High-Fidelity Long-term Whole-embryo Lineage and Fate Reconstruction by Iterative Tracking with Error Correction

Mengfan Wang, Qinghua Zhang, Congchao Wang, Yunfeng Chi, Wei Zheng, Zeyu Mu, Xiangyu Cao, Weizhan Zhang, Boao Yang, Alexander F. Schier, Joaquín Navajas Acedo, Yinan Wan, Guoqiang YubioRxiv 2026 · DOI

TL;DR

This paper introduces ITEC, a novel, fully unsupervised method for reconstructing cell lineages and fate maps in whole embryos with unprecedented accuracy. By leveraging spatiotemporal information and an iterative error correction strategy, ITEC achieves over 99.7% accuracy on terabyte-scale datasets, enabling detailed analysis of embryonic development and gene expression dynamics.

What Problem Does This Paper Address?

Understanding how a single cell develops into a complex organism is a fundamental question in developmental biology. This requires precisely tracking every cell's lineage and fate throughout the entire embryonic development. However, analyzing the massive, terabyte-scale datasets generated by modern live imaging of whole embryos is a monumental task. Manual annotation is infeasible and prone to bias, while existing automated methods struggle with the inherent challenges of low signal-to-noise ratios, heterogeneous imaging conditions, and the sheer volume of data, leading to significant errors in lineage reconstruction. This paper aims to bridge this gap by providing a highly accurate, automated, and robust solution for long-term, whole-embryo cell tracking.

Methods

ITEC (Iterative Tracking with Error Correction) is a full-process, unsupervised pipeline. Its core innovation lies in an iterative error correction strategy. This strategy uses tracking results to refine cell detection and then uses the improved detections to further refine tracking, creating a stable feedback loop that progressively reduces errors.

Key components include:

  • Optional Pre-processing: Deconvolution, rigid registration, and motion flow estimation to enhance image quality.
  • Initial Cell Detection: Utilizes high-order image information and a min-cut algorithm for precise pixel-level instance segmentation.
  • Tracking: Formulates cell linkage as a global discrete optimization problem solved by a minimum-cost circulation framework, enabling efficient handling of millions of cells.
  • Iterative Error Correction: The heart of ITEC, this module corrects segmentation and linkage errors by leveraging spatiotemporal information. It includes modules for missing cell redetection and segmentation correction.
  • Post-processing: Includes tracklet association, division detection, and multi-view stitching to ensure lineage continuity and completeness.
  • Visualization and Curation: An interactive interface and integration with Mastodon facilitate user analysis and curation.

The method is entirely unsupervised, requiring only the adjustment of a few biologically meaningful parameters.

Datasets

ITEC was validated on four cross-species datasets:

  • FISH1 & FISH2: Zebrafish embryo datasets.
  • MOUSE: Mouse embryo dataset.
  • DRO: Drosophila embryo dataset.

These datasets were chosen for their diverse imaging characteristics (signal-to-noise ratio, temporal/spatial resolution, cell density) to test the robustness of the method. Ground truth was established through a combination of public annotations and manual curation, resulting in over 44,555 ground-truth edge linkages.

Key Insights and Ideas

  • Spatiotemporal Information is Key: The paper emphasizes that the inherent redundancy and continuity in long-term, 3D time-lapse data are crucial for correcting errors that are difficult to resolve from single frames or short time windows.
  • Iterative Refinement is Powerful: The core idea of ITEC is that by iteratively feeding tracking results back into the detection and linkage processes, errors can be systematically identified and corrected, leading to a stable and highly accurate final reconstruction.
  • Minimum-Cost Circulation for Scalability: The adoption of a minimum-cost circulation framework is critical for handling the massive scale of data (millions of cells, terabytes of information) efficiently.
  • Unsupervised Approach Reduces Bottlenecks: Eliminating the need for manual annotation significantly accelerates the analysis pipeline and makes it more accessible.

Key Results

  • High Accuracy: ITEC achieved an estimated accuracy of over 99.7% on a terabyte-scale zebrafish embryo dataset (18.5 million cells).
  • Superior Performance: On four benchmark datasets, ITEC outperformed three state-of-the-art methods (TGMM, Linajea, Ultrack) by a significant margin, reducing errors per ground-truth cell linkage by at least 51.6% compared to the second-ranked method.
  • Lineage Continuity: ITEC demonstrated superior performance in long-term lineage reconstruction, with 85.5% of lineages remaining accurate over 100 frames in one dataset, compared to 67.7% for the next best method.
  • Scalability: The method successfully reconstructed lineages from terabyte-scale data, handling over 18.5 million cells cumulatively.
  • Biological Insights: ITEC enabled the visualization of complex morphogenetic processes like somite boundary formation, cell mixing patterns, and revealed associations between cell movement and spatial transcriptomics.
  • Efficiency: While runtime can still be long for very large datasets, ITEC's use of advanced techniques and parallelization allows for processing terabyte-scale data in days, a significant improvement over manual annotation.

Limitations and Future Work

  • Runtime Optimization: For extremely large datasets, the runtime can still be considerable if not using distributed computing. The authors plan to improve operational efficiency by designing better parallel strategies and optimizing underlying modules.
  • Cell Type Generalization: The current method is primarily designed for nucleus-labeled cells. Expanding its versatility to segment and track other cell types with diverse morphologies or internal expression patterns is a future goal.
  • Integration with Other Technologies: Future work includes integrating ITEC with other biological technologies, such as membrane channel imaging for morphological dynamics and deeper integration with spatial transcriptomics to correlate developmental rhythms with gene expression.

Summary updated

Abstract

Abstract Reconstructing the complete cell lineages and fate maps of a living embryo has remained a central challenge in developmental biology. Here we introduce ITEC (Iterative Tracking with Error Correction), a fully unsupervised method that automatically reconstructs the lineage of every cell in the embryo with high fidelity. ITEC was validated with manually annotated lineages on four cross-species datasets including zebrafish, mouse, and Drosophila . We reconstructed the developmental lineages of a zebrafish embryo from terabyte-scale data of total 18.5 million cells with an estimated accuracy of over 99.7%. With ITEC, we revealed spatiotemporal dynamics of key morphogenetic processes such as somite boundary formation, demonstrated the various patterns of spatial sorting among adjacent organs or anatomical regions, and identified associations between cellular movement and spatial transcriptomics. ITEC provides a powerful platform for retrospective fate mapping and systematic exploration of developmental dynamics at the cellular scale and single-embryo level.
BibTeX
@misc{wang2026high,
  author = {Mengfan Wang and Qinghua Zhang and Congchao Wang and Yunfeng Chi and Wei Zheng and Zeyu Mu and Xiangyu Cao and Weizhan Zhang and Boao Yang and Alexander F. Schier and Joaquín Navajas Acedo and Yinan Wan and Guoqiang Yu},
  title = {{High-Fidelity Long-term Whole-embryo Lineage and Fate Reconstruction by Iterative Tracking with Error Correction}},
  year = {2026},
  doi = {10.64898/2026.03.12.711203},
  howpublished = {bioRxiv},
}
Figure