Automated Replicative Lifespan Determination in Diploid Yeast Using a Microfluidic Chip and ResNet-Based Image Analysis

This chapter describes an integrated system combining a high-throughput microfluidic diploid yeast long-term culturing (DYLC) chip with a deep-learning-based computational pipeline for yeast replicative aging studies. The microfluidic system features 1100 "leaky bowl"-shaped traps to accommodate diploid cells with a bipolar budding pattern. By leveraging hydrodynamic forces to induce mother-cell rotation, the traps reorient buds to facilitate concerted daughter dissection, thereby ensuring reliable lifelong tracking of individual mother cells. A dedicated algorithm, combining computer vision and residual neural network (ResNet), is presented to efficiently process massive datasets of micrographs. Pivotal steps encompass image segmentation, multi-feature annotation, network training, and subsequent data conversion, merging, and correction, ensuring accurate recognition of dynamic cell-budding events. This integrated strategy enables the bias-free, high-throughput determination of replicative lifespan (RLS) and budding time interval (BTI), serving as a standardized toolkit for eukaryotic aging research.