Imaging quality assessment of multi-modal miniature microscope

We are developing a multi-modal miniature microscope (4M device) to image morphology and cytochemistry in vivo and provide better delineation of tumors. The 4M device is designed to be a complete microscope on a chip, including optical, micro-mechanical, and electronic components. It has advantages such as compact size and capability for microscopic-scale imaging. This paper presents an optics-only prototype 4M device, the very first imaging system made of sol-gel material. The microoptics used in the 4M device has a diameter of 1.3 mm. Metrology of the imaging quality assessment of the prototype device is presented. We describe causes of imaging performance degradation in order to improve the fabrication process. We built a multi-modal imaging test-bed to measure first-order properties and to assess the imaging quality of the 4M device. The 4M prototype has a field of view of 290 μm in diameter, a magnification of -3.9, a working distance of 250 μm and a depth of field of 29.6±6 μm. We report the modulation transfer function (MTF) of the 4M device as a quantitative metric of imaging quality. Based on the MTF data, we calculated a Strehl ratio of 0.59. In order to investigate the cause of imaging quality degradation, the surface characterization of lenses in 4M devices is measured and reported. We also imaged both polystyrene microspheres similar in size to epithelial cell nuclei and cervical cancer cells. Imaging results indicate that the 4M prototype can resolve cellular detail necessary for detection of precancer. 2003 Optical Society of America OCIS codes: (350.3950) Micro-optics; (110.3000) Image quality assessment; (110.0180) Microscopy References and links 1. M.R. Descour et al. “Toward the Development of Miniaturized imaging systems for Detection of PreCancer,” IEEE J. Quantum. Electron. 38, 122-120 (2002). 2. G. Dallen Bach-Hellueg, H. Doulson, Histopathology of the Cervix Uteri (Springer-Verlag, New York, 1990). 3. J. Lee, J. D. Rogers, C. Liang, R. Richard-Kortum and M. R. Descour, “Stray light analysis for miniature multimodal microscope (4M device),” in Current Developments in Lens Design and Optical Engineering III, R. E. Fischer, ed, Proc SPIE 4767-09, 53-61, (2002). 4. D. S. Goodman, “Basic Optical Instruments,” in Geometrical and Instrumental Optics, D. Malcacara, ed. (Academic Press, San Diego, CA 1988). 5. M.A.A Neil, R. Juskaitis, and T. Wilson, “Method of obtaining optical sectioning by using structured light in a conventional microscope,” Opt. Lett. 22, 1905-1907 (1997). (C) 2003 OSA 16 June 2003 / Vol. 11, No. 12 / OPTICS EXPRESS 1436 #2394 $15.00 US Received April 17, 2003; Revised June 09, 2003 6. J. Lee, J. D. Rogers, and M. R. Descour, “Shack-Hartmann Wavefront Sensor for Micro-optical Imaging Systems,” in Diffractive Optics and Micro-optics R. Magnusson, ed., Technical Digest (Optical Society of America, Tucson, AZ, 2002), post deadline paper. 7. J. D. Rogers, M. R. Descour and A.H.O. Kärkkäinen, “Fabrication and assembly of miniature imaging systems using lithographically patterned micro-optics and silicon microstructures,” in Diffractive Optics and Micro-optics R. Magnusson, ed., Technical Digest (Optical Society of America, Tucson, AZ, 2002), 117-118. 8. Ari H. O. Kärkkäinen, Juha T. Rantala, Michael R. Descour, “Fabrication of micro-optical structures by applying negative tone hybrid glass materials and greyscale lithography,” IEEE Electron. Lett. 3, 23-4 (2002) 9. B. Cassarly, Design of Efficient Illumination systems (Optical Research Associates, Pasadena, CA 2000), Section 4. 10. UP 1800, (Uniq Vision Inc, 2002), http://www.uniqvision.com/ 11. DL7032-001, (Sanyo Corp., 2001), http://www.optima-prec.com/DL7032.htm 12. R.R.Shannon, “Testing of Complete Objectives,” in Applied Optics and Optical Engineering Vol.III, R.Kingslake, ed. (Academic Press, San Diego, CA 1965) 13. J.W.Coltman, “The specification of imaging properties by response to a sine wave input,” J. Opt. Soc. Am. 44, No.6, 468-471 (1954). 14. G.D. Boreman and Sidney Yan, “Modulation Transfer Function measurement using threeand four-bar target,” App. Opt. 34, 8050-8052 (1995). 15. J.D.Gaskill, Linear Systems, Fourier Transforms, and Optics, (John Wiley & Sons, New York 1978). 16. W. J. Smith, Modern Optical Engineering (McGraw-Hill, New York 1990). 17. K. Creath and A. A. Morales, “Contact and Noncontact Profilers,” in Optical Shop Testing, D. Malacara, ed. (John Wiley & Sons, New York, 1992). 18. Wyko NT 2000, (Veeco Instruments, 2002), http://www.veeco.com/ 19. J. C. Wyant and K. Creath, “Basic Wavefront Aberration Theory,” in Applied optics and Optical Engineering Vol. XI, R. R. Shannon and J. C. Wyant, ed. (Academic Press, San Diego, CA 1992). 20. IDL, (Research Systems Inc, 2002), http://www.rsinc.com/ 21. W. H. Press et al, Numerical recipe in C++ (Cambridge University Press, New York, 2002), Chap. 2 22. Plain polystyrene microsphere (PS06N/2110), (Bangs Laboratories Inc., 2002), http://www.bangslabs.com/ 23. F. Friedl et al., “Studies on a new human cell line (SiHa) derived from carcinoma of uterus. I. Its establishment and morphology,” Proc. Soc. Exp. Biol. Med. 135, 543-545 (1970). 24. A.a.M. Brunsting, “Differential light scattering from spherical mammalian cells,” Biophy. Jour. 14, 439453 (1974). 25. J. Maier, et al., “Possible correlation between blood glucose concentration and the reduced scattering coefficient of tissues in the near infrared,” Opt. Lett. 19, 2062-2064 (1994). 26. H. Liu, B. Beauvoit, M. Kimura and B. Chance, “Dependence of tissue optical properties on solute-induced changes in refractive index and osmolarity,” J. Biom. Opt. 1, 200-211 (1996). 27. S. G. Jennings, “Attenuated total reflectance measurements of the complex refractive index of polystyrene latex at CO2 laser wavelengths,” J. Opt. Soc. Am. 71, 923-927 (1981).


