Super-Achromatic Rapid Scanning Microendoscope for Ultrahigh-Resolution OCT Imaging

A. Microlens Design

The design of a miniature achromatic compound lens took into account a number of challenges, including: 1) small diameter (1.5 mm in optics and 2 mm with a housing tube); 2) superb secondary spectrum control over the broad spectrum of an OCT light source (i.e., 650–950 nm); 3) low magnification to image the beam at the fiber tip with a size of the fiber mode-field diameter to a small spot size with an ~1:0.8 reduction ratio or better; and 4) good telecentricity so that the beam after the compound lens will be nearly perpendicular to the surface of the target tissue during lateral scanning. With these considerations in view, a preliminary design of a multiple-element compound lens for the scanning microendoscope was generated. All elements had a 1.5 mm diameter (with a 1.4 mm clear aperture) so that they could be housed in a protective tube of a 2 mm outer diameter. A 0.5-mm-thick protective glass window was also considered. Beyond the window, body fluid of a refractive index of ~1.38 was assumed. The designed working distance between the last lens element and the focal plane was about 2.8 mm, which provides ample room for the lens assembly to translate with respect to the window for performing focus tracking over a depth of 1–2 mm, if needed. The target beam scanning range was 320 μm in any transverse direction. Glasses of various refractive indices and dispersion properties were used for achieving sufficient beam bending and chromatic correction. This preliminary design had excellent achromatic characteristic, and the on-axis focus position at 700, 800, and 900 nm differed by less than 2 μm, as shown in Fig. 1. A modulation transfer function (MTF) analysis indicated that an approximately 45% image contrast would be achieved even at a frequency of 200 line pairs/mm (corresponding to a 5-μm resolution) at a test wavelength of 632 nm (see Fig. 2). The design also took the field curvature traced out by the scanning fiber tip into consideration (see discussion in the endoscope design, Section III-A). The maximum deviation from perfect telecentricity was less than 0.3°, which occurred near the edge of the scan field and at the edge of the wavelength band. This deviation is small enough to keep the change in the collection of the backscattered light by the fiber within a few percent.

B. Microlens Performance Characterization

The achromatic microlens has been successfully fabricated. The end surfaces (except the outer surface of the window that was designed to directly contact body fluid) were antireflection (AR)-coated over a broad spectrum bandwidth (650–1050 nm). All the lenses were housed in a thin-wall hypodermic tube with an overall diameter of 2 mm. A photo of the finished miniature compound lens is shown in Fig. 3(A). The measured MTF is shown in Fig. 2, which is very close to the theoretical design. Even at 250 lines/mm spatial frequency, the measured MTF is about 0.25, suggesting that an image contrast of ~1.7 is achievable for resolving a structure with a 4 μm spatial dimension. Once implemented in a scanning fiber-optic endoscope (detailed in the next section), we were able to achieve a measured transverse resolution of about 4 μm with a fiber tip of a 4.6 μm mode-field diameter. This corresponds to a magnification of M = −0.87, which is within our design tolerance (considering an expected value of M_theory = −0.8). Compared to a previous endoscope made of a GRIN lens [9], the use of the achromatic microlens dramatically improves the transverse resolution by more than a factor of 2.5.

In order to characterize the chromatic aberration, a customized Ti:Sapphire laser with an ~150 nm full-width at half-maximum (FWHM) spectrum bandwidth was used as a broadband light source. The laser light, sent through the fiber-optic microlens-equipped endoscope, was initially focused on a mirror placed at the focal plane. The spectra backreflected by the mirror were measured when the mirror was placed at the focal plane and at other planes with different distances away from the focal plane [see Fig. 3(B)]. As shown in Fig. 3(C), the backreflected spectra exhibit minimal change over a large distance from 0 to 3 in the units of the confocal parameter of the microlens b_microlens (30.5 μm), indicating the superb achromaticity of the microlens. In comparison, the same experiment was performed with an endoscope made of a 1.8 mm GRIN lens. The backreflected spectra were again measured with the mirror placed at planes over a distance of 0–2 from the focal plane in the units of the GRIN lens confocal parameter b_GRIN (350 μm), and the spectra are shown in Fig. 3(D). Obviously, the backreflected spectra experienced dramatic changes due to the strong chromatic aberration of the GRIN lens. The superb achromatic performance of the microlens is critical to maintain a high axial resolution at different depths for OCT imaging.

A. Basic Layout for the Scanning Microendoscope

The small size, superb achromatic performance, and the capability of producing a tight beam focus make the compound microlens very suitable for implementation in a scanning fiber-optical microendoscope. Compared to many other customized microlenses [16], the one reported here has a much smaller size (e. g. 1.5 mm versus 4 mm).

With the miniature compound objective lens, we have developed a scanning fiber-optic UHR endoscope with an overall diameter of 2.4 mm. Fig. 4(A) shows the schematic of the fiber-optic scanner head for the endoscope that adopts a layout similar to that in [9]. The use of the miniature compound lens dramatically improves the transverse resolution (1/e² diameter) from the previous ~13–16 μm (GRIN lens) to ~4 μm in addition to the benefit of achromaticity. In essence, the endoscope consists of a tubular piezoelectric actuator, a single-mode fiber (SMF) cantilever going through the piezo tube and glued to the end of the tube, and the compound microlens. A photo of the entire flexible microendoscope is shown in Fig. 4(B), and the length of the rigid portion at the distal end (including the piezo actuator) is about2.5–3.0 cm, depending upon the piezo tube used. The small endoscope diameter was intended to facilitate its delivery to internal organs [e.g., gastrointestinal (GI) tract] through the working channel of a standard GI endoscope. Rapid lateral beam scanning is achieved by: 1) driving the piezoelectric actuator at the mechanical resonant frequency of the fiber-optic cantilever; 2) imaging the sweeping fiber tip to the target through the microlens. In our case, the resonant frequency was about 1.22 kHz for an 8.2-mm-long SMF cantilever, and the fast beam scanning permits OCT imaging in real time. In addition to rapid 1-D lateral beam scanning, the fiber-optic scanner is also capable of circular or 2-D spiral scanning [17]. Compared to a recently reported superb side-viewing UHR endoscope [18], the UHR microendoscope presented here offers better transverse resolution (i.e. 4 μm versus 7.5 μm 1/e² diameter) and significantly higher imaging speed (i.e., 1220 scans/s versus 240 scans/s). The piezo drive voltage is generally not high, e.g., for about ±3.0° angular fiber cantilever scanning, the peak-to-peak drive voltage is only about 40 V. This corresponds to a lateral sweeping distance of approximately 0.86 mm by the tip of an 8.2-mm-long fiber cantilever, and an ~0.69 mm beam focus scanning when using an imaging lens of 0.8 magnification (as in our URH microendoscope).

