Toward optical coherence tomography on a chip: in vivo three-dimensional human retinal imaging using photonic integrated circuit-based arrayed waveguide gratings

Optical coherence tomography (OCT) is a leading non-invasive ophthalmological imaging technique for visualizing retinal subsurface layers. Its resolution and contrast have significantly improved in recent decades, establishing it as a standard in ophthalmologic care. The current standard, spectral-domain OCT (SD-OCT), employs broadband light in an interferometer. Light reflected from the sample and reference arms interferes, chromatically diverges via a diffraction grating, and is projected onto a camera. Fourier transformation of the spectrum produces a depth profile. Recent advancements have increased SD-OCT performance with wider bandwidth light sources improving axial resolution and faster cameras enabling faster volumetric imaging. However, the size and cost of SD-OCT systems (approximately 1 m³ and up to $100,000) are significant barriers. This necessitates miniaturization and cost reduction. One approach involves using smaller and cheaper components, such as MEMS mirrors for scanning and 3D-printed housings. Another promising approach, requiring further research, is the use of photonic integrated circuits (PICs), compatible with CMOS fabrication for cost-effective mass production. PICs, with their small footprint and monolithic integration of optical and optoelectronic components, can significantly reduce size, cost, and increase stability. While PIC development for OCT applications has progressed using Michelson, Mach-Zehnder, and multimode interferometers in the 1300-1500 nm range, achieving sensitivities up to 91 dB and 13 µm axial resolution, PIC-based SD-OCT requires more complex building blocks. Specifically, the diffraction grating in conventional SD-OCT can be replaced by an arrayed waveguide grating (AWG), enabling spectral separation of light. Previous AWG-based OCT implementations, however, suffer from laborious packaging needs and insufficient sensitivity for in vivo imaging, especially retinal imaging, which requires high acquisition rates to minimize motion artifacts. This paper presents the first in vivo retinal tomograms using AWGs with 256 output channels, eliminating time-intensive averaging. The CMOS compatibility allows for the integration of photodiodes and electronics on the same chip, eliminating the need for an external CCD camera, improving robustness, and simplifying packaging.

The literature review section examines existing OCT systems and their limitations, emphasizing the need for miniaturization and cost reduction. It highlights previous attempts to reduce the size and cost of OCT systems through the use of smaller optical components and low-cost materials. The review also explores the use of photonic integrated circuits (PICs) as a potential solution for creating smaller, more cost-effective OCT systems. Specific examples of previous research using AWGs for OCT applications are presented and their limitations discussed. The authors highlight the drawbacks of previously demonstrated implementations of AWG PICs for OCT, emphasizing the need for higher sensitivities to enable faster image acquisition rates without extensive averaging and the complexities in packaging for in vivo applications. The existing AWG-based OCT systems are found to require laborious packaging due to the need for many external components and also lack the sensitivity required for in vivo imaging, especially retinal imaging. The need for higher sensitivities and faster acquisition rates is particularly important in retinal imaging to minimize the effects of eye movement artifacts. This review sets the stage for the authors' contribution, which addresses these limitations by presenting the first in vivo retinal tomograms using AWGs with 256 output channels.

Two compact 256-channel AWGs were designed and fabricated. AWG 1 had a center wavelength of 794 nm and a 0.09 nm wavelength spacing (22 nm bandwidth), while AWG 2 had a center wavelength of 875 nm and a 0.19 nm wavelength spacing (48 nm bandwidth). Both AWGs were 13 x 14 mm² on a 20 x 20 mm² semiconductor chip. Optical characterization involved coupling a tunable Ti:sapphire laser source (800–900 nm) to the AWGs using a polarization-maintaining fiber. Transmission losses were measured for 33 output channels. The AWGs were integrated into a fiber-based interferometer. Sensitivity, roll-off with depth, and axial resolution were measured using a neutral density filter to control signal strength. Maximum sensitivity was determined using the signal-to-noise ratio (SNR) of the point spread function (PSF). Axial resolution was calculated from the full width at half maximum (FWHM) of the PSF. Finally, in vivo retinal imaging was performed on a healthy volunteer's right eye, under an approved protocol, using both AWG systems. The systems were optimized individually resulting in different coupler splitting ratios, powers on the eye and imaging speeds. The system differences however, do not influence the AWG design-specific parameters, such as axial resolution or signal roll-off with depth.

The study successfully demonstrated the first in vivo three-dimensional human retinal imaging using chip-based OCT with AWGs. Two AWG designs were characterized: AWG 1 (22 nm bandwidth) achieved a sensitivity of 91 dB with 830 µW on the sample and an axial resolution of 10.7 µm; AWG 2 (48 nm bandwidth) achieved a sensitivity of 90 dB with 480 µW on the sample and an axial resolution of 6.5 µm. The mean transmission over 33 measured channels was -15.51 dB for AWG 1 and -11.64 dB for AWG 2. The 6-dB roll-off depth was approximately 625 µm for AWG 1 and 380 µm for AWG 2. In vivo retinal tomograms were successfully acquired using both AWG designs, demonstrating the feasibility of this technology for clinical applications. The measured sensitivity, axial resolution and imaging depths for both AWG designs are comparable to the results acquired with a commercial SD-OCT system, supporting the potential of this approach for clinical translation.

The successful in vivo retinal imaging using AWG-based OCT demonstrates a significant advancement towards compact, cost-effective, and robust ophthalmic imaging systems. The achieved sensitivities and resolutions are comparable to commercially available SD-OCT systems, highlighting the clinical potential of this technology. The CMOS compatibility of the fabrication process enables future integration of dedicated photodiodes and electronics, leading to fully integrated chip-based OCT systems. The miniaturization and improved robustness offered by this approach could facilitate the development of point-of-care diagnostic tools for retinal imaging, increasing accessibility and affordability of this crucial technology. While the current study focuses on retinal imaging, the underlying technology is adaptable to other areas needing miniaturized OCT systems. Future research should focus on integrating the light source and further enhancing the sensitivity and imaging speed to improve the quality and efficiency of the OCT imaging. The use of this technology may help reduce costs and increase availability in various clinical settings.

This work successfully demonstrated the feasibility of in vivo three-dimensional human retinal imaging using a photonic integrated circuit-based arrayed waveguide grating. The achieved performance metrics (sensitivity, resolution) are comparable to existing clinical systems, indicating significant potential for miniaturized, robust, and cost-effective ophthalmic imaging. Future work will focus on integrating the light source and enhancing the sensitivity and imaging speed for broader clinical applications. This technology holds promise for increasing accessibility and affordability of OCT technology.

The study utilized only two AWG designs, and further optimization might be possible. While the in vivo results are promising, a larger clinical study is needed to validate the system's performance and reliability in a diverse patient population. Furthermore, potential limitations of the CMOS-compatible fabrication process for extremely high-resolution or high-speed applications should be considered and addressed in future iterations. The current system is limited in terms of imaging depth and the specific wavelength range and therefore additional improvements are needed to meet all requirements for a wide range of clinical applications.