Introduction
Recent advances to miniaturize optics are revolutionizing optical imaging.Miniature imaging systems have many advantages including compact size, reduced fabrication cost and simplified mass production; these are of particular importance for medical imaging, where one desires inexpensive, small imaging systems to detect disease without the need to remove tissue.Previously, we have proposed battery-powered, pen-sized multi-modal miniature microscopes designed to specifically image microscopic and molecular features of pre-cancers that can be characterized by morphologic and biochemical changes [1,2].In this paper, we describe the performance of the first such prototype miniature microscope.
Although we introduced the concept of multi-modal miniature microscopes (4M) in earlier publications [1,3], it is briefly described here for clarity.The 4M device is designed to be a complete microscope on a chip, including optical, micro-mechanical, and electronic components which are capable of imaging the morphologic and biochemical features of precancerous cells in vivo.The 4M device is designed to utilize the interaction of light with tissues in many modalities to image morphology and biochemistry in vivo, yielding tools that provide better delineation of tumors.The combination of miniaturization and imaging capability provides broad applicability in many organ sites.Our goal is to first use 4M devices to improve the detection of pre-cancers in organs such as the oral cavity and uterine cervix which can be easily accessed.The optical design and development procedure of the 4M device are reported in Section 2.
In this paper, we present for the first time a prototype version of the 4M device, which consists only of optical elements.Micromechanical and electronic components are not yet integrated into the device.The motivation of this paper is to assess the quality of micro optics and the imaging quality of actual 4M device before completing the integration of a 4M device by adding illumination optics and a digital imaging device.Unlike the optical design of a 4M device and the fabrication of micro-optics, the metrology of micro-optics and the imaging quality assessment of 4M device are still un-explored areas.It is necessary to explore the improvement of imaging quality which can come from the systematic analysis of possible causes of image quality degradation including the metrology of micro-optics.
We built a multi-modal imaging test-bed for 4M devices to test the imaging ability of the optics-only prototype.Within the multi-modal imaging test-bed, optical power is delivered into a 4M device.Test-bed optics relay the image created by the 4M device onto an external CCD camera.The test-bed details are explained in Section 3. Section 4 shows measurement result of first-order properties of the 4M device, including field of view, depth of field, working distance, and magnification.Section 5 shows three assessment results of imaging quality: (1) MTF measurement results for a quantitative estimate of imaging quality (2) Surface profile measurement results of hybrid lenses and (3) Images of a resolution test target (USAF1951), polystyrene beads and cervical cancer cells for a qualitative estimate of imaging quality.Finally we discuss the potential of 4M device as a medical imaging tool and future improvements suggested by the metrology results.

