OrganoidTracker: Efficient cell tracking using machine learning and manual error correction

Area of Science:

• * Developmental Biology
• * Cell Biology
• * Bioengineering

Background:

• * Time-lapse microscopy is crucial for studying cell division and differentiation in organoids.
• * Manual cell tracking in dense, dynamic organoid cultures is time-consuming and impractical.
• * Automated solutions are needed to enable single-cell level analysis of organoid growth and homeostasis.

Purpose of the Study:

• * To develop a semi-automated cell tracking tool for organoids.
• * To overcome challenges posed by high cell density and rapid cell movement.
• * To provide a faster alternative to manual cell tracking without compromising data quality.

Main Methods:

• * Nuclei detection using a machine learning approach based on convolutional neural networks.
• * Cell trajectory reconstruction via a min-cost flow solver linking detections across time points.
• * Implementation of a warning system to flag potential tracking errors for manual review.

Main Results:

• * The developed semi-automated tracker significantly accelerates cell tracking in organoids.
• * A warning system identifies potential errors like rapid volume/position changes and nucleus division/appearance/disappearance.
• * With optimized warnings, over 98% of detected nuclei positions require no manual analysis, yielding high-quality lineage trees.

Conclusions:

• * The semi-automated cell tracker offers a substantial speed improvement over manual methods.
• * The tool provides high-quality, single-cell level tracking data for organoid research.
• * This approach facilitates more efficient study of organoid growth, development, and homeostasis.