolerances are considered commercial or state-of-the-art.¹⁸

Table 2

Tolerance values for the NA0.55 needle achromatic objective.

2.3.

Specifications for the SI-HRME

Figure 3 shows a schematic for the endomicroscope. The excitation light is generated by a blue LED (M455L2, Thorlabs, Newton, New Jersey) at 455 nm and passes through a $10\text{-}\mathrm{lp}/\mathrm{mm}$ grid (59–545, Edmund Optics Inc., Barrington, New Jersey) on a motorized stage (T-LS28E, Zaber Technologies Inc., Vancouver, British Colombia, Canada). An image of the grid is focused onto the proximal face of a fiber bundle (FTIFHR800S, F & T Fibers and Technology GmbH, Berlin, Germany) with an NA0.3 commercial objective (RMS10X-PF, Thorlabs). The grid’s image is relayed through the fiber bundle and focused onto the fluorescent sample by the integrated objective. The modulated emission light returns through the system and is imaged onto a charge-coupled device camera (GRAS-14S5M, Point Grey Research, Inc., Richmond, British Colombia, Canada), where it is reconstructed into an optically sectioned image. When the grid is outside the path of the excitation light, the system is functionally identical to a conventional HRME. The SI-HRME can take advantage of this by using the motorized stage to translate the grid out of the light path; this allows conventional HRME images to be compared with SI-HRME images with identical FOVs without the need for a separate imaging system.

Fig. 3

Structured illumination high-resolution microendoscope (SI-HRME) design layout.

3.1.

Measuring the Tilt and Decenter of Ultra-Slim Doublet Lenses

Before assembling a complete objective, five sample doublet lenses were made to validate the custom alignment features’ abilities to create systems which meet the given tolerances. Each doublet was fabricated as described above, using flat optical surfaces to facilitate tilt and decenter measurements. To measure tilt, each doublet was placed on a microscope slide beneath a NewView 5020 white light interferometer (Zygo Corp., Middlefield, Connecticut) after the tilt of the stage was reduced to $<0.01\mathrm{deg}$ [Fig. 5(a)]. The tilt of each doublet was measured five times and averaged. The average tilt of the five doublets was 0.17 deg with a standard deviation of 0.07 deg.

Fig. 5

(a) Sample doublet undergoing tilt measurements with a Zygo white light interferometer. (b) Sample doublet undergoing decenter measurements with a Zeiss microscope. (c) Set of lenses ready for assembly, fully assembled achromatic objective, and a commercial objective, next to a ruler for scale.

To measure the decenter, each doublet was placed on a microscope slide beneath an Imager.Z1 upright microscope (Carl Zeiss, Oberkochen, Germany) [Fig. 5(b)]. Images of the first and the last surfaces were captured without adjusting the lateral position of the stage or the doublet itself. This process was repeated four more times. The decenter between each doublet was measured by using ImageJ to find the center of each surface and to calculate the displacement between pairs of surfaces. The average decenter for the set of doublets was 9.6 μm with a standard deviation of 9.0 μm. When compared with the toleranced values for tilt and decenter, 0.2 deg and 15 μm (Table 2), the measured values indicate that some, though likely not all, of the objectives built will have diffraction-limited performance. These measurements will also allow further optical designs to be more accurately toleranced.

3.2.

Validation Setup of the SI-HRME with an Integrated Achromatic Ultra-Slim Objective

A complete set of individual lenses and an assembled achromatic objective are shown in Fig. 5(c). A layer of heat shrink tubing was added to block outside light from entering the system and to improve the objective’s durability. To reduce stray light through the system, baffles were added within the objective by applying permanent marker up to the edges of the optical surfaces [the blue surfaces in Fig. 5(c)]. An integrated achromatic objective is shown in Fig. 6(a). The fully assembled SI-HRME is shown in Fig. 6(b) and matches the schematic in Fig. 3.

Fig. 6

(a) Achromatic objective integrated to a fiber bundle. (b) Assembled SI-HRME. Col collection lens; Ex excitation filter; DM dichroic mirror; Com commercial objective, FB fiber bundle; AO achromatic objective; Em emission filter; Con condenser lens; FM folding mirror.

4.1.

Achromatic, Ultra-Slim Objective Performance

The performance of the achromatic objective was characterized before incorporating it into the endomicroscope. Images from the custom objective were relayed through an Imager.Z1 upright microscope as shown in Fig. 7(a). A fluorescent 1951 USAF resolution target (38–256, Edmund Optics Inc.) was placed at the object plane of the custom objective; while in water immersion, an image of the target was relayed with a $10\times$ NA0.45 objective (Carl Zeiss). An example image showing good performance is displayed in Fig. 7(b). There is some loss of resolution at the edge of the FOV; this is because the objective is designed to image a curved object, and the target is flat. Adjusting the focal plane by a few microns brings the edge features into sharp focus.

Fig. 7

(a) Commercial Zeiss microscope objective relaying images from the custom achromatic objective. (b) Fluorescence image of a 1951 USAF resolution target imaged through the achromatic objective.

