1. Introduction

This article presents the development of a high numerical (NA) aperture diffraction limited microendoscope objective that is part of a fiber confocal reflectance microscope (FCRM) system [1]. Previous miniature objectives have been designed for this system to work with a probe that has a 10 mm outside diameter (O.D.), which includes an image bundle (Sumitomo IGN-15/30) that has a 2.5 mm O.D. and two hydraulic suction lines that measure 0.3 mm O.D. The intended use of the original system is for in vivo imaging of oral and cervical precancers where probes of this size are appropriate. However, to extend the ability of the FCRM system to image the interior of hollow organs such as the colon or urinary bladder, the maximum probe O.D. must be equal to or less than 4 mm, which would make it compatible with current endoscopy probes. The objective presented in this paper meets this criteria with an outer diameter of 3.85 mm and length of 14.65 mm (object plane to end of fiber ferrule) without sacrificing optical performance of the original design (see table 1).

In addition to miniaturization, cost is also an important factor in the objective design. Disposable components are often desired in the medical community because it eliminates the need for repeated sterilization procedures. Since the only part of the fiber confocal microscope that comes in contact with the tissue is the objective, there is strong motivation to make this component disposable. Typical microendoscope objectives are not disposable since they require precise miniature optics that are expensive and difficult to align. A disposable high-NA miniature objective has been developed in our group by using only plastic injection molded optics and optomechanical features [2,3]. This miniaturized objective had an O.D. of 8 mm, which was acceptable for imaging oral and cervical regions; however, it exceeds the size required for endoscopic imaging (< 4mm). To meet this new size limitation and disposability requirement, a hybrid approach was proposed that combines both plastic injection-molded optics and glass optics. The final design utilizes inexpensive components with one off-the-shelf glass element, two plastic injection-molded elements, and LIGA fabricated opto-mechanics, which are all easily mass producible. We believe this is the first microendoscope objective for fiber confocal reflectance microscopy that offers high performance at a low cost.

2. Fiber confocal reflectance microscope system overview

The FCRM is a reflectance based microscope that is currently used to detect precancerous lesions by confocal imaging in the oral cavity and in the cervix. The FCRM is a point scanning, epi-illumination system in which monochromatic light is focused to a point and then raster scanned across a coherent fiber bundle composed of 30,000 individual fibers. At the distal end of the bundle, each fiber is imaged to a specific location on the tissue by a miniature objective. In the presence of an index mismatch between the nucleus and the cytoplasm of an abnormal cell, light is reflected back through the system. This index mismatch is roughly Δn = 0.05 which translates to 0.034% of incident flux reflected, but can be increase to Δn = 0.07 with acetic acid solution [4]. A beamsplitter redirects the reflected signal toward an avalanche photodiode (APD) which collects the reflected light converting it to an electrical signal for image acquisition and processing. To obtain images in 3D, hydraulic suction lines are used to translate different tissue layers through the focus region as it is being raster scanned in the lateral direction.

The low reflected signal from the tissue requires a design which optimizes the signal-to-noise ratio (SNR). A large contribution to noise comes from reflections off the fiber bundle surfaces since they are conjugate to the object plane. To reduce these reflections, both surfaces are immersed in an index matching oil. The objective elements must be AR coated and designed to minimize the irradiance from back reflections at the fiber bundle surface. In addition to reducing the background noise, the system is designed with a large NA objective to maximize collection of the reflected signal. A number of groups are also developing contrast agents to boost the reflected signal from abnormal cells [5–9]. A more detailed description of the FCRM can be found in previous papers [1,4].

3. Optical Design

3.1 Design requirements

The design requirements for the new microendoscope objective are similar to previous designs [2,10] as it is intended to work with the same FCRM system described in the section above. The main difference in design specification is size constraint of the objective. For endoscopy applications, the O.D. of the probe cannot exceed 4 mm. Therefore, the optics, opto-mechanics, and hydraulics, which are necessary for imaging at multiple depths, must now be integrated into one small package. In addition, the design wavelength has been changed from 1064 nm to 850 nm for use with metallic nanoparticle contrast agents. Table 1 lists all the design specifications.