Miniature multi-modal microscope
Since the 4M device will image target organs in vivo, it is designed to be water-immersion, assuming an aqueous film between the tissue and the 4M device.The object to be imaged is located in aqueous medium beneath a silicon substrate.An illumination system and microoptical elements, such as hybrid lenses and an objective lens, create a reflected light image on a CMOS active-pixel image sensor [see Fig. 1].The illumination system in Fig. 1 consists of a condenser lens and multi-mode fiber.It adopts the Koehler illumination concept due to efficiency and simplicity [4].The 4M device also implements optical sectioning by using a scanning sinusoidal grating; the scanning-grating optical sectioning method is described elsewhere [5].Every 4M-device component is mounted on a silicon substrate [a.k.a., Micro Optical Table , MOT] whose size is 13mm (L) × 10 mm (W) [1].
The prescription data of the 4M device are reported in Table 1 and Table 2; the lens design is shown in Fig. 2. The actual 4M device has a folding mirror between an objective lens and the first hybrid lens to reduce total track.In order to reduce the cost and difficulty of fabrication and assembly, the 4M device is designed to have one objective lens from Edmund Industrial Optics; three hybrid lenses, which have the same surface shapes, are vertically mounted on the MOT with equal spacing among them.
The constraints of lens design come from the fabrication constraints of the hybrid lenses.This fabrication process creates only positive lenses and limits maximum sag to 60 µm [1].However, the process also allows for aspherical surfaces.All of hybrid lenses used in the 4M device are plano-convex.Radii of curvature listed in Table 2 are for the convex surface.Hybrid lenses are made of sol-gel material on a glass substrate.The hybrid lenses are only 1.3 mm in diameter.The surface shape is fabricated by a photo-lithographic technique utilizing a grayscale photomask, so that, a lens surface can easily have an aspherical or nonrotationally symmetric profile.The transmission of hybrid glass is reported in [1].
The fabrication procedure of the hybrid lenses and assembly procedure of the 4M device are beyond the scope of this paper and are reported extensively in previous publications [1,7,8].They are briefly summarized here.Conventionally, the fabrication of micro-optical glass structures is a multi-step process that includes UV exposure of a deposited photo-resist layer through a photomask, development and wet or dry etching transfer of the optical structure into the substrate material.The fabrication technique that we developed eliminates the multiple steps and uses a single lithography step: No etching transfer of patterned structures is required when using hybrid sol-gel material.The hybrid sol-gel method provides a variety of options for optimization of material properties such as refractive index and optical loss.
The hybrid glass material functions as a negative tone material.The polymerization of solgel material is controlled by UV dose.This characteristic can be used to fabrication various shapes of structure, for example, a lenslet with grayscale mask.The grayscale mask consists of 127 radial zones that have different optical densities from 0.12 to 1.5.The resulting spatial distribution of UV exposure is rotationally symmetric and leads to decreasing polymerization from the center of mask to the edge of the lens surface.Therefore the center of a surface would be most polymerized and have the thickest cross section.The decreased exposure offcenter makes the sol-gel material less polymerized and the developed surface has less thickness.
The first procedure of fabrication is the liquid phase deposition of sol-gel films by a phased single step spin-coating process.The sol-gel solution is spread on a glass substrate at 300 rpm for 30 seconds and then instantly spun on at 800 rpm for 30s.After spin-coating, the samples are pre-baked at 90 °C.The mercury UV lamp used in the exposure step has irradiance of 10.6 mW/cm 2 at 365 nm.Exposure time is 2 seconds (UV exposure dose of 21.2 mJ/cm 2 ).The exposed sampled is developed in isopropanol and then post-baked at 145 °C for 1 hour.
The hybrid sol-gel material has an index of refraction of 1.531 at 632.8 nm, 1.524 at 830 nm and 1.515 at 1550 nm.Using the Conrady dispersion formula, the index of refraction at 587 nm is estimated to be 1.53 and the Abbe number is estimated to be 45.The material exhibits a maximum extinction coefficient of 2.0 ×10 -4 µm -1 between 450 and 1600 nm.
Before the assembly of 4M device, a hybrid lens is tested by Shack-Hartmann wavefront sensor that is especially built for hybrid lens testing [6].Ten hybrid lenses were manufactured and the best three lenses were selected based on the RMS wavefront error [see Table 3].Two views of the actual "optics-only" 4M device are shown in Fig. 3 that consists only of one objective lens and three hybrid lenses.Fig. 3(b) shows a folding mirror and one objective lens embedded in MOT as well as hybrid lenslets.The folding mirror is added to bend optical path.

