Sustained 3D isotropic imaging of subcellular dynamics using adaptive VCD light-field microscopy 2.0

Abstract

Long-term and high-speed 3D imaging of live cells beyond diffraction limit remains a big challenge for current super-resolution microscopy implementations, owing to their severe phototoxicity and limited volumetric imaging rate. While emerging light-field microscopy (LFM) has mitigated this issue through rapid and mild 3D imaging of dynamic biological processes with single 2D snapshots, it suffers from a suboptimal spatial resolution close to diffraction limit, which greatly compromises its applications for live-cell imaging. Here, we propose a super-resolution light-field reconstruction strategy based on an optics-aware view-channel-depth (VCD 2.0) deep-learning strategy. Through a stepwise inversion of the physical degradation process during light-filed imaging, VCD 2.0 strongly pushes the light-field microscopy beyond diffraction limit, and achieves instant 3D reconstructions from raw 2D light-field snapshots with an ~14-fold resolution improvement. After being combined with a simple-and-robust light-field add-on design, VCD 2.0 further enables long-term 3D super-resolution imaging (over 30000 volumes) of various intracellular dynamics (FOV of ~220*220*10 micro m3) even on an ordinary 2D fluorescence microscope. Using this novel and readily-accessible approach, we successfully capture the fast morphology changes of diverse organelles, such as the extension and constriction in lysosome, at a near isotropic resolution of ~120 nm, and a high volume rate up to 50 Hz.

Kilohertz volumetric imaging of in-vivo dynamics using squeezed light field microscopy

“…Additionally, the literature presents several strategies to enhance SLIM’s performance, such as background rejection by hardware 17 and computation 52 , multi-focus optics for extended DoF 15,53 , and sparsity-based resolution enhancement 43,54 . Furthermore, ongoing advancements in data-driven reconstruction algorithms, particularly physics-embedded deep learning models 19,20,45,55 , hold great promise for addressing the ill-posed inverse problems associated with limited space-bandwidth and compressive detection in SLIM. These developments are expected to significantly enhance SLIM’s capabilities and broaden its utility across diverse imaging scenarios.…”