Table 1. Design Specifications for Third Generation Objective.

View Table | View all tables in this article

3.2 Design process

For this objective, it is imperative that the optics be miniaturized as much as possible to allow room for the opto-mechanical and hydraulic features that must reside within the reduced 4mm O.D. To determine if this is a realistic goal, the theoretical limit for the optics clear aperture was examined for a doubly telecentric system. Equations (1) and (2) are used to predict the minimum diameter for both object and image space [11].

Where WD is the working distance, m is the objective magnification, NA is the numerical aperture, and z’ is the image distance measured from the last optical surface. Inserting the design requirements from table 1 into equations (1) and (2) yields a diameter of 1.28 mm. The clear aperture does not need to remain at this theoretical limit, since that would over constrain the design preventing the necessary freedom for aberration correction but this limit does indicate that a solution is possible within a 4 mm diameter. The largest minimum diameter of 1.28 mm is from the object side where the rays are expanding at the fastest rate due to the high numerical aperture. For miniaturization, the highest power elements should be placed closest to these rapidly expanding rays to constrict them early in the design, resulting in the smallest possible system clear aperture diameter. Attempts to miniaturize systems using only plastic optics have been unsuccessful due to low refractive indices, i.e., n_plastic~1.5 of plastics. Glass optics have much higher refractive indices of up to n_glass~ 2.0, but are expensive for high precision miniature elements. In an attempt to keep the cost of the objective low while maintaining a small OD, it was decided that an off-the-shelf high index glass element would be used. Typically off-the-shelf optics cannot be used for miniature high NA objectives since the latter require tight tolerances. However, by using an innovative lens mounting feature, as discussed later in the paper, the loose lens tolerances can be mitigated to an acceptable level. Since most of the power resides in the glass element, plastic injection-molded aspheric elements can be used primarily for aberration correction.

3.3 Final design form

The final design is a three element objective with a clear aperture diameter of 2.75mm. The design meets all of the criteria described in Table 1. The first element in the objective is a standard 2 mm diameter plano convex lens (Edmund Optics P/N: R45-955). The last two elements are custom plastic aspheric lenses. The lenses are made out of ZEONEX 480R plastic. Manufacturing issues were addressed in the design with consultation from a local plastic injection-molding company, Aurora Optical Inc., who also produced the initial prototypes of the plastic lenses. Fig. 1 illustrates a two dimensional layout of the lens system and Table 2 lists the lens prescription.

 

Fig. 1. Miniature Microscope Objective Layout

Download Full Size | PDF

Table 2. Miniature Microscope Objective Lens Prescription

View Table | View all tables in this article

The optical performance of the design is shown in Fig. 2. The design is diffraction limited over the entire field of view for λ = 850 nm as shown in the spot diagrams in Fig. 2(a) and the MTF plots in Fig. 2(c). Fig. 2(b) displays the percent of distortion versus the field position (y-axis). The maximum distortion is approximately 0.25% or 1.5 microns at the 0.707 field position. This error is acceptable since it is smaller than the 7 micron fiber-to-fiber spacing within the fiber bundle and individual fiber displacement error of +/- 2 microns [10].

 

Fig. 2. Theoretical performance of the microendoscope objective. (a) Geometric spot diagrams for three radial image positions: on-axis, 0.707 field, and full field. (b) Distortion plot in percentage of deviation from chief ray location. (c) Modulation transfer function plot for the same image field positions in (a).

Download Full Size | PDF

3.4 Tolerance analysis

A tolerance analysis was performed to predict fabrication success for the final design. Tolerances for the glass element are specified in the Edmund Optics catalog and displayed in table. Tolerances for the plastic optics are specified by the plastic injection-molder, and are also listed in Table 3.

Table 3. Tolerances for miniature optics in microendoscope objective.