Multi-modal imaging test-bed
The 4M device is designed to have an imaging quality comparable to a conventional compound microscope.A 4M device can be designed to enable, for example, reflectance imaging, or fluorescence imaging and optical sectioning.Since the concept and actual development of the 4M device has not yet been reported and its imaging quality is still unknown and subject to improvement, it is essential to develop the methodology for assessment of imaging quality at an early stage of fabrication and assembly.
The "optics-only" 4M device creates images that indicate imaging quality directly but there is not yet an integrated illumination system or an imaging device.The imaging testing system that we have developed consists of a Koehler illumination system and an image relay system to capture images created by the optics-only 4M device without adding any artifacts.The imaging test-bed also implements multiple modes of imaging to evaluate the full potential of 4M devices.
The schematic diagram of multi-modal imaging test-bed is shown in Fig. 4. The red lines in Fig. 4(a) are the chief ray and the blue lines are marginal rays.The adjustable iris diaphragm is an aperture stop of the test-bed.A collimated beam is coupled into two power delivery optical fibers whose diameters are 600 µm.These two fibers implement two illumination modes: fiber "A" for trans-illumination mode and fiber "B" for epi-illumination mode.Both modes adapt the Koehler illumination concept that usually requires objects to be located at an exit pupil or back focal plane of a condenser lens [9].In principle, the irradiance at the object is related to the angular distribution of light source.
In the case of the trans-illumination mode, the object is located underneath the 4M device.The object is illuminated from beneath [see Fig. 4(b)].Images created in this mode are relayed onto a CCD camera [Uniq Vision (UP1800)] by two achromats, L1 and L3 that make up what is called an imaging arm [10].Although we used only trans-illumination for imaging quality assessment in this paper, the multi-modal imaging test-bed reported here will be used to test future 4M devices under multi-illumination modes.
The epi-illumination mode has two "arms."Light from fiber "B" is transferred to an object underneath the 4M device via achromats L2 and L1.This is called the illumination arm.The 4M device captures reflected or fluorescent-emission light from the object and creates an image which is then relayed onto the CCD camera in the same way as in the transillumination mode.Figure 4(a) shows that a scanning sinusoidal grating is inserted into illumination arm.Since the grating is located at an image plane within the illumination arm, the sinusoidal pattern is projected onto the object.This arrangement results in structured illumination that is required for optical sectioning.It will test the optical sectioning power of 4M device.
The beamsplitter that is inserted at the intersection of two arms can be a dichroic beamsplitter for a fluorescing object or an amplitude (e.g., 50/50) beamsplitter for a reflective object.Therefore, the multi-modal imaging test-bed can assess reflectance and fluorescence imaging quality only by exchanging the type of beamsplitter [see Fig. 4 Figure 5 shows the actual multi-modal imaging test-bed.The 3-axis actuators shown within the red oval are used to control the object, the 4M device and the trans-illumination system.The 4M device is located inside the oval.Figure 5(b) provides the magnified view of the 4M device being used with an imaging chamber used to hold cells.The lenses in the testbed are AR-coated and achromatized in the near infrared (NIR) region where operating wavelength (830 nm) of the tested 4M device is located.The light source is a laser diode [Sanyo (DL7032-001)] [11].The magnification of the image relay system is m = -3.This value of the magnification was chosen through consideration of 4M-device field of view, dimensions of multi-modal imaging test-bed and imaging performance of the image relay system.An illumination system was designed so that the etendue was held nearly constant through the 4M device and multi-modal imaging test-bed.For instance, the etendue in object space of the 4M device is 0.0078 (mm 2 ) since the 4M device numerical aperture (NA) is 0.4 and the half field of view is 0.125 mm.The etendue at the end of fiber "B" is 0.0064 because the NA of the condenser lens is 0.15 and the radius of fiber "B" is 0.3 mm.