To ensure that the custom objective is achromatized, the chromatic focal shift was measured. A VariSpec liquid crystal tunable filter (Cambridge Research & Instrumentation, Inc., Hopkinton, Massachusetts) was used to provide narrowband illumination starting at 450 nm. A resolution target was brought into focus, and the objective’s position was measured. This procedure was performed in 20-nm steps up to 630 nm, and then repeated four times. The average chromatic focal shift is plotted below along with the nominal chromatic focal shift (Fig. 8). The measurements demonstrate that the objective is achromatic, though the chromatic focal shift range of 6.9 μm is slightly larger than the 6.0-μm diffraction-limited shift range. This analysis was compared with a different custom, ultra-slim NA0.55 plastic objective that conforms to the specifications in Table 1, except that it is designed for only a single wavelength; it is optimized for 515 nm, the emission peak of proflavine, and performs well at this wavelength. The nominal chromatic focal shift for the monochromatic objective is 96.7 μm, more than an order of magnitude greater than the diffraction-limited shift range. The excitation and the emission peaks of proflavine are marked on the measured achromatic objective plot ($X$) and on the nominal monochromatic objective plot (+). Note that the peaks are within the diffraction-limited range for the achromatic objective and are outside this range for the monochromatic objective.

Fig. 8

The diffraction limited depth of field for a $2\times$, NA0.55 objective represented by the gray rectangle is 6.0 μm. The nominal shift range plotted by the solid line is 4.0 μm. The measured shift range plotted by the small dashed line is 6.9 μm. The nominal shift range of an NA0.55 monochromatic objective plotted by the large dashed line is 96.7 μm.

Achromatizing the objective keeps the contrast of the structured illumination grid pattern near the diffraction limit. The pattern, which has a frequency of $19.4\mathrm{lp}/\mathrm{mm}$ in the image plane of the custom objective, has expected contrasts of approximately 0.98 at 452 nm and 0.97 at 515 nm, which leads to a high-intensity, optically sectioned image. The monochromatic objective’s expected contrast at 452 nm is also 0.98, but the contrast at 515 nm has dropped to approximately 0.70, indicating a weaker-structured illumination signal.

4.2.

Performance of the Integrated SI-HRME

The lateral resolution limit of the endomicroscope was measured before and after integrating the achromatic objective (Fig. 9). With only a bare fiber bundle, the SI-HRME is limited by the fiber spacing, and the smallest resolvable features on the USAF 1951 resolution target are group 7, element 1 [Fig. 9(a)] indicating a resolution limit of 7.8 μm. With the achromatic objective at the distal tip, the effective fiber spacing at the object is doubled due to the objective’s magnification. All features on the target are resolvable [Fig. 9(b)] indicating a resolution of at least 4.4 μm (this nearly matches the expected 3.9-μm resolution limit).

Fig. 9

Fluorescence images of a 1951 USAF resolution target through the HRME (a) before integrating the achromatic objective to the distal tip of the fiber bundle and (b) after integrating the objective. The lateral resolution limit improves from 7.8 μm (group 7, element 1) to 4.4 μm (group 7, element 6).

4.3.

Imaging Results with the Integrated SI-HRME

The endomicroscope’s ability to image breast tissue was evaluated using freshly resected murine mammary tissue. A non-neoplastic murine mammary gland and a murine mammary gland bearing an invasive carcinoma lesion were sliced to a thickness of approximately 300 μm using a tissue slicer (Krumdieck MD4000, Alabama Research and Development, Munford, Alabama), stained with a 0.05% proflavine solution for 5 min before washing with $1\times$ PBS (Sigma-Aldrich Co., St. Louis, Missouri), and then imaged with an excitation of 488 nm using the Vivascope 2500 confocal microscope (Caliber Imaging and Diagnostics Inc., Rochester, New York). Once imaging with the confocal microscope was complete, regions of interest from the ex vivo sample were imaged with the integrated SI-HRME both with and without optical sectioning via structured illumination.

Selected sites were matched, and the results are shown in Fig. 10. Each row shows images which were obtained from the same tissue slice, and attempts to match the FOV. The same region of interest is guaranteed for the conventional HRME and the SI-HRME, because the fiber bundle is held in place at a region of interest while the grid is translated into or out of the path of the excitation light. However, minor morphological changes in the tissue lead to some uncertainty when matching the FOV for images from the confocal microscope. Even when the FOV is not precisely matched, the selected features provide a viable comparison between the various modalities.

Fig. 10

Ex vivo mouse tissue images through the endomicroscope and a benchtop confocal microscope.

Cells are visible in the images of the neoplastic mammary tissue obtained with all three systems. However, in the denser, more highly scattering tumor specimens, image contrast is degraded in the conventional HRME images, and individual cells cannot be resolved. The SI-HRME improves contrast, and individual tumor cells can clearly be resolved providing images with similar quality to those images obtained with benchtop confocal microscopy. Note that there is greater contrast improvement for the tumor images, which are dense with nuclei, than for the relatively sparse normal tissue images.