Compressed Blind Deconvolution and Denoising for Complementary Beam Subtraction Light-Sheet Fluorescence Microscopy

Deep learning enhanced light sheet fluorescence microscopy for in vivo 4D imaging of zebrafish heart beating

Time-resolved volumetric fluorescence imaging over an extended duration with high spatial/temporal resolution is a key driving force in biomedical research for investigating spatial-temporal dynamics at organism-level systems, yet it remains a major challenge due to the trade-off among imaging speed, light exposure, illumination power, and image quality. Here, we present a deep-learning enhanced light sheet fluorescence microscopy (LSFM) approach that addresses the restoration of rapid volumetric time-lapse imaging with less than 0.03% light exposure and 3.3% acquisition time compared to a typical standard acquisition. We demonstrate that the convolutional neural network (CNN)-transformer network developed here, namely U-net integrated transformer (UI-Trans), successfully achieves the mitigation of complex noise-scattering-coupled degradation and outperforms state-of-the-art deep learning networks, due to its capability of faithfully learning fine details while comprehending complex global features. With the fast generation of appropriate training data via flexible switching between confocal line-scanning LSFM (LS-LSFM) and conventional LSFM, this method achieves a three- to five-fold signal-to-noise ratio (SNR) improvement and ~1.8 times contrast improvement in ex vivo zebrafish heart imaging and long-term in vivo 4D (3D morphology + time) imaging of heartbeat dynamics at different developmental stages with ultra-economical acquisitions in terms of light dosage and acquisition time.

Parameter-free super-resolution structured illumination microscopy via a physics-enhanced neural network

With full-field imaging and high photon efficiency advantages, structured illumination microscopy (SIM) is one of the most potent super-resolution (SR) modalities in bioscience. Regarding SR reconstruction for SIM, spatial domain reconstruction (SDR) has been proven to be faster than traditional frequency domain reconstruction (FDR), facilitating real-time imaging of live cells. Nevertheless, SDR relies on high-precision parameter estimation for reconstruction, which tends to suffer from low signal-to-noise ratio (SNR) conditions and inevitably leads to artifacts that seriously affect the accuracy of SR reconstruction. In this Letter, a physics-enhanced neural network-based parameter-free SDR (PNNP-SDR) is proposed, which can achieve SR reconstruction directly in the spatial domain. As a result, the peak-SNR (PSNR) of PNNP-SDR is improved by about 4 dB compared to the cross-correlation (COR) SR reconstruction; meanwhile, the reconstruction speed of PNNP-SDR is even about five times faster than the fast approach based on principal component analysis (PCA). Given its capability of achieving parameter-free imaging, noise robustness, and high-fidelity and high-speed SR reconstruction over conventional SIM microscope hardware, the proposed PNNP-SDR is expected to be widely adopted in biomedical SR imaging scenarios.

Extended field of view of light-sheet fluorescence microscopy by scanning multiple focus-shifted Gaussian beam arrays

Light-sheet fluorescence microscopy (LSFM) facilitates high temporal-spatial resolution, low photobleaching and phototoxicity for long-term volumetric imaging. However, when a high axial resolution or optical sectioning capability is required, the field of view (FOV) is limited. Here, we propose to generate a large FOV of light-sheet by scanning multiple focus-shifted Gaussian beam arrays (MGBA) while keeping the high axial resolution. The positions of the beam waists of the multiple Gaussian beam arrays are shifted in both axial and lateral directions in an optimized arranged pattern, and then scanned along the direction perpendicular to the propagation axis to form an extended FOV of light-sheet. Complementary beam subtraction method is also adopted to further improve axial resolution. Compared with the single Gaussian light-sheet method, the proposed method extends the FOV from 12 μm to 200 μm while sustaining the axial resolution of 0.73 μm. Both numerical simulation and experiment on samples are performed to verify the effectiveness of the method.

Minutes-timescale 3D isotropic imaging of entire organs at subcellular resolution by content-aware compressed-sensing light-sheet microscopy

Rapid 3D imaging of entire organs and organisms at cellular resolution is a recurring challenge in life science. Here we report on a computational light-sheet microscopy able to achieve minute-timescale high-resolution mapping of entire macro-scale organs. Through combining a dual-side confocally-scanned Bessel light-sheet illumination which provides thinner-and-wider optical sectioning of deep tissues, with a content-aware compressed sensing (CACS) computation pipeline which further improves the contrast and resolution based on a single acquisition, our approach yields 3D images with high, isotropic spatial resolution and rapid acquisition over two-order-of-magnitude faster than conventional 3D microscopy implementations. We demonstrate the imaging of whole brain (~400 mm3), entire gastrocnemius and tibialis muscles (~200 mm3) of mouse at ultra-high throughput of 5~10 min per sample and post-improved subcellular resolution of ~ 1.5 μm (0.5-μm iso-voxel size). Various system-level cellular analyses, such as mapping cell populations at different brain sub-regions, tracing long-distance projection neurons over the entire brain, and calculating neuromuscular junction occupancy across whole muscle, are also readily accomplished by our method.