First order properties of 4M device
Table 4 gives the first-order properties of the 4M device reported in this paper.The field of view and magnification were measured by observing the image of a square grid.The square grid has 50 µm line spacing and 5 µm line thickness.The working distance is defined as the gap between the object plane and the underside of the 4M device.The depth of field is defined as the axial range where the contrast is higher than half of maximum contrast (FWHM) at a single frequency of a square wave.The depth of field was measured at 200 line-pairs/mm.

Field of View
Magnification Working Distance Depth of field 292 µm -3.9 0.257 mm 29.6 µm

Imaging quality of 4M device
We addressed two types of measurements to assess the imaging quality of an imaging system: quantitative assessment such as modulation transfer function (MTF) and the surface characterization of optics in the 4M device and qualitative assessment such as images of specific objects of known dimensions.Here we report results of both approaches.Since the 4M device was designed for detection of pre-cancer, we imaged objects with similar size, shape and optical properties expected in precancerous epithelial tissue: polystyrene spheres and cervical cancer cells.

Quantitative assessment: MTF measurement
The MTF is a commonly used representation of imaging performance in various contexts.It is a more appropriate metric for imaging performance than conventional aberration coefficients because surface shape has some level of non-rotational symmetry that will be shown in next sub-section.Seidel aberration coefficients assume a rotationally symmetric surface.Among the many methods of MTF measurement, we chose the scanning method that is to measure the change of contrast when an object is a sinusoidal grating [12].Although the MTF expresses the change of contrast after imaging a sinusoidal grating as a function of spatial frequency, it is advantageous to use a square wave grating in the actual experiment.The advantage stems from the practical fact that even a very high spatial frequency square wave grating is easily fabricated.Therefore the basic approach here is to measure square-wave response [i.e., contrast transfer function (CTF)] and convert the CTF to ordinary MTF.The equation of such conversion was published previously [13,14].However since our actual implementation is slightly different, we present a modified form of the conversion equation.
We begin with the well-known relationship between cosine wave response, that is, MTF ) (ξ m , and a square wave response, that is, CTF ) (ξ s .ξ is the spatial frequency.
The conversion equation from CTF to MTF is given in J. W. Coltman's paper [13] and is can be +1, 0, or -1 according to the formulas in J. W. Coltman's paper.Equation (2) appears impractical because of an infinite summation but considering the cut-off-frequency property of an optical imaging system, one can limit the summation to only a finite number of terms.Assuming that the cut-off frequency is c ξ , is also zero here because of Eq. ( 1).Therefore, below the cut-off frequency ) (ξ m does not have a contribution from ) (ξ s at frequencies above the cut-off frequency.For instance, if , then only the first term in Eq. ( 2) contributes to ) (ξ m . The same logic can be applied repeatedly until ) (ξ m is determined at every spatial frequency below the cut-off spatial frequency.The result is In the actual measurement, the square-wave gratings used as objects have discrete spatial frequencies separated by equal spatial-frequency increments.The square-wave gratings (Max Levy Autograph, Inc.) range in spatial frequency from 5 LP/mm to 600 LP/mm and the spatial frequency increment is 5 LP/mm from 5 LP/mm up to 115LP/mm and it is 10 LP/mm at higher spatial frequencies.
The field position at which the MTF was measured was chosen to obtain the best contrast and was fixed during entire MTF measurement.Two images were taken at each frequency: one contained maximum irradiance and the other contained minimum irradiance at given field position.The contrast is min max min max The measured contrast data are interpolated in Mathematica to create the CTF ) (ξ s , for the conversion by Eq. (3) to MTF ) (ξ m .Interpolation works by fitting polynomial curves between specified points.The order of fitting polynomial curves is 3, yielding cubic curves.Figure 6 shows plots of contrast data, interpolated CTF and MTF.For comparison, the diffraction limited MTF is included under the assumption of 4M-device NA = 0.15 which is actual NA after reducing aperture size to avoid scattering from the edges of the hybrid lenses.
In a multi-modal imaging test-bed, the adjustable iris diaphragm is an aperture stop.The numerical aperture is calculated by measuring the size of iris diaphragm and magnification.The spatial frequency in Fig. 6 is in object space.According to the MTF curve in Fig. 6, the 4M device shows non-zero contrast even after 200 LP/mm where an object has a similar size to epithelial cell nuclei.
The Strehl ratio can be expressed as the ratio between integrals of the measured MTF and the diffraction limited MTF, according to the central ordinate theorem [15].Based on this definition and the data in Fig. 6, the Strehl ratio is 0.59.The main reason for MTF degradation is analyzed in next sub-section.