View Table | View all tables in this article

Decentration of the first lens (glass) is eliminated with the opto-mechanics (see section 3.6), and is not specified in Table 3. Surface tilt for a plano-convex spherical element can also be considered as decentration, and likewise, is also not specified. Element decentration is assumed to be ±3 microns.

The tolerance analysis used a Monte Carlo (MC) simulation that statistically adjusts design parameters outlined in Table 3 using a normal distribution, and then evaluates the performance of the system based on the RMS spot size. Compensation parameters are the object position, image position, and object shape. Object shape is allowed to vary due to the object sectioning capability of the system. The acceptance criteria is that fifty percent (50%) of the Monte Carlo simulations must have a RMS spot size equal to or less than the diffraction limited Airy disk spot size. Table 4 lists the proportion of Monte Carlo simulations which result in an RMS spot size smaller than a threshold value.

Table 4. Monte Carlo tolerance results for objective.

View Table | View all tables in this article

50% of the MC simulations resulted in an RMS spot size less than or equal to 3.86 microns which is close enough to the Airy disk spot size of 3.46 microns that it is acceptable, especially when comparing these sizes to the diameter of the individual fibers (~4 microns) which couple the light at the objectives image plane back through the system.

3.5 Ghost reflection analysis

As mentioned previously, the reflected signal from abnormal cells is ~ 0.034% of the incident flux. The reflected signal is so small it is imperative that back reflections in the objective be reduced to minimum levels. Optical surfaces are coated with an anti-reflection (AR) coating to decrease reflections from 4-5% to 0.25% of the incident flux, except for the surface which contacts the index matching fluid where AR coating is not needed. Even with AR coatings, the back reflections are still an order of magnitude larger than the reflected signal from tissue. If a back reflection comes to focus near the image plane, its irradiance level is comparable to the object signal, increasing the need for increased dynamic range of the detector. Spot sizes at image plane are analyzed using ZEMAX Ghost Focus Generator [12]. Irradiance values are computed assuming uniform flux density across the field of view. In cases where reflected rays are vignetted, throughput is estimated by ratioing solid angles. Fig. 3 shows the reflected light rays from some of the surfaces. Table 5 lists the irradiance contributions from each surface.

 

Fig. 3. Schematic of back reflections from some of the surfaces in objective. (a) surface 3, (b) surface 2, (c) surface 1, (d) tissue.

Download Full Size | PDF

Table 5. Back reflection irradiance comparison to signal irradiance.

View Table | View all tables in this article

The irradiance level from the tissue to the fiber bundle location is 4 orders of magnitude (~ 10⁴) larger than that from the other surfaces. From this analysis it is evident that back reflections will have a minimal impact on the system performance.

3.6 Optomechanical design

The opto-mechanical design needs to have high precision, small feature size, low cost, and ease of assembly. Traditionally machined, molded, or stamped parts are limited in feature size and precision but offer high throughput and low cost depending on complexity of design. Other processes, such as electro-discharge machining (EDM), produce parts with high precision and small feature size, but are expensive and low volume. An alternative to these approaches are LIGA fabricated components [13]. LIGA is a German acronym, for X-ray lithography (X-ray Lithographie), Electroplating (Galvanoformung), and Molding (Abformung)). The process offers high precision (sub micron dimensional control), repeatable processing based on lithography, and low cost due to high throughput from batch or parallel fabrication. Feature sizes range from millimeter to sub-millimeter dimensions [13]. The use of LIGA fabricated parts also allows for easy implementation of complex miniaturized geometries without additional component cost. These attributes make LIGA fabricated components ideal for this application. The process does have limitations in that the surface shapes must be vertical in nature, and typical surface layers do not exceed 500 microns in thickness. However, with proper design these limitations can be overcome.

