Label-free multiplexed microtomography of endogenous subcellular dynamics using generalizable deep learning

Abstract

Simultaneous imaging of various facets of intact biological systems across multiple spatiotemporal scales would be an invaluable tool in biomedicine. However, conventional imaging modalities have stark tradeoffs precluding the fulfilment of all functional requirements.Here we propose the refractive index (RI), an intrinsic quantity governing light-matter interaction, as a means for such measurement. We show that major endogenous subcellular structures, which are conventionally accessed via exogenous fluorescence labeling, are encoded in 3D RI tomograms. We decode this information in a data-driven manner, thereby achieving multiplexed microtomography. This approach inherits the advantages of both highspecificity fluorescence imaging and label-free RI imaging. The performance, reliability, and scalability of this technology have been extensively characterized, and its application within single-cell profiling at unprecedented scales has been demonstrated..

Nanoscale resolution imaging of the whole mouse embryos and larval zebrafish using expansion microscopy

“…The use of machine learning to identify specific structures from fluorophore NHS ester images of expanded specimens would be the subject of further study. Recently, machine learning has been applied to data from label-free optical microscopy imaging, 86,87 vibrational imaging of expanded specimens, 88 and EM imaging [89][90][91][92] for cellular structure segmentation and reconstruction. In applying ML to image data, the most critical part is acquiring a large amount of labeled data, which can be used as ground truth training.…”

Quantitative phase velocimetry measures bulk intracellular transport of cell mass during the cell cycle

“…allowing imaging of more than one cell, or cells with larger spread area) and increase computational cost accordingly. QPV can also potentially be extended to use with 3D QPI data available from 3D tomographic QPI imaging 45,46 . This would come at the cost of increased computation time.…”

“…QPV can also potentially be extended to use with 3D QPI data available from 3D tomographic QPI imaging 45,46 . This would come at the cost of increased computation time.…”

Molecularly Specific and Functional Live Cell Imaging by Label-Free Interference Microscopy

“…Data scientists have worked with the optical microscopists to elevate the molecular specificity of label-free phase-sensitive imaging. This approach relies on a hypothesis that there is a (nontrivial) relationship between the refractive index of the sample and the fluorescence images of molecules of interest. , By acquiring paired data sets of the label-free and fluorescence images, a computational model that connects the two imaging modalities can be established through machine learning. Once established, this model is then used to predict the fluorescence images of a sample based on its label-free image (Figure a).…”

“…Remarkably, the accuracy of digital staining is enhanced by acquiring label-free phase images with rich subcellular structural information by a sensitive interference reflection microscopy . The general connections between the refractive index and fluorescence intensity of several subcellular structures (actins, mitochondria, lipid droplets, plasma membranes, nuclei, and nucleoli) are closely examined with a high-resolution, three-dimensional refractive index tomographic mapping . Importantly, it was shown that the cell organelles generally exhibit local refractive index contrast for identification, elucidating the underlying principles of predicting the fluorescence images of cell organelles with the label-free phase images.…”