Quantitative assessment: Surface measurements of Hybrid Lens
Two major characteristics of optical surfaces are measured: surface quality and surface accuracy.Surface roughness, characterize as surface quality, is reported in Table 5. Radii of curvature and Zernike coefficients plots as surface accuracies are reported in Table 6 and Fig.
7 [16].Convex surfaces of the objective lens and three hybrid lenses mounted on 4M device were measured under a Wyko NT 2000 optical profilometer which is basically a Mirau-type interferometer [17,18].Surface measurements were implemented in two different fields of view (FOV) to investigate the degradation of surface qualities at larger aperture sizes of the hybrid lenses.The two microscope objectives used in the Wyko NT-2000 have 0.3 and 0.55 of N.A. respectively and each provides 240 µm × 180 µm and 610 µm × 460 µm of FOV.Its spatial resolution is 1.33 µm.
In order to characterize surface data correctly, surface data is decomposed with Zernike polynomials, as in Eq. ( 5), where ) , ( y x S is a measured surface data and i c is a Zernike coefficients and ) , ( y x Z i is Zernike polynomial.We used 36 Zernike polynomials whose equations are listed in [19].There are two reasons why this decomposition is useful to characterize a surface.The first one comes from measurement process.In the measurement of a surface shape, it is unavoidable to introduce unknown tilt and de-center.If a decomposition is successful, the contribution from tilt and de-center can be eliminated by discarding first three Zernike coefficients (Piston and Tilt).The second reason is that decomposition can extract specific information that one wants from measurement data.For instance the decomposition readily separate rotational symmetry and non-rotational symmetry information of surface data.The decomposition is implemented by singular value decomposition (SVD) in commercial software [20,21].
Surface roughness data in Table 5 is defined as RMS error between measured surface data and designed surface data after disregarding first three Zernike coefficients.The unit is nanometer.The surface roughness of hybrid lenses is higher than previously reported in Ref. [1] because we use a different definition of surface roughness here.In this paper, the surface roughness includes the departure of a measured surface from the correct surface shape, which is more rigorous, and not considered in Ref. [1].FOV 1 has 240 µm × 180 µm of FOV and FOV 2 has 610 µm × 460 µm of FOV.RMS roughness of hybrid lenses is controlled to within λ/4 (207.5 nm) in FOV1 while it increases above λ/4 in FOV 2. Comparing the hybrid lens roughness to that of the objective lens shows that the RMS roughness of hybrid lenses is roughly 3 times bigger than commercial optics.This high level of RMS roughness has a detrimental influence on imaging performance, shown in previous section.Degradation of surface qualities at larger apertures introduces high background noise such as stray light from scattering and limits the usable aperture size of hybrid lenses, reducing the actual numerical aperture of the 4M device.RMS roughness of the objective lens in FOV 2 was not measured because the surface of the objective lens is too steep for Wyko NT 2000 to collect reflected light.
In Eq. ( 6), r is the radius of curvature and i c is a Zernike coefficient of i th Zernike polynomial.ρ is a radius of measured surface.If a surface is spherical, only 3 c is non-zero because other rotationally symmetric terms have aspherical contributions.Comparing with design parameters in Table 2, there are noticeable errors in radii of curvature of the objective lens and hybrid lens 1 that cause can cause imaging performance degradation as well as change first order properties [see Table 4].Although radii of curvature and RMS roughness are popular metrics for representing surface characterization, these metrics show only first order and second order statistics of surface data.It is sometimes necessary to make a closer observation to surface data, especially when optics is fabricated by a novel process.Figure 7 shows reconstructed Zernike coefficients of surface data.Each plot in Fig. 7  The same surface data of the objective lens and three hybrid lenses are shown in terms of contour plot in Fig. 8. Contour plots are graphical indicators of uniformity of radius of curvature and non-rotationally symmetric shapes.One can observe the defects of surfaces visually unnoticed by simple metrics reported above.It is shown that contours of hybrid lens surfaces are distorted and departed from spherical shape.The black spots are missing data due to lack of reflectance from testing surface.Since the objective lens has a high slope, the reflectance does not go back to the optical profiler, which causes the data missing in plot (a).

