Deep learning-powered virtual fluorescent labeling of fibers in Fibrin hydrogels

The intricate fibrillar composition of the extracellular matrix (ECM) establishes viscoelastic and non-linear elastic properties, which are essential factors in modulating intracellular signaling pathways and providing structural support for cells. Commonly used amorphous synthetic scaffolds lack a natural ECM's fibrillar properties, whereas naturally-derived biomaterials such as fibrin and collagen can mimic native fibrillar structure, propagate forces and mechanical tension, and thus influence cell migration during processes such as tissue morphogenesis. Longitudinal confocal imaging of these fibers often relies on fluorescent labeling, where fluorescent dyes may impact cell viability and state, and tend to photobleach early in an experiment. An alternative technique is reflection confocal microscopy (RCM) that provides high resolution images of fibers. However, RCM is sensitive to fiber orientation along the optical axis and consequently fibers not parallel to the image plane are not detected. We successfully detected these fibers without the use of labels by employing a deep-learning model that can virtually stain fibers.