The final opto-mechanical design for the miniature microscope objective is shown in Fig. 4. The elements in the drawing are all to scale. Ray tracing through the objective illustrates the object and image locations and the optical element clear apertures. The housing is composed of a series of layers made of nickel. Most layers are 500 microns in thickness and have dimensional tolerances of +/- 2 microns. The disadvantage of using this stacking approach is that the individual layer tolerances in the axial direction increase by the number of layers between the elements. Fortunately, the optical system is less sensitive in this direction and can tolerate this error. The lateral tolerances are the critical parameters which are still held. As the layers are stacked, the possibility that the location of each layer can shift giving a “Tower of Pisa” effect is possible. However, with this design the tightest tolerances are associated with the first two elements which are near the distal tip of the objective mitigating this concern. The advantage of using this stacking approach is that the design becomes very modular. Layers can be easily added or removed without having to redesign the optomechanics. The layers are intended to be diffusion bonded/soldered to one another providing a rigid watertight objective.

The tissue (object) is located in the bottom right corner of Fig. 4 and is immersed in saline solution. The saline solution is delivered by the hydraulic channels located in an isolated ring between the optics and outer diameter of the objective. The ring is split into two crescent shaped channels by two through pins used for alignment during assembly (see Fig. 5(a)). The channels converge at the object chamber. One channel is used for suctioning and the other is used for filling and removing bubbles from the chamber. Each hydraulic channel connects to a standard silicone or PTFE tube at the end of the objective. The hydraulics are used to move tissue through the objective focal plane for 3D imaging. In addition to providing 3D sectioning, the hydraulic channels can also be used for delivery of contrast agents or for marking areas of tissue with dye for later analysis.

 

Fig. 4. Opto-mechanical assembly schematic of microendoscope objective.

Download Full Size | PDF

The distal tip of the fiber bundle is attached to a ferrule that fits into the objective at the rear. The objective includes a latch to lock the fiber bundle in optimum position for imaging. The latch is a bi-stable spring, which locks in both an outward and inner position. Fig. 7 shows a sequence of pictures where the fiber bundle is inserted into the objective and then locked in place.

The plastic injection-molded optics is held in proper position by a mechanical flange molded into the lens that centers on a rim located in the housing. A retaining ring is used to secure the optic in place from the back side. The glass element is positioned in the objective using a self centering ring. The ring actively engages the lens surface flexing away as it makes contact, while centering the lens with respect to its optical axis as shown in Fig. 5. This retaining ring eliminates any decentration associated with off-the-shelf lens manufacturing error.

 

Fig. 5. LIGA manufactured self-centering ring for glass element in objective. (a) Front view of self-centering ring. (b) Side view of self-centering ring showing the flexing direction of ring.

Download Full Size | PDF

4. Assembly/Testing

4.1 Assembly

For prototyping purposes, the plastic optics were diamond turned instead of injection molded. Diamond turning a soft material like Zeonex 480R, is difficult, and leaves tool marks across the surface. White light profilometry measured an rms roughness of ~ ½ wave and an average peak-to-valley of ~ 3 waves. For initial performance testing this surface is acceptable, however, the performance will be slightly degraded. The lenses were also not anti-reflection (AR) coated. The objective is assembled using a vertical stacking approach in which two precision through pins hold alignment as the layers are slid into place. The optics are added to the objective as it is being assembled. The LIGA components are secured in place using UV curable epoxy. The epoxy is applied to the outside wall of the objective after the layers are pressed together, and does not add to the thickness of the objective. In the future, layers of the objectives will be diffusion/solder bonded together making it water tight for use with hydraulics. The fiber bundle ferrule is assembled in the same manner as the objective. Fig. 6(a) shows some individual components used in the objective assembly. Fig. 6(b) is a picture of the complete microendoscope objective. Both pictures are taken with a U.S. penny to show its relative size.

 

Fig. 6. Pictures of actual microendoscope objective and components. (a) Some components used in the objective (lenses, objective layers, and ferrule). (b) Assembled microendoscoe objective.

Download Full Size | PDF

Since the objective is intended to be a disposable component of the FCRM system, it is important that its integration with the system is easy and simple. The next sequence of pictures demonstrate the seamless mating of the objective with the fiber bundle ferrule. Once the bundle is fully inserted into the objective, the operator pushes down on the latch handle, ‘snapping’ the latch spring into its inner locking position. The two pieces are securely held together until the operator lifts the latch handle up, at which time, the latch spring ‘snaps’ back to its unlocked position. The fiber bundle ferrule can now slide freely from the objective for disposal. The objective is designed with a ferrule stop that prevents the fiber bundle from damaging the objective lens or itself during insertion and removal.

