Research

Our lab develops computational imaging systems by jointly designing optics, devices, signal processing, and algorithms. We combine optical engineering, physics, and computation to expand what imaging systems can measure. Our work spans biomedical imaging, neuroscience, and semiconductor metrology and inspection.

Computational Fluorescence & Neural Imaging

We develop computational fluorescence microscopes for large-scale, three-dimensional imaging and fast measurements of biological and neural dynamics. Related publications are here.

Computational Miniature Mesoscope (CM²)

We develop compact computational fluorescence microscopes that combine distributed optics with model-based and learning-based reconstruction for single-shot 3D imaging across large fields of view. These platforms support studies of biological structure and dynamics at cellular scales.

• • • • • • • • • Xue et al. Single-Shot 3D Widefield Fluorescence Imaging with a Computational Miniature Mesoscope. Science Advances 6(43), eabb7508 (2020).
                • Xue et al. Deep-learning-augmented computational miniature mesoscope. Optica 9, 1009–1021 (2022).
                • Yang et al. Wide-field, high-resolution reconstruction in computational multi-aperture miniscope using a Fourier neural network. Optica 11, 860–871 (2024).

Event-Driven Dynamic Microscopy

We combine event-based sensors, tailored illumination, and computational reconstruction to capture fast biological dynamics. Our approaches include volumetric light-field imaging and dual-channel fluorescence microscopy.

• • • • • • • • • Guo et al. Dual-channel event microscopy for ultrafast biological imaging. PNAS (2026).
                • Guo et al. EventLFM: Event Camera integrated Fourier Light Field Microscopy for Ultrafast 3D imaging. Light: Science & Applications (2024).

Computational Metrology & Inspection

We develop computational imaging methods for surface measurement, three-dimensional material characterization, and semiconductor metrology and inspection. Related publications are here.

Substrate-Enhanced Diffraction Tomography

We develop reflection-mode tomography methods that use substrate-enhanced scattering and multiple-scattering models to recover three-dimensional refractive-index structure from intensity measurements. 

• • • • • • • • • Li et al. Transfer-function approach to substrate-enhanced diffraction tomography. Optica (2026).
                • Zhu et al. Reflection-mode Multi-slice Fourier Ptychographic Tomography. IEEE Transactions on Computational Imaging (2026).
                • Li et al. Reflection-mode diffraction tomography of multiple-scattering samples on a reflective substrate from intensity images. Optica (2025).

Fourier Ptychographic Topography

We develop Fourier ptychographic topography (FPT) to recover surface height from reflection-mode measurements over wide fields of view. Single- and dual-wavelength approaches combine high lateral resolution with sensitive height measurements.

• • • • • • • • • Shen et al. Dual-wavelength Fourier Ptychographic Topography. IEEE Transactions on Computational Imaging (2026).
                • Wang et al. Fourier ptychographic topography. Optics Express 31, 11007–11018 (2023).

Physics-based AI

We combine physical imaging models with neural networks, simulator-based learning, and structured priors to solve inverse problems. Explore our Physics-based AI publications and related work on imaging through scattering media.

Diffusion models

We develop structured diffusion priors for image reconstruction, combining learned image statistics with measurement models.

• • • • • • • • • Alido et al. Whitened Score Diffusion: A Structured Prior for Imaging Inverse Problems. NeurIPS (2025).

Neural fields

We use neural fields to represent continuous object structure and recover quantitative phase and three-dimensional refractive-index maps from optical measurements.

• • • • • • • • • Wang et al. NeuPh: scalable and generalizable neural phase retrieval with local conditional neural fields. Advanced Photonics Nexus 3, 056005 (2024).
                • Liu et al. Recovery of Continuous 3D Refractive Index Maps from Discrete Intensity-Only Measurements using Neural Fields. Nature Machine Intelligence 4, 781–791 (2022).

Reliable Deep Learning with Uncertainty Quantification

We develop uncertainty-aware learning methods that assess prediction confidence and improve the reliability of quantitative phase imaging and biomedical measurements.

• • • • • • • • • Xue et al. Reliable deep learning-based phase imaging with uncertainty quantification. Optica 6, 618–629 (2019).
                • Liu et al. Deep spectral learning for label-free optical imaging oximetry with uncertainty quantification. Light: Science & Applications 8, 102 (2019).

AI for Biomedicine

We develop AI methods for biomedical image reconstruction, denoising, digital staining, and quantitative analysis. Related publications are here.

Physics-informed deep learning

We integrate physical models with neural networks to estimate biological quantities from optical measurements, including vascular and tissue dynamics measured with laser speckle contrast imaging.

• • • • • • • • • Li et al. Physics-informed neural network for mapping vascular and tissue dynamics using laser speckle contrast imaging. Neurophotonics 13, 035012 (2026).

Deep Learning for In Vivo Neural Signal Modeling & Extraction

We develop self-supervised denoising methods to recover neural signals from low-photon fluorescence measurements and support high-speed voltage imaging. Explore related neural-imaging publications.

• • • • • • • • • Liu et al. DeepVID v2: self-supervised denoising with decoupled spatiotemporal enhancement for low-photon voltage imaging. Neurophotonics 11(4), 045007 (2024).
                • Platisa et al. High-speed low-light in vivo two-photon voltage imaging of large neuronal populations. Nature Methods 20, 1095–1103 (2023).

Digital Staining & Cross-Modality Learning

We learn relationships between imaging modalities to predict molecular and structural information from label-free measurements. Applications include multiplexed single-cell cytometry, digital staining, and computational tissue assessment.

• • • • • • • • • Cheng et al. Single-cell cytometry via multiplexed fluorescence prediction by label-free reflectance microscopy. Science Advances 7(3), eabe0431 (2021).

Quantitative Phase Imaging & Tomography

We develop label-free methods to recover optical phase, refractive index, and three-dimensional sample structure. Explore our phase-imaging and tomography publications.

Intensity Diffraction Tomography

We combine programmable illumination, scattering models, and reconstruction algorithms to recover three-dimensional refractive-index structure from intensity-only measurements.

• • • • • • • • • Li et al. High-speed in vitro intensity diffraction tomography. Advanced Photonics 1(6), 066004 (2019).

Fourier Ptychography 

We combine illumination diversity with computational reconstruction to achieve high-resolution phase imaging across wide fields of view.

• • • • • • • • • Tian et al. Computational illumination for high-speed in vitro Fourier ptychographic microscopy. Optica 2(10), 904–911 (2015).