CELLECT: contrastive embedding learning for large-scale efficient cell tracking.

Area of Science:

• Cellular biology
• Bioinformatics
• Computational neuroscience

Background:

• Quantitative analysis of large-scale cellular behaviors is vital for understanding physiopathological processes.
• Efficient and high-performance cell tracking remains a significant challenge in practical applications.
• Existing methods often lack the generalization capabilities required for diverse biological systems.

Purpose of the Study:

• To introduce CELLECT, a novel contrastive embedding learning method for large-scale and efficient cell tracking.
• To demonstrate the broad applicability and generalization of CELLECT across different imaging modalities and species.
• To validate CELLECT's performance on challenging datasets, including the Cell Tracking Challenge.

Main Methods:

• Development of CELLECT, a contrastive embedding learning framework for cellular structures.
• Pretraining CELLECT models on public datasets for enhanced generalization.
• Application of CELLECT using advanced two-photon imaging for real-time 3D cell tracking and signal extraction.

Main Results:

• CELLECT demonstrates broad generalization, performing effectively across different imaging modalities and species after pretraining on a single dataset.
• Real-time 3D tracking of large-scale B cells during germinal center formation in mouse lymph nodes was achieved.
• Quantitative identification of cell-bacterium interactions in the mouse spleen and high-fidelity neural signal extraction during nonrigid motions were successfully performed.

Conclusions:

• CELLECT provides a high-performance and efficient solution for large-scale cell tracking.
• The method exhibits significant potential for diverse applications in immunology, pathology, and neuroscience.
• CELLECT's ability to generalize across datasets and imaging types represents a significant advancement in cell tracking technology.