Qualitative assessment: Imaging experiment in trans-illumination mode
The first object imaged with the optics-only prototype 4M device is a polystyrene bead (PS06N/2110) whose nominal mean diameter is 5.10 µm (Bangs Laboratories, Inc) [22].Since illuminating light comes from beneath the object, negative contrast is expected because a polystyrene bead is equivalent to an opaque disk.
In order to evaluate the imaging performance for actual objects, we used a suspension of 5 micron polystyrene beads in solution of 3 mg/ml bovine serum albumin (BSA) in water.The polystyrene beads were diluted 100 or 1000 times from the stock suspension (10% w/w) before the measurements.
Figure 9 shows images of polystyrene beads in trans-illumination mode.Each image shows negative contrast (i.e., a dark ring pattern) around the microspheres as expected if the beads are in the object plane.The dark ring pattern is collapsed when the bead is out of focus as shown in Fig. 9(b).It is notable that there are two ring patterns clearly distinguishable even when two beads are stuck together, as in Fig. 9(d).The second object consisted of cervical cancer cells (SiHa [23]).SiHa cells are derived from a squamous cell carcinoma of the cervix.SiHa cells were obtained from the Lotan laboratory (University of Texas M.D. Anderson Cancer Center) and grown to confluency in a 75 cm 2 tissue culture flask with DMEM/F12 (Invitrogen, CA) supplemented with 5% FBS (HyClone, Logan, UT) and penicillin and streptomycin (Invitrogen).Cells were harvested using 0.25% trypsin-EDTA (Invitrogen) and re-suspended in phenol-red free DMEM/F12 (Invitrogen) at a concentration of 5~10 million cells/ml.
The cells in Fig. 10 have diameters on the order of 10~15 µm.The cell membrane is shown due to refractive index mismatch between cells and media that is known to be 05 .0 ≈ ∆n [24,25,26].The refractive index mismatch is much smaller than mismatch between polystyrene bead and media ( 23 .0 ≈ ∆n ) so that contrast in Fig. 10 is much weaker than Fig. 9 [27].

Conclusion
We report on the current development of the 4M device and represent quantitative and qualitative assessment of the device's imaging performance.The imaging experiments reported in this paper were carried out using monochromatic trans-illumination of the object.Although the 4M device has the potential to become a powerful and versatile tool for early detection of pre-cancer, it is important to evaluate imaging performance while in development of the device.The imaging quality assessment of 4M device is the main purpose of this paper.In order to fulfill this requirement, we built a multi-modal imaging test-bed that can assess the imaging quality of 4M device under different imaging modalities.The multi-modal imaging test-bed delivers illumination to the 4M device and relays 4M-device images without artifacts to a CCD camera.
The first-order properties of the 4M device such as field of view, magnification, working distance and depth of field were measured and reported.For the assessment of imaging performance, we measured the MTF for a quantitative assessment and took images of microspheres and cervical cells for a qualitative assessment under trans-illumination.We demonstrated that 4M device has a potential to be a legitimate medical imaging tool by presenting imaging experimental results.Each optics surface was decomposed by Zernike polynomials that provide quantitative information about surface shape.We found that the main source of imaging performance degradation is surface inaccuracy of hybrid lenses.Due to the error on radii of curvature, first order properties of actual 4M device are a little different from designed parameters.
The 4M device shows encouraging imaging results (see Fig. 9 and Fig. 10).The device can resolve nuclei of cervical cancer cells under very low contrast conditions (unstained, trans-illumination).In order to recognize abnormalities of cancer cells such as nuclear size variation, an imaging tool must be able to resolve features smaller than 2 µm which is equal to the line width of 250 line/mm.The MTF curve presented here indicates the possibility that a 4M device would satisfy this requirement if surface accuracy hybrid lenses are improved.
The next step of 4M device development is to improve surface accuracy of hybrid lens and to test more imaging modalities with larger N.A, e.g., epi-illuminated reflectance, fluorescence imaging and/or optical sectioning which are not implemented yet.These imaging modalities require higher quality of hybrid lenses than trans-illumination imaging modality.The image contrast is much higher when unstained cells are imaged in reflectance or fluorescence mode than in trans-illumination mode as reported here; thus we anticipate that delineation of pre-cancer can be achieved definitely with successful implementation of multimodalities.Possible background noise can be diminished by improving surface quality and accuracy of hybrid lenses and AR-coating all 4M-device optics at the operating wavelengths.Such improvements are expected to make reflectance imaging and optical sectioning possible.Finally, we note the possibility of using the hybrid lens manufacturing process to make usable yet disposable large NA lenses capable of high resolution for medical diagnostic purposes.

