DeepSea is an efficient deep-learning model for single-cell segmentation and tracking in time-lapse microscopy

Area of Science:

• Cell Biology
• Microscopy Techniques
• Bioinformatics

Background:

• Time-lapse microscopy offers high temporal resolution for observing cellular dynamics and heterogeneity at the single-cell level.
• Automated cell segmentation and tracking are essential for analyzing large datasets from time-lapse microscopy.
• Current methods struggle with accurate segmentation and tracking, especially using non-toxic phase-contrast imaging.

Purpose of the Study:

• To develop a versatile and trainable deep-learning model for automated single-cell segmentation and tracking.
• To improve the precision of cell tracking in phase-contrast live microscopy image sequences.
• To demonstrate the model's utility in analyzing cell size regulation in embryonic stem cells.

Main Methods:

• Development of DeepSea, a deep-learning model for simultaneous segmentation and tracking.
• Training and validation of the model on phase-contrast live microscopy image sequences.
• Application of DeepSea to analyze cell size regulation in embryonic stem cells.

Main Results:

• DeepSea achieves higher precision in single-cell segmentation and tracking compared to existing models.
• The model demonstrates robustness in handling phase-contrast microscopy images.
• Successful analysis of cell size regulation dynamics using the DeepSea model.

Conclusions:

• DeepSea provides a significant advancement in automated single-cell analysis for time-lapse microscopy.
• The model enhances the capability to study cellular dynamics and heterogeneity using accessible imaging modalities.
• DeepSea is a valuable tool for cell biology research, particularly in areas like cell size regulation.