175 / 2024-06-14 18:58:18

Enhanced Intracellular Interaction Observation through Multi-Color Super-Resolution via Deep Learning

Multi-color super-resolution, Deep learning

In life science, microscopy is an essential tool to examine live cells. In the past decades, the ongoing research of intracellular organelles interaction lies on efficient means to specifically reveal multiple sub-diffraction structures, in another term, multi-color super-resolution microscopy.  

For intracellular interaction research, the limited spectral channels constrain the number of organelles, optical systems with multiple lasers and detectors increases system  complexity, potential cross-talk and imaging time. For high spatial resolution, super-resolution has been universally applied to multi-color imaging. However, its high phototoxicity hinders long term living cells observation. Further the scope of fluorescence dyes has been narrowed by spectral intervals and high requirements for super-resolution imaging. Meanwhile, despite of universal applications of deep neuron networks in spatial improvements like super-resolution and denoising, the potential for simultaneous multi-color super-resolution has not been intensively explored.

In order to boost intracellular interaction observation, here we propose a deep learning based method to enlarge the number of organelles and spatial resolution all at once, named decouple super-resolution microscopy(DSRM). We demonstrate its performance on both our synthetic and experimental dataset. DSRM has achieved decomposition and super-resolution ability up to 3 different organelles and 75.4nm, respectively.