SC-Track: a robust cell-tracking algorithm for generating accurate single-cell lineages from diverse cell...

Area of Science:

• Cell biology
• Computational biology
• Image analysis

Background:

• Accurate single-cell analysis relies on precise cell segmentation and classification from microscopy.
• Deep learning, particularly Convolutional Neural Networks (CNNs), shows promise but often yields noisy results, hindering lineage construction.

Purpose of the Study:

• To develop a robust algorithm for accurate single-cell tracking and lineage construction from timelapse microscopy images.
• To address the challenge of noisy cell segmentations and classifications produced by current methods.

Main Methods:

• Developed Single Cell Track (SC-Track), a novel algorithm utilizing a hierarchical probabilistic cascade model.
• Incorporated biological insights into cell division and movement dynamics.
• Integrated a cell class correction function for multiclass segmentation time series.

Main Results:

• SC-Track demonstrated superior performance compared to existing cell trackers across various segmentation types.
• Achieved robust cell-tracking without parameter tuning, adaptable to diverse imaging conditions and cell appearances.
• The cell class correction function enhanced classification accuracy in complex datasets.

Conclusions:

• SC-Track provides a generalized and robust solution for accurate single-cell lineage construction.
• The algorithm effectively handles noisy outputs from CNNs, improving quantitative analysis in cell fate studies.
• SC-Track is a valuable tool for advancing single-cell dynamics research.