Super-resolution optical imaging is crucial to the study of cellular processes. Current super-resolution fluorescence microscopy is restricted by the need of special fluorophores or sophisticated optical systems, or long acquisition and computational times. In this work, we present a deep-learning-based super-resolution technique of confocal microscopy. We devise a two-channel attention network (TCAN), which takes advantage of both spatial representations and frequency contents to learn a more precise mapping from low-resolution images to high-resolution ones. This scheme is robust against changes in the pixel size and the imaging setup, enabling the optimal model to generalize to different fluorescence microscopy modalities unseen in the training set. Our algorithm is validated on diverse biological structures and dual-color confocal images of actin-microtubules, improving the resolution from ~ 230 nm to ~ 110 nm. Last but not least, we demonstrate live-cell super-resolution imaging by revealing the detailed structures and dynamic instability of microtubules.

Stimulated emission depletion (STED) microscopy is a typical laser-scanning super-resolution imaging technology, the emergence of which has opened a new research window for studying the dynamic processes of live biological samples on a nanometer scale. According to the characteristics of STED, a high depletion power is required to obtain a high resolution. However, a high laser power can induce severe phototoxicity and photobleaching, which limits the applications for live cell imaging, especially in two-color STED super-resolution imaging. Therefore, we developed a low-power two-color STED super-resolution microscope with a single supercontinuum white-light laser. Using this system, we achieved low-power two-color super-resolution imaging based on digital enhancement technology. Lateral resolutions of 109 and 78 nm were obtained for mitochondria and microtubules in live cells, respectively, with 0.8 mW depletion power. These results highlight the great potential of the novel digitally enhanced two-color STED microscopy for long-term dynamic imaging of live cells.

Although conventional fluorescence intensity imaging can be used to qualitatively study the drug toxicity of nanodrug carrier systems at the single-cell level, it has limitations for studying nanodrug transport across membranes. Fluorescence correlation spectroscopy (FCS) can provide quantitative information on nanodrug concentration and diffusion in a small area of the cell membrane; thus, it is an ideal tool for studying drug transport across the membrane. In this paper, the FCS method was used to measure the diffusion coefficients and concentrations of carbon dots (CDs), doxorubicin (DOX) and CDs-DOX composites in living cells (COS7 and U₂OS) for the first time. The drug concentration and diffusion coefficient in living cells determined by FCS measurements indicated that the CDs-DOX composite distinctively improved the transmembrane efficiency and rate of drug molecules, in accordance with the conclusions drawn from the fluorescence imaging results. Furthermore, the effects of pH values and ATP concentrations on drug transport across the membrane were also studied. Compared with free DOX under acidic conditions, the CDs-DOX complex has higher cellular uptake and better transmembrane efficacy in U₂OS cells. Additionally, high concentrations of ATP will cause negative changes in cell membrane permeability, which will hinder the transmembrane transport of CDs and DOX and delay the rapid diffusion of CDs-DOX. The results of this study show that the FCS method can be utilized as a powerful tool for studying the expansion and transport of nanodrugs in living cells, and might provide a new drug exploitation strategy for cancer treatment in vivo.

To noninvasively and precisely discriminate among the growth stages of multiple cultivars of navel oranges simultaneously, the fusion of the technologies of near-infrared (NIR) hyperspectral imaging (HSI) combined with machine vision (MV) and deep learning is employed. NIR reflectance spectra and hyperspectral and RGB images for 740 Gannan navel oranges of five cultivars are collected. Based on preprocessed spectra, optimally selected hyperspectral images, and registered RGB images, a dual-branch multi-modal feature fusion convolutional neural network (CNN) model is established. In this model, a spectral branch is designed to extract spectral features reflecting internal compositional variations, while the image branch is utilized to extract external color and texture features from the integration of hyperspectral and RGB images. Finally, growth stages are determined via the fusion of features. To validate the availability of the proposed method, various machine-learning and deep-learning models are compared for single-modal and multi-modal data. The results demonstrate that multi-modal feature fusion of HSI and MV combined with the constructed dual-branch CNN deep-learning model yields excellent growth stage discrimination in navel oranges, achieving an accuracy, recall rate, precision, F1 score, and kappa coefficient on the testing set are 95.95%, 96.66%, 96.76%, 96.69%, and 0.9481, respectively, providing a prominent way to precisely monitor the growth stages of fruits.

Bound states in the continuum (BIC) have recently attracted extensive research interest, owing to their exceptional capability for light confinement and high quality-factor (Q-factor) resonances. In this work, we design an all-dielectric terahertz metasurface based on the symmetry-breaking properties of silicon rectangular pillars. Numerical simulations demonstrate that breaking the in-plane inversion (C₂) symmetry transforms the BIC into a quasi-BIC (q-BIC), which features a high Q-factor and a narrow resonance linewidth. Through far-field multipole decomposition and near-field electromagnetic field distribution analysis, the excitation mechanism of the q-BIC mode is elucidated, revealing that the magnetic dipole (MD) contribution remains predominant over a certain range of asymmetry. Furthermore, the q-BIC resonance exhibits high sensitivity to the change in refractive index, achieving a theoretical sensitivity of 110 GHz/RIU even with an analyte thickness of only 15 μm. When MS2 and PRD1 bacteriophages are employed as model analytes with distinct effective refractive indices, the sensor produces a larger resonance frequency shift for MS2 than for PRD1, demonstrating its capability for theoretical dielectric discrimination based on refractive index differences. These findings provide a valuable reference for the development of all-dielectric terahertz metasurface platforms for label-free refractive index sensing and model virus differentiation.