Liquid-shaped microlens for scalable production of ultrahigh-resolution optical coherence tomography microendoscope

Optical coherence tomography (OCT) enables label-free, real-time, three-dimensional imaging of luminal organs with 1–3 mm tissue penetration and near-histologic microstructural detail. Most endoscopic OCT systems operate at 1300 nm with ~10 µm resolution. Operating at ~800 nm improves resolution to ~2–4 µm and enhances contrast but reduces imaging depth. Imaging small and convoluted lumens (e.g., narrow arteries, peripheral bronchioles) requires ultrathin, flexible microendoscopes. An 800-nm OCT endoscope with ultrahigh resolution (axial <3 µm in air) and ultrathin form factor (<1 mm OD) is desirable to detect subtle pathology with minimal invasiveness. Conventional miniature OCT probes use GRIN optics or fiber ball lenses but suffer from chromatic aberration at 800 nm, limited achromatization options, low transmission for diffractive corrections, restricted design flexibility with fiber-melted ball lenses, and labor-intensive angle-polished reflectors. Recent two-photon 3D microprinting can form freeform micro-optics but is costly, less scalable, and has insufficient surface quality for OCT. The study aims to develop a scalable liquid shaping technique to fabricate custom, aberration-corrected, sub-nanometer-roughness microlenses on fiber tips for ultrathin, ultrahigh-resolution 800-nm OCT microendoscopy.

Prior OCT endoscopes at 1300 nm deliver ~10 µm resolution but are limited for detecting subtle pathology. Shifting to ~800 nm improves resolution (~2–4 µm) and contrast but demands better aberration control in miniature optics. GRIN-based microprobes at 800 nm exhibit strong chromatic aberration. Achromatic diffractive correctors suffer from lower transmission efficiency. Fiber-melted ball lenses can be angle-polished to achieve achromatization but offer limited control over working distance, resolution, and depth of focus, and require labor-intensive polishing that can degrade surface quality. Two-photon 3D microprinting enables freeform side-deflecting micro-optics and submillimeter probes, yet remains expensive, lacks scalability, and yields surface roughness on the order of 10–200 nm, suboptimal for OCT imaging. These gaps motivate a new, scalable method to form custom, aberration-corrected microlenses with superior surface finish and flexible geometry control.

The authors developed a liquid shaping technique to form freeform microlenses by controlling the minimum energy state of a curable optical liquid (NOA 81) on substrates with tailored wettability and boundary geometry. Key steps: (1) Liquid dispensing: A piezoelectrically actuated dispenser with thermal control (~75 °C) reduced viscosity (from 300 to ~30 cps) to precisely dispense optical liquid with ~0.1 nL volume precision. (2) Wettability control: Glass and 3D-printed PEGDA cylinder substrates (circular/elliptical boundaries, ~400 µm height) were oxygen-plasma treated then fluorinated using POTS to tune contact angles (measured precision ±0.1°). This allowed continuous tuning of droplets’ contact angles from ~50° to ~110°. (3) Geometry control: On boundary-free glass, lens radii (120–720 µm) were tuned by volume (5–800 nL) at fixed contact angle (e.g., 90°). With physical boundaries, spheroid (circular aperture) and ellipsoid (elliptical aperture) lenses were formed, with shape governed initially by wettability and volume and ultimately by boundary constraints. (4) Polymerization and shrinkage compensation: Droplets were UV-cured (365 nm, 65 mW/cm²) for ~30 min. A consistent polymerization shrinkage of 7.69% (volume) preserved contact angle and profile; volumes were pre-compensated by 8.33% to reach target geometry. (5) Microendoscope assembly: A single-mode fiber (780HP) was spliced to a non-core fiber (FG125LA) and cleaved to 500 ± 5 µm. A semi-spherical microlens (target radius 140 µm, contact angle ~90°) was aligned and bonded to the fiber tip at an incident angle of 52.5° using NOA 81 adhesive and UV curing, maintaining ~5 µm standoff during bonding. Five probes were assembled simultaneously on a custom four-dimensional stage with top and side inspection microscopes; one-way transmission efficiency ≥94%. Probes were packaged as rigid (hypodermic tube plus glass capillary, OD ~617 µm, wall ~40 µm) or flexible (torque coil OD 450 µm, FEP sheath OD 626 µm, wall 45 µm) microendoscopes, total length ~215 mm. Total fabrication time per batch ~90 min for five devices. (6) Optical design and simulation: Using Zemax OpticStudio (v17), material dispersion of silica and NOA 81 was included. Non-sequential stray light analysis evaluated back-reflection versus incident angle, lens radius, and NCF length. Mixed sequential/non-sequential ray-tracing (source NA 0.13, 750–950 nm bandwidth) provided chromatic focal shift, focused spot size, astigmatism ratio, effective depth of focus (DOF), and working distance across NCF lengths and lens radii. An optimal design region was identified; the chosen design used NCF length 500 µm, lens radius 140 µm, incident angle 52.5°. (7) Characterization: Lens topography was measured with a 3D confocal surface profiler; surface roughness with white-light interferometry. An 800-nm spectral-domain OCT system characterized axial resolution and achromaticity; the focused beam profile (out of protective sheath) was measured with an optical beam profiler to obtain focused spot sizes, DOF, and astigmatism. (8) Animal imaging: Ex vivo rat esophagus and mouse aorta imaging used euthanized animals; perfusion (mouse aorta) removed blood. The flexible microendoscope performed pullback imaging at 10 frames/s and 20 µm frame pitch. For in vivo mouse brain, burr holes were made; a rigid probe was inserted at ~10 µm/s, imaging a 5-mm-deep cylindrical volume in ~50 s. Post-imaging, tissues were fixed and processed; H&E histology was correlated with OCT.