 

Fig. 7. Sequence of pictures showing the integration of the fiber bundle with objective.

Download Full Size | PDF

4.2 Testing

The objective was tested on a modified optical bench test setup described in detail in M. Chidley’s paper [2]. The source was changed from a 1064 nm Nd:YAG laser (Amoco 1064-50P) to an 850 nm IR LED (LEDtronics L200CWIR851). The spinning diffusers in the test setup were also eliminated, since the LED source is inherently incoherent. The objective is tested in trans-illumination. The test targets are placed in an immersion oil bath at the objective’s image location and uniformly illuminated from behind. The objective images the target to the water immersed object location. A water immersion IR objective (Olympus UM575) with matching NA (NA=1.0 WI) and a tube lens (f=120mm doublet) relays the image onto a CCD camera (Pulnix TM-745E) for analysis. Two methods were used for evaluating the optical performance of the objective. A simple first test involves imaging a standard 1951 United States Air Force (USAF) Resolution target. The target contains groups of vertical and horizontal bars with known spatial frequencies. The highest resolved spatial frequency roughly indicates the resolving capability of the system. The test is only meant to provide a quick evaluation of performance since it is susceptible to human error. Fig. 8(a) shows the objective resolving the smallest elements on the resolution chart, the Group 7 element 6 bars, which correspond to a spatial frequency of 228 line pairs per millimeter (LPMM). This spatial frequency corresponds to a resolution of 760 LPMM at the tissue or 1.3 microns (object plane). The FCRM’s lateral resolution is limited by the fiber-to-fiber spacing of the fiber-optic bundle and not the objective’s lateral resolution. The fiber-to-fiber spacing is ~ 7 microns (71 LPMM) in image space, or ~ 2.1 microns (238 LPMM) in object space. The axial resolution of the system is estimated to be ~ 2 microns [14]. The second more quantitative test used to evaluate the optical performance of the objective is the slanted-edge technique [15]. MTF results are extracted from an image of a slanted edge using data processing. Images are taken at two corners, the upper right and lower left, of a square box. Since MTF results are only valid for the axis perpendicular to the edge, both a horizontal and vertical edge were analyzed to better evaluate the objective performance. Fig. 8(b) shows the location of two images taken from the upper right corner.

 

Fig. 8. Optical performance results for microendoscope objective. (a) USAF resolution target imaged by objective. (b) Slanted-edge data acquisition locations for upper right corner.

Download Full Size | PDF

The MTF results (in image space) are shown in Fig. 9. Each curve corresponds to the vertical and horizontal edges of the square. The average Strehl ratio (SR) for the entire system is 0.61 at 850 nm. The Strehl Ratio (SR) compares the peak irradiance in the measured Airy disk to that of the theoretical Airy disk providing a convenient single value measure, ranging from 0 (poor) to 1 (best), of system image quality. Improving the optical surface quality by injection molding the plastic lenses instead of diamond turning should increase this result. The objective was tested in transmission instead of reflection to reduce the effect of back reflections since the optics were not AR coated.

 

Fig. 9. MTF and SR performance for microendoscope object at four edges of a square.

Download Full Size | PDF

5. Conclusion

A miniature microscope objective has been designed, fabricated, and tested, for use in endoscopy applications. The objective integrates both imaging and hydraulic capabilities into a package whose size is only 3.85 mm outside diameter and 14.68 mm in length. The objective has a water immersion NA = 1.0 for maximum light collection at the tissue surface. Testing the prototype showed a Strehl ratio of 0.61 at the design wavelength of 850 nm. The objective is inexpensive ~ $35 and intended for disposable use with a fiber confocal reflectance microscope. Future development is in process to fully implement and test the hydraulic capabilities of the objective and miniaturize the fiber confocal reflectance system for portable use in clinical studies.