Deep Temporal Sequence Classification and Mathematical Modeling for Cell Tracking in Dense 3D Microscopy Videos of...

Area of Science:

• * Computational Biology
• * Image Analysis
• * Microbiology

Background:

• * Automatic cell tracking in dense environments is challenging due to inaccurate correspondences and parent-offspring misidentification.
• * Existing methods struggle with crowded scenarios, limiting their application in complex biological systems like bacterial biofilms.

Purpose of the Study:

• * To introduce DenseTrack, a novel algorithm for accurate automatic cell tracking in dense environments.
• * To improve the identification of parent-offspring relationships and cell division events in crowded 3D time-lapse image sequences.

Main Methods:

• * Formulating cell tracking as a deep learning-based temporal sequence classification task.
• * Employing a constrained one-to-one matching optimization problem using classifier confidence scores.
• * Utilizing an eigendecomposition-based strategy for cell division detection based on cellular geometry.

Main Results:

• * DenseTrack effectively establishes correspondences between consecutive frames in crowded scenarios.
• * The algorithm accurately detects cell division events, improving parent-offspring relationship identification.
• * Demonstrated superior performance in both qualitative and quantitative evaluations on simulated and experimental fluorescence image sequences of bacterial biofilm development.

Conclusions:

• * DenseTrack offers a significant advancement in automatic cell tracking for dense biological environments.
• * The integration of deep learning and mathematical models provides robust solutions for complex tracking challenges.
• * The method shows promise for applications in studying bacterial biofilm development and other crowded cellular systems.