Abstract
The development of epidermis plays a central role in driving the morphogenesis of the Caenorhabditis elegans embryo. However, current research on epidermal morphogenesis focuses disproportionately on overt phenotypic abnormalities, potentially overlooking the crucial role of developmental timing. In this study, we developed a modular two-step deep learning-based image analysis pipeline. First, we used ResU-Net to extract completely developed embryos and suppress noise; second, ResNet was used to predict the corresponding embryonic stage. The predicted probabilities and their corresponding embryonic time points were subsequently utilized to construct a developmental timeline. Combining this pipeline with differential interference contrast time-lapse microscopy, we dynamically tracked the timeline of epidermal morphogenesis in RNAi-treated embryos (ajm-1, tes-1, leo-1) and mutant embryos (clk-1). By statistically comparing the duration of each embryonic stage, our approach enabled the detection of stage-specific developmental timing without relying on overt phenotypic abnormalities or fluorescent markers, successfully recapitulating and extending the known roles of these genes from a temporal perspective. Our work underscores the importance of incorporating developmental timing into morphogenetic analysis, offering a novel framework for revealing subtle developmental processes, deepening the understanding of morphogenetic dynamics, and bridging the methodological gap in C. elegans embryology.