B. Real-Time Imaging with the Scanning Microendoscope and Performance Comparison

In order to generate a Doppler frequency shift and perform highly sensitive heterodyne detection in the time domain, a pair of acousto-optic modulators (AOMs) were used in the reference arm of our time-domain OCT system, with one AOM upshifting the light frequency and the other downshifting the frequency, producing a net frequency shift of about 1.0 MHz for a double-pass configuration [19], [20]. The dispersion of the AOM crystals in the reference arm and the microlens in the endoscope (i.e., in the sample arm) were compensated by using our previously developed method based on the combination of a single-mode fiber and a pulse stretcher [19], [20]. To fine-tune the dispersion compensation, a CaF₂ glass window and a pair of BK7 wedges were placed in the reference arm of the OCT system. Fig. 5 illustrates the interference signal when using a light source (Ti:Sapphire laser) with a center wavelength of 825 nm and a spectrum bandwidth of ~150 nm. The measured axial resolution (i.e. the FWHM of the envelope of the interference signal in Fig. 5) was approximately 3 μm in air (corresponding to ~2.2 μm in tissue when the tissue refractive index of ~1.38 is taken into account). The interference signal is nearly symmetric, indicating the dispersion in the two arms of the OCT system was reasonably balanced. Owning to the broadband AR coating on the microlens, the loss of the entire endoscope including the uncoated fiber tip was low (i.e., with a throughput of approximately 90%).

The performance of the scanning microendoscope was tested by imaging different tissues ex vivo with a measured imaging resolution of 4.0 × 3.0 μ m (transverse × depth in air). The light power delivered to the sample was about 6.5–7 mW, and the OCT system sensitivity was about 96 dB. A representative image of rat tail acquired by the scanning endoscope in real-time is shown Fig. 6(A), where the tendons (i.e., collagen fiber bundles) can be clearly identified as indicated with the red arrows. The image size was 811 × 1230 pixels, corresponding to a physical dimension of 0.8 × 0.6 mm (transverse × depth). The 1.22 kHz lateral beam scanning rate (e.g., 1220 lateral scans per second) permitted real-time imaging at a speed of about 1.5 frames per second (with each frame consisting of 811 lateral scans). For comparison, the same rat tail was also imaged by a bench-top time-domain UHR OCT system at nearly the same position, with an imaging speed of ~100 depth scans per second, a resolution of approximately 5.0 × 2.8 μm (transverse × depth in air) and an image size of 800 × 1000 pixels (0.8 × 0.6 mm, transverse × depth) [Fig. 6(B)]. It should be mentioned that the image acquisition sequence with the scanning endoscope is fast lateral scanning followed by a slow depth scanning, which is opposite to the image acquisition sequence by the bench-top system (which performed fast depth scanning followed by slow lateral scanning). As evident in Figs. 6(A) and (B), the image quality achieved by the rapid scanning microendoscope is very comparable to the one achieved by the bench-top system even though the bench-top system affords the use of large optics. One significant difference, however, is that the UHR imaging speed enabled by the endoscope is at least 12 times faster than that of the bench-top system.

We have also carried out another comparison study by imaging the same tissue (e.g. rabbit trachea ex vivo) with a scanning endoscope made of the achromatic compound microlens and by a scanning endoscope made of a 1.8-mm-diameter GRIN lens. Representative images of rabbit trachea are shown in Fig. 7. The two images were acquired at two frames per second and have a physical dimension of about 0.8 × 0.35 mm. The image resolution was ~4.0 × 3.0 μm for the compound microlens endoscope and ~13.5 × 3.5 μm for the GRIN lens endoscope (transverse × depth in air). Fine tissue structures such as the epithelium (E), lamina propria (LP), submucosa (SM), and glands (G) are evident on both images. Although the depth of focus for the microlens endoscope is much shorter than for the GRIN lens endoscope (i.e., 30.5 μm versus 350 μm), the image quality achieved by the microlens endoscope [Fig. 7(A)] deteriorates much slower than the one acquired by the GRIN lens endoscope [Fig. 7(B)]. In particular, the cartilage can be clearly identified in Fig. 7(A), while it is barely visible in Fig. 7(B). These findings strongly suggest the superior performance of the compound microlens endoscope over a conventional GRIN lens endoscope, and the excellent achromaticity of the compound microlens could be the key factor contributing to the superb performance of the microendoscope made of the microlens. We also noticed that the imaging depth in Fig. 7(B) seems shallower than the one in [15], where rabbit esophagus was imaged with a similar GRIN lens endoscope, and this depth reduction most likely resulted from the difference in tissue types (esophagus versus trachea), tissue freshness, and the incident power [6.5–7 mW for Fig. 7(B) versus 9.3 mW in Reference [15]].