Label-free Cell Tracking and Dynamic Rearrangements in Epithelial Monolayers

Particle image velocimetry (PIV) has been successfully adapted from fluid dynamics to investigate collective cell motion, due to its strength in analyzing label-free images in particular phase-contrast images. In PIV-based analysis, the motion of any features is measured, without distinction of the two most notable features in phase-contrast images, cell boundaries and nuclei. Here we describe the use of deep learning to identify cell nuclei based on a UNet convolutional neural network for nucleus segmentation. This enables nuclear tracking and analysis of the individual and collective motion of cell groups. We will compare both individual and collective motion for sheets of epithelial cells of varying metastatic potential as an example of the behaviors that can be captured by our techniques.