- Fabrication and scalability: Five ultrathin 800-nm OCT microendoscopes (rigid and flexible) were fabricated simultaneously in ~90 min with comparable performance; total probe diameters (with sheath) ~0.6 mm. Microlens geometry control was precise (measured semi-spherical lens radius 140.5 ± 0.5 µm). UV-cured microlenses exhibited a consistent volume shrinkage of 7.69% with negligible profile change upon curing. - Surface quality: Sub-nanometer surface roughness was achieved: 0.84 ± 0.11 nm RMS on the curved surface and 0.53 ± 0.11 nm on the flat reflective surface, minimizing scattering. - Optical design/performance: Selected design (NCF 500 µm, lens radius 140 µm, incident angle 52.5°) achieved simulated chromatic focal shift ~5.4 µm across 750–950 nm, focused spot size ~4.6 µm, astigmatism ratio ~1.05, effective DOF ~197 µm, working distance ~238 µm, and back-reflection < −56 dB. - Experimental beam metrics: Measured focused spot sizes out of the protective sheath were ~4.5 µm (x) and ~4.3 µm (y) at ~240 µm from the sheath surface (mean 4.5 µm; astigmatism ratio 1.05). Effective DOF ~200 µm. - Axial resolution and achromaticity: Axial resolution ~2.43 µm in air with <5% variation over 1-mm imaging depth; back-reflected spectra remained stable with axial mirror translation, confirming achromatic behavior. - Esophagus imaging (rat): 3D imaging (10 fps, 20 µm frame pitch) over 36 mm length revealed clear layered microanatomy (stratified squamous epithelium, lamina propria, muscularis mucosae, submucosa, circular and longitudinal muscle) in agreement with H&E histology. - Aorta imaging (mouse, descending thoracic, lumen ~0.6 mm): 14.6 mm length imaged (10 fps, 20 µm pitch) showed tunica intima (TI), media (TM), adventitia (TA), adipose tissue, and elastic lamellae. Quantified layer thicknesses (n=4 mice; 40 A-lines across 10 sections) were TI 11.0 ± 1.2 µm, TM 69.2 ± 3.4 µm, TA 28.2 ± 2.2 µm. - Deep brain imaging (mouse, in vivo): A 5-mm-deep cylindrical volume imaged in 50 s revealed cerebral cortex, corpus callosum, caudate putamen, ventral striatum; en face projection matched histology. Fine striatopallidal fiber bundles in caudate putamen were visualized. - Back-reflection dependence: Simulations showed back-reflection decreased with increasing lens radius and incident angle; weak dependence on NCF length. An incident angle of 52.5° with 125–150 µm lens radii yielded < −56 dB back-reflection.

The work demonstrates that liquid shaping enables precise, scalable fabrication of custom freeform microlenses with sub-nanometer surface roughness directly on fiber probes to produce ultrathin (≈0.6 mm OD), ultrahigh-resolution 800-nm OCT microendoscopes. By tuning substrate wettability, droplet volume, and physical boundary, lens geometry can be engineered to correct chromatic aberration, optimize working distance, DOF, and spot size, and minimize back-reflection without angle polishing. The resulting devices achieve axial resolution ~2.43 µm, transverse ~4.5 µm, ~200 µm DOF, and stable spectral response across 750–950 nm. In situ imaging in small lumens (rat esophagus, mouse aorta) and minimally invasive interstitial deep-brain imaging validate clinical and preclinical relevance, enabling delineation and quantification of fine tissue layers (e.g., aortic TI, TM, TA) and microstructures (elastic lamellae, fiber bundles). Compared to GRIN or fiber-melted ball lens approaches and 3D microprinting, this method offers superior surface quality, optical performance at 800 nm, reduced fabrication complexity and cost, and batch scalability. The findings address the need for ultrathin, flexible, high-resolution OCT microendoscopes capable of imaging small and tortuous lumens and deep solid tissues with minimal invasiveness, highlighting potential for improved diagnostic yield and guidance of interventions.

This study introduces a liquid shaping technique to fabricate custom, aberration-corrected microlenses with sub-nanometer roughness for ultrathin 800-nm OCT microendoscopes. The approach enables batch, rapid, low-cost production of rigid and flexible probes (≈0.6 mm OD) that deliver achromatic ultrahigh resolution (axial ~2.43 µm; transverse ~4.5 µm), appropriate working distance (~238 µm), low back-reflection (< −56 dB), and large effective DOF (~200 µm). Imaging of rat esophagus, mouse aorta (with quantitative layer measurements), and in vivo mouse deep brain demonstrates high-quality visualization of fine microanatomy. Future work includes adding active control modalities (thermal/electromagnetic) to expand freeform shaping capabilities, further automation for higher throughput and yield, validation in large-animal models and comprehensive safety assessments, increasing imaging speed to reduce motion artifacts and procedural time, and extending the approach to other fiber-based modalities (confocal, two-photon, coherent Raman).

The current work demonstrates feasibility with passive lens-shaping controls (substrate wettability, volume, boundary) and small-animal studies. Limitations include: reliance on passive shaping without active control of complex freeform surfaces; limited batch size (five devices) and need for further automation to improve yield and reduce cost; imaging speed (10 fps) that may be insufficient for motion-prone clinical applications; and absence of large-animal and human validation and safety studies. While long-term stability is promising, comprehensive durability and sterilization assessments for clinical use remain to be completed.