Spectrally-encoded non-scanning imaging through a fiber

Ningzhi Xie, Quentin A. A. Tanguy, Johannes E. Fröch, K. F. Böhringer, Arka Majumdar

Abstract: With the advent of neuroimaging and microsurgery, there is a rising need for capturing images through an optical fiber. We present an approach of imaging through a single fiber without mechanical scanning by implementing spatial-spectral encoding. The spectral encoding is achieved through a microfabricated spectral filter array, where light from different spatial pixels is coded with a highly orthogonal spectrum. The image is then computationally recovered via pseudo inverse of the encoding process. We demonstrate imaging of a $4 \times 4$ binary object at the proximity of the spectral filter array using $560-625nm$ wavelength band. The recovered image maintains an error rate of $<11\%$ when measured using a spectrometer with a spectral resolution of $1.5nm$. The image remains unchanged with fiber bending or moving. Thus our approach shows a more robust way to image through a single optical fiber, with potential applications in compact endoscopes and angioscopes.

Introduction

Traditional endoscopic imaging faces a significant trade-off between device miniaturization and image stability. Mechanical scanning techniques require bulky actuators at the fiber tip, limiting their use in the narrowest biological channels. Coherent fiber bundles offer flexibility but are constrained in resolution by the number of individual cores. Multi-mode fibers (MMF) have emerged as a high-resolution alternative, yet they are notoriously sensitive to environmental changes; bending or moving the fiber alters its transmission matrix, necessitating constant and complex recalibration. This paper presents a robust, non-scanning imaging approach using Spatio-Spectral Encoding (SSE) to overcome these hurdles.

Spatio-Spectral Encoding System

The core innovation is a microfabricated spectral filter array placed at the distal tip of a single optical fiber. This array "tags" each spatial pixel with a unique, designable spectral signature.

Schematic of the spatio-spectral encoding (SSE) imaging system showing the object, filter, multi-mode fiber, and reconstruction process.

The device is constructed as a Fabry-Pérot (FP) cavity, consisting of a dielectric metasurface sandwiched between two Distributed Bragg Reflectors (DBRs). The metasurface is composed of Silicon Nitride (SiN) nano-posts. By precisely varying the lateral dimensions of these posts, the researchers control the local phase shift, which in turn shifts the resonance peak of the FP cavity for each of the $4 \times 4$ pixels in the array.

Experimental Characterization

The researchers validated the system's design through detailed fabrication and spectral analysis.

Experimental setup and characterization: (a) Optical setup, (b) micrograph of the filter array, (c) SEM of nano-posts, (d) individual pixel spectra, and (e) combined pattern spectrum.

Each pixel is assigned a narrow-band spectral "code." A critical insight of this work is the importance of orthogonality: by designing filters with minimal spectral overlap, the decoding process becomes significantly more stable. When light from an object passes through the array, spatial information is converted into the frequency domain. This encoded light is transmitted through a single fiber to a spectrometer. Because the spatial mapping is fixed at the tip, the image can be recovered using a simple linear reconstruction (a pseudo-inverse of the encoding matrix) regardless of fiber deformation.

Results and Robustness

The system was tested using 24 different $4 \times 4$ binary patterns, including letters like "G". The imaging system achieved an average error rate of 9.8% using high-resolution spectral data.

A major advantage of this approach is its immunity to fiber bending. Since the spatial-to-spectral mapping occurs before the light enters the fiber, the fiber acts merely as a "bucket" to collect the signal. The reconstruction remains accurate even when the fiber is coiled or moved. Additionally, the system demonstrated flexibility regarding hardware requirements; the error rate only slightly increased to 11% when the spectral resolution was reduced to 1.5nm, suggesting compatibility with compact, low-cost spectrometers.

Conclusion and Future Work

While the current proof-of-concept is limited to 16 pixels within the $560–625nm$ band, the technology is highly scalable. Increasing the pixel count to high-definition resolutions would require broader spectral bandwidths or higher Q-factor resonances, potentially achievable through photonic crystal slabs. Furthermore, integrating a lens system at the fiber tip would enable far-field imaging. This SSE approach offers a promising path toward ultra-thin, robust endoscopes for neuroimaging, angioscopy, and other minimally invasive medical procedures.

Summary updated

Abstract

With the advent of neuroimaging and microsurgery, there is a rising need for capturing images through an optical fiber. We present an approach of imaging through a single fiber without mechanical scanning by implementing spatial-spectral encoding. The spectral encoding is achieved through a microfabricated spectral filter array, where light from different spatial pixels is coded with a highly orthogonal spectrum. The image is then computationally recovered via pseudo inverse of the encoding process. We demonstrate imaging of a $4 \times 4$ binary object at the proximity of the spectral filter array using $560-625nm$ wavelength band. The recovered image maintains an error rate of $&lt;11\%$ when measured using a spectrometer with a spectral resolution of $1.5nm$. The image remains unchanged with fiber bending or moving. Thus our approach shows a more robust way to image through a single optical fiber, with potential applications in compact endoscopes and angioscopes.
BibTeX
@inproceedings{xie2024spectrally,
  author = {Ningzhi Xie and Quentin A. A. Tanguy and Johannes E. Fröch and K. F. Böhringer and Arka Majumdar},
  title = {{Spectrally-encoded non-scanning imaging through a fiber}},
  booktitle = {ACS},
  year = {2024},
  doi = {10.1021/acsphotonics.3c01582},
}
Figure