In vivio single-cell analysis of dynamic cell behaviors - Imaging plays an increasingly important role in studying development and complex tissue formation across molecular, cellular and tissue levels. Fluorescence 3D time-lapse imaging allows every cell in a tissue or entire organism to be imaged, tracked and measured over hours and days to analyze lineage differentiation and dynamic cell behaviors. Other modalities, e.g., electron and expansion microscopy reveal fine structural phenomena, while emerging genomics technologies, e.g., spatial transcriptomics, seek to merge with microscopy to further provide systematic molecular information. Effective tools of image analysis are crucial to extract, integrate and interpret the information. We propose to leverage the power of deep learning to develop image analysis tools for systematic in vivo single-cell analysis, and to leverage such analysis to study collective cell behaviors in tissue morphogenesis, with three Aims. First, we will develop deep learning methods for accurate cell tracking, which is the essential first step to trace cell lineages and measure dynamic cell behaviors. We aim to produce a tool that can deliver substantial cell lineages with hundreds to thousands of cells imaged over hours to days in a wide range of model organisms and organoid cultures. Second, we will automate landmark-based image registration, which is crucial for cross-modality data integration. We propose a generalizable approach that uses statistical templates and Neural Networks to address the unique challenge in developmental images, namely heterochrony of developmental processes that creates combinatorial configurations of landmarks and complex systematic co-variance. Third, we study a novel Planar Cell Polarity (PCP) scheme in C. elegans, which we discovered through cell tracking and deep learning of cell movement patterns. By dissecting the compound polarity scheme and context specific regulators, we aim to understand how the conserved core PCP pathway can coordinate with different polarity pathways in order to orchestrate diverse motile behaviors in diverse developmental processes. By integrating technology development and hypothesis driven research, our proposal will further our understanding of embryogenesis and complex tissue formation.