Deep diffractive optics for high-fidelity microscopy with an extended depth-of-field network
High-resolution microscopy is fundamentally constrained by the trade-off between lateral resolution and depth of field (DOF), which limits single-shot imaging of thick or topographically complex specimens. To address these limitations, we propose the extended depth-of-field network (EDFNet), a deep diffractive optics framework for high-fidelity extended DOF fluorescence microscopy. EDFNet jointly optimizes a pupil-plane diffractive optical element and an nonlinear activation free network (NAFNet)-based digital decoder within a fully differentiable pipeline, while incorporating fabrication-aware constraints to ensure practical feasibility. In simulation, EDFNet delivers consistent all-in-focus reconstruction across a 16-fold DOF, with an average peak signal-to-noise ratio (PSNR) of 25.86 dB and an average structural similarity index measure (SSIM) of 0.7989, while significantly outperforming representative baselines. Experimentally, we demonstrate single-shot, high-fidelity imaging of calibration slides and biological specimens, including ∼30 μm-thick cellular samples, over ±20 μm axial defocus. This capability paves the way for rapid all-in-focus observation of 3D biological dynamics, facilitating high-throughput diagnostic imaging and noninvasive biomedical screening.
Authors
- Liangcai Cao (ORCID: https://orcid.org/0000-0002-8099-2948)
- Qiaofeng Tan (ORCID: https://orcid.org/0000-0003-3475-6394)
Institutions
- Research Institute of Precision Instruments (Russia) (RU)
- Tsinghua University (CN)
Publication Details
- Journal
- Advanced Photonics Nexus
- Published
- 2026-11-13
- DOI
- https://doi.org/10.1117/1.apn.6.1.016010
- Primary Topic
- Digital Holography and Microscopy
- Type
- article
- Field-Weighted Citation Impact
- 0.00