Revealing architectural order with quantitative label-free imaging and deep learning

Abstract

Methods for imaging architecture of biological systems are currently not as scalable as genomic, transcriptomic, or proteomic technologies, because they rely on labels instead of intrinsic signatures. Label-free visualization of diverse biological structures is feasible with phase and polarization of light. However, distinguishing structures from information-dense label-free images is challenging. Recent advances in deep learning can distinguish structures from label-free images based on their shape. Here, we report joint use of polarization resolved imaging, reconstruction of optical properties, and deep neural networks for scalable analysis of complex structures. We reconstruct quantitative three-dimensional density, anisotropy, and orientation from polarization-and depthresolved images. We report a computationally efficient variant of U-Net architecture to predict 3D fluorescent structure from its density, anisotropy, and orientation. We evaluate the performance of our models by predicting anisotropic F-actin and isotropic nuclei. We report label-free prediction of myelination in human brain tissue sections and demonstrate the model's ability to rescue inconsistent labeling. We anticipate that the proposed approach will enable quantitative analysis of architectural order across scales of organelles to tissues. Label-free Microscopy | Polarization | Phase | Computational Imaging | Deep Learning Guo, Yeh, Folkesson et al. | bioRχiv | November 18, 2019 | 1-26 Guo, Yeh, Folkesson et al. | Learning architectural order bioRχiv | 3

Interpretable deep learning of label-free live cell images uncovers functional hallmarks of highly-metastatic melanoma

“…However, the often cited weakness of these techniques is the lack of an intuitive explanation of which parts of the data are particularly meaningful in defining the extracted pattern. While in some applications, such as image segmentation, image restoration or mapping between imaging modalities, a well-validated outcome of a network has been satisfactory (Christiansen et al, 2018;Fang et al, 2019b;Guo et al, 2019;Hershko et al, 2019;Hollandi et al, 2019;LaChance and Cohen, 2020;Moen et al, 2019;Nehme et al, 2018;Ounkomol et al, 2018;Ouyang et al, 2018;Rivenson et al, 2019;Wang et al, 2019;Weigert et al, 2018;Wu et al, 2019), there is increasing mistrust in results produced by 'black-box' neural networks. Aside from increasing the confidence, the analysis of the properties -also referred to as 'mechanisms'of the pattern recognition process can potentially generate insight of a biological/physical phenomenon that escapes the analysis driven by human intuition.…”

DynaMorph: self-supervised learning of morphodynamic states of live cells

“…Additionally, transfection efficiency of primary human microglia is usually low, providing a further challenge for cell labeling. To image the natural behavior of microglia over long periods, we used quantitative label-free imaging with phase and polarization (QLIPP) (16). QLIPP measures physical properties of the specimen in terms of its optical phase and retardance.…”

“…The phase, or optical path length, reports the density of molecular assemblies in a cell. The retardance, or polarization-resolved optical path length, reports the density of anisotropic molecular assemblies such as cytoskeletal networks and lipid bilayers (16).…”

“…Recent work on analysis of morphological states of cells has relied on images of fixed cells labeled with a panel of fluorescent markers (1), live three-dimensional imaging of the membrane labeled with genetic markers (2), and phase contrast imaging of live cells (3)(4)(5)(6). The morphological states have been analyzed with low dimensional representations computed with geometric or biophysical models (3,(7)(8)(9)(10)(11), supervised learning of morphological labels (4,(12)(13)(14)(15)(16)(17), and, recently, self-supervised learning of latent representations of morphology (5,6). These analytical approaches have been inspired by the need for quantitative descriptions of specific, complex biological functions, such as motility of single cells (2,3,7,8,18), collective cell migration (9,11), cell cycle (4,12,13), spatial gene expression (17), and spatial protein expression (14,16).In addition, data-driven integration of the morphology and gene expression (13,17,(19)(20)(21)(22) is now enabling rapid analysis of functional roles of genes.…”

“…We acquired reproducible measurements of cellular architecture and dynamics of microglia under immunogenic perturbations ( Figure 1A) using quantitative label-free imaging with phase and polarization (QLIPP) (16). To identify morphodynamic states for human microglia, we developed DynaMorph, a deep-learning enabled framework that automatically learns the morphodynamic states and transitions among states from high-dimensional live cell imaging data.…”