High-throughput space-time Fourier ptychography for motile microorganisms
Abstract
High-fidelity imaging of live microorganisms is essential for understanding dynamic biological processes but remains limited by motion blur and insufficient temporal resolution. We present a space-time Fourier ptychography (ST-FP) system optimized for high-throughput, time-resolved quantitative phase imaging of fast-evolving biological specimens. To overcome photon budget and timing constraints, we develop a custom shift-register-based LED panel that enables direct, flicker-free control of illumination patterns, achieving a ∼1100× increase in effective irradiance over normal LED panels. This allows for stable imaging at exposure times as short as 800 µs. On the computational side, we formulate an auto-differentiable complex-valued reconstruction framework with GPU acceleration for the object set solution. Our temporal regularization can improve consistency and robustness under complex motion with different illumination conditions. The computational pipeline achieves a ∼23× speedup, reducing reconstruction time for a 120 × 1200 × 1200 sequence from 13.7 to 0.59 hours (CPU vs. GPU). We first evaluate our optimized ST-FP through simulations of swirling motion patterns. Then, in a real experiment with translation motion condition, a ∼45× improvement in space-bandwidth-time product (SBP-T) over our prior ST-FP implementation is quantitatively demonstrated using a binary phase target. We further validate the system on grayscale phase samples. We finally test our system for live imaging of vinegar eels and brine shrimp with uncontrolled motion at up to 260 Hz, demonstrating its capability to push the dynamic upper limits of high-throughput Fourier ptychography. The motion-aware reconstruction further enables trajectory tracking and flow field analysis of dynamic biological behaviors.
Methodology and Validation
Overview of the high-performance ST-FP system. (a) By leveraging our fast and stable illumination hardware, ST-FP captures high-speed raw data and reconstructs high-fidelity amplitude, phase, and velocity fields. (b) Schematic of the ST-FP system, based on a shift-register-controlled LED panel for fast and stable illumination. (c) LED module design using an array of shift-register LED drivers. This configuration allows the system to simultaneously illuminate all LEDs based on preloaded data, which maximizes the effective brightness of each LED. (d) Comparison of LED control logics under multiplexed illumination. Conventional sequential refreshing with short exposures, causing serious noise. Sequential refreshing with long exposures (low fps) leads to motion blur and reduced temporal resolution. Our static shift-register control (high fps) enables high-contrast, high-SNR imaging without flicker. (e) ST-FP reconstruction framework: the optimization problem is split into object and flow subproblems, solved alternately using automatic differentiation and optical flow solvers. (f) GPU acceleration results: compared to the old implementation, our GPU-accelerated version dramatically reduces computational time for long-sequence reconstructions.
Visualization of LED scanning mode and pulse-width modulation (PWM)–based brightness control. (a) The designed illumination pattern, where each column contains LEDs set to different brightness levels. (b) Illustration of PWM signals with varying duty cycles: a lower duty cycle (top) results in shorter LED on-time, while a higher duty cycle (bottom) in- creases the on-time and thus the perceived brightness. (c) Eight representative frames captured using a short exposure time (10 μs ), demonstrating the scanning behavior of the LED panel. In each frame, only a single column is active due to the panel’s internal column-scanning scheme. Within each active column, LEDs with higher brightness settings are more frequently visible, consistent with their longer PWM duty cycles. This supports the presence of both column-wise scanning and PWM-based brightness modulation.
Comparison of ST-FP reconstructions under different frame rates with simulated swirl motion. (a) Amplitude reconstructions from ST-FP at low fps, high fps (ours), and ground truth for selected frames. Insets in the lower right corners of the amplitude panels show the corresponding raw data for low-fps and high-fps cases. (b) Phase reconstructions from ST-FP at low fps, high fps, and ground truth for selected frames. (c) Accuracy metrics, including SSIM and MSE for amplitude and phase reconstructions across frames, comparing low-fps and high-fps results.
Comparison of ST-FP under different frame rate conditions using a moving phase target. (a) Raw intensity image of USAF phase target from a single central LED under the static condition. (b) Zoomed-in region of interest (ROI) from (a). (c) Reference phase reconstruction from traditional FP. (d) Raw intensity images acquired at low fps. (e) Phase reconstructions from ST-FP using the low-fps dataset in (d). (f) Raw intensity images acquired at high fps. (g) Phase reconstructions from ST-FP using the high-fps dataset in (f).
Media Results
This video compares raw data, zoomed-in views, and amplitude and phase reconstructions from CS-FP and ST-FP under low-fps and high-fps acquisition of a freely swimming vinegar eel.
This video illustrates the motion comparison of low-fps CS-FP, low-fps ST-FP, high-fps CS-FP, and high-fps ST-FP (ours) reconstructions, showing amplitude reconstruction, estimated velocity fields, and trajectories of a freely swimming vinegar eel.
This video illustrates the motion analysis capability of ST-FP, showing amplitude reconstruction, estimated velocity fields, and trajectories of a stretching-type motion of a brine shrimp.
BibTeX
@article{STFP2025,
author = {Ming Sun and Kaizhang Kang and Yogeshwar Nath Mishra and Xinge Yang and Hadi Amata and Wolfgang Heidrich},
journal = {Opt. Express},
number = {19},
pages = {39438--39451},
publisher = {Optica Publishing Group},
title = {High-throughput space-time Fourier ptychography for motile microorganisms},
volume = {33},
month = {Sep},
year = {2025},
url = {https://opg.optica.org/oe/abstract.cfm?URI=oe-33-19-39438},
doi = {10.1364/OE.570438}
}