Fig. 1 .
Fig. 1.Conceptual geometry of 4M device.Every component is mounted on one silicon substrate [a.k.a., the Micro Optical Table, MOT] including an imaging sensor and an illumination system.The substrate dimensions are 13mm (L) × 10 mm (W).Blank lines are geometrical rays traced non-sequentially in the 4M device.

Fig. 2 .Fig. 3 .
Fig. 2. Lens Design of 4M Device.4M devices consist of one objective lens from Edmund industrial optics and three hybrid lenses that are fabricated by photo-lithography.The object is assumed to be in water.

Fig. 4 .
Fig. 4. Schematic diagram of multi-mode imaging test-bed: (a) Top view of entire system.It shows two fibers for two different illumination modes: Fiber A for trans-illumination mode and Fiber B for epi-illumination mode.(b) Side view of region inside the blue box in (a).It shows trans-illumination mode.

Fig. 5 .
Fig. 5. Actual multi-modal imaging test-bed in figure (a).The 3-axis actuators shown within the red oval are used to control the object, the 4M device and the trans-illumination system.The 4M device is located inside the oval.Figure (b) provides the magnified view of the 4M device being used with an imaging chamber used to hold cells.An object, usually cells in liquid media, is contained in sample holder underneath 4M device.

Fig. 6 .
Fig.6.MTF curve for the 4M device reported in this paper (Green Line).The black dots are the measured CTF data and the red line is the interpolated CTF.The blue line above other lines is the diffraction limited MTF of a 4M device with NA = 0.15.

Fig. 7 .
Fig. 7. Zernike coefficients of objective lens and 3 hybrid lenses: (a) objective lens, (b) hybrid lens 1, (c) hybrid lens 2, and (d) hybrid lens 3. Red triangles are correct Zernike coefficients and black rectangles are reconstructed Zernike coefficients from measured surface profiles.

Fig. 9 .
Fig. 9. Images of polystyrene beads in trans-illumination mode in figure (a) and (b).As expected, the negative contrasts (dark ring patterns) are visible around polystyrene beads in all parts of the Figure.Parts (c) and (d) show that two polystyrene beads can adhere to each other and still be clearly resolvable.Parts (b) and (d) are the magnified sections indicated by red boxes in Parts (a) and (c), respectively.

Fig. 10 .
Fig. 10.Images of cervical-cancer cells (SiHa) in trans-illumination mode.The cell membrane is shown due to small change of refractive index between cells and media (a).Figure (b) shows the magnified view of region inside red box in figure (a).Within the cell images in figure (b), the unstained nuclei are visible.
Fig. 10.Images of cervical-cancer cells (SiHa) in trans-illumination mode.The cell membrane is shown due to small change of refractive index between cells and media (a).Figure (b) shows the magnified view of region inside red box in figure (a).Within the cell images in figure (b), the unstained nuclei are visible.

Table 1 .
First order design of 4M devices

Table 2 .
Lens prescription data of 4M devices

Table 3 .
RMS wavefront error in waves of three hybrid lenses used in 4M device

Table 4 .
First-order properties of 4M device.

Table 5 .
Surface roughness of objective lens and hybrid lenses.Surface roughness is defined as RMS error between measured surface data and designed surface data.The unit is nano-meter.FOV 1 has 240 µm × 180 µm of FOV and FOV 2 has 610 µm × 460 µm of FOV.Surface accuracies are reported in terms of radii of curvature in Table6.Radii of curvature are calculated by Zernike coefficients whose Zernike polynomials are rotationally symmetric and includes 2 nd order dependence on radius such as 3

Table 6 .
Radii of curvature of objective lens and hybrid lenses in mm.