Abstract
Cell division, migration, and collective cell behaviors, along with less frequent events such as cell fusion, are fundamental biological processes underlying physiological and pathological mechanisms. However, accurate cell tracking and lineage reconstruction from time-lapse microscopy remain challenging because of failure modes such as identity switching and division misassignment, arising from dynamic cell behavior, morphological heterogeneity, and intermittent segmentation errors. We introduce Tui tracker, a multigenerational tracking framework that integrates integer linear programming optimization with lineage-aware corrections for both cell division and fusion. Although fully automated, the Tui tracker includes an expert-correctable graphical interface, allowing users to optionally inspect and refine lineage trees. We validate the Tui tracker using synthetic datasets and real-world data, including the DIC-C2DH-HeLa and Fluo-N2DH-SIM+ datasets from the Cell Tracking Challenge and T98G glioblastoma cell electrotaxis, achieving tracking accuracies of 0.983, 0.999, and 0.999, respectively. Beyond standard tracking accuracy metrics, the Tui tracker demonstrates superior performance on biologically motivated lineage fidelity measures such as BIO (averaging complete track reconstruction, branching correctness, track continuity, and cell cycle accuracy) and OP(CLB) (integrating BIO with temporal association quality), thereby providing a more faithful reconstruction of cellular events. The framework accurately detects mitosis events and incorporates explicit modeling of fusion events, which are validated in controlled synthetic datasets due to the limited availability of annotated biological fusion data. By enabling robust, automated, and correctable analysis of complex cellular dynamics, the Tui tracker has broad applications in quantitative cell biology, cancer research, and high-content drug screening. The source code is available at https://github.com/hftsai/tui.