Large-scale compressive microscopy via diffractive multiplexing across a sensor array

Kevin C. Zhou, Chaoying Gu, Muneki Ikeda, Tina M. Hayward, Nicholas Antipa, Rajesh Menon, Roarke Horstmeyer, Saul Kato, Laura WallerarXiv 2025 · DOI · arXiv

Abstract

Microscopes face a trade-off between spatial resolution, field-of-view, and frame rate -- improving one of these properties typically requires sacrificing the others, due to the limited spatiotemporal throughput of the sensor. To overcome this, we propose a new microscope that achieves snapshot gigapixel-scale imaging with a sensor array and a diffractive optical element (DOE). We improve the spatiotemporal throughput in two ways. First, we capture data with an array of 48 sensors resulting in 48x more pixels than a single sensor. Second, we use point spread function (PSF) engineering and compressive sensing algorithms to fill in the missing information from the gaps surrounding the individual sensors in the array, further increasing the spatiotemporal throughput of the system by an additional >5.4x. The array of sensors is modeled as a single large-format "super-sensor," with erasures corresponding to the gaps between the individual sensors. The array is placed at the output of a (nearly) 4f imaging system, and we design a DOE for the Fourier plane that generates a distributed PSF that encodes information from the entire super-sensor area, including the gaps. We then computationally recover the large-scale image, assuming the object is sparse in some domain. Our calibration-free microscope can achieve ~3 μm resolution over >5.2 cm^2 FOVs at up to 120 fps, culminating in a total spatiotemporal throughput of 25.2 billion pixels per second. We demonstrate the versatility of our microscope in two different modes: structural imaging via darkfield contrast and functional fluorescence imaging of calcium dynamics across dozens of freely moving C. elegans simultaneously.

Introduction

Traditional microscopy is constrained by a "triple trade-off" between spatial resolution, field-of-view (FOV), and frame rate. Achieving high resolution over a large area typically necessitates mechanical scanning, which is inherently slow and unsuitable for capturing the dynamics of moving organisms. While multi-camera arrays offer a potential solution, they are often bulky and expensive, as each sensor typically requires its own dedicated optics. Furthermore, physical gaps between sensors in an array create "dead zones" where information is lost. This work introduces a computational imaging architecture that overcomes these limitations by treating a 48-sensor array as a single "super-sensor" and using diffractive optics to recover information from inter-sensor gaps.

System Architecture and Design

The hardware consists of a 6×8 array of monochrome sensors positioned at the output of a nearly 4f imaging system. The key innovation is the insertion of a custom-designed Diffractive Optical Element (DOE) into the pupil plane.

System architecture showing the DOE-based multiplexing onto a 48-sensor array.

Instead of a standard point-like Point Spread Function (PSF), the DOE is engineered to generate a "distributed" PSF consisting of 15 discrete spots. Consequently, every point in the sample is projected onto 15 different locations across the sensor array. This redundancy ensures "pan-visibility": even if the primary image of a feature falls into a physical gap between sensors, at least one of the other 14 copies is likely to land on an active sensor area.

Compressive Reconstruction and Gap Filling

The captured data is a multiplexed measurement where 15 shifted copies of the scene overlap on the sensor array. To recover the original scene, the authors solve an inverse problem using a fully shift-variant forward model that accounts for lens distortions. By employing L1-sparsity and Total Variation (TV) regularizers, the system can "demix" the signals and computationally "fill in the gaps." This approach effectively increases the FOV by >5.4x compared to the active area of the sensors alone.

A significant advantage of this system is its calibration-free design. Rather than requiring a physical scan of a point source, the distortion parameters are optimized directly from the captured data, enhancing robustness to minor hardware misalignments.

Experimental Validation and Biological Imaging

The system's performance was validated using USAF resolution targets, photolithography masks, and live biological samples. It achieves a record-breaking spatiotemporal throughput of 25.2 billion pixels per second, maintaining ~3 μm resolution over a FOV exceeding 5.2 cm².

Reconstruction results for darkfield and fluorescence configurations compared to raw data and low-resolution references.

The microscope was demonstrated in two modes:

  1. Darkfield Structural Imaging: Tracking the movement and "pharyngeal pumping" (feeding behavior) of dozens of wild-type C. elegans simultaneously.
  2. Fluorescence Functional Imaging: Monitoring calcium dynamics in GCaMP8f-expressing C. elegans. In this mode, the authors leveraged axial chromatic aberration to mitigate radial smearing caused by the DOE, improving the signal-to-noise ratio.

Discussion and Limitations

While the system offers unprecedented throughput, it relies on the assumption of sample sparsity, which is common in darkfield and fluorescence microscopy but may limit performance for dense, high-contrast samples. Current resolution at the FOV edges is limited by objective and tube lens aberrations. Future work may involve incorporating depth encoding into the DOE for 3D imaging or utilizing hybrid diffractive-refractive lenses to further correct for aberrations and spectral blurring.

Summary updated

Abstract

Microscopes face a trade-off between spatial resolution, field-of-view, and frame rate -- improving one of these properties typically requires sacrificing the others, due to the limited spatiotemporal throughput of the sensor. To overcome this, we propose a new microscope that achieves snapshot gigapixel-scale imaging with a sensor array and a diffractive optical element (DOE). We improve the spatiotemporal throughput in two ways. First, we capture data with an array of 48 sensors resulting in 48x more pixels than a single sensor. Second, we use point spread function (PSF) engineering and compressive sensing algorithms to fill in the missing information from the gaps surrounding the individual sensors in the array, further increasing the spatiotemporal throughput of the system by an additional >5.4x. The array of sensors is modeled as a single large-format "super-sensor," with erasures corresponding to the gaps between the individual sensors. The array is placed at the output of a (nearly) 4f imaging system, and we design a DOE for the Fourier plane that generates a distributed PSF that encodes information from the entire super-sensor area, including the gaps. We then computationally recover the large-scale image, assuming the object is sparse in some domain. Our calibration-free microscope can achieve ~3 μm resolution over >5.2 cm^2 FOVs at up to 120 fps, culminating in a total spatiotemporal throughput of 25.2 billion pixels per second. We demonstrate the versatility of our microscope in two different modes: structural imaging via darkfield contrast and functional fluorescence imaging of calcium dynamics across dozens of freely moving C. elegans simultaneously.
BibTeX
@misc{zhou2025large,
  author = {Kevin C. Zhou and Chaoying Gu and Muneki Ikeda and Tina M. Hayward and Nicholas Antipa and Rajesh Menon and Roarke Horstmeyer and Saul Kato and Laura Waller},
  title = {{Large-scale compressive microscopy via diffractive multiplexing across a sensor array}},
  year = {2025},
  eprint = {2507.14437},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2507.14437},
  howpublished = {arXiv},
}

Loading related content...

Figure