CellSAM: A Foundation Model for Cell Segmentation.

Area of Science:

• Bioimage analysis
• Cellular imaging
• Deep learning

Background:

• Accurate cell segmentation is crucial for biological research.
• Existing deep learning models are often domain-specific and lack scalability.
• A universal cell segmentation model is needed to address these limitations.

Purpose of the Study:

• To develop CellSAM, a universal model for generalized cell segmentation.
• To enable accurate cell identification across diverse cellular imaging data.
• To provide a scalable and adaptable solution for bioimage analysis workflows.

Main Methods:

• Utilized the Segment Anything Model (SAM) with a novel prompt engineering approach.
• Developed CellFinder, an object detector for automatic cell detection and SAM prompting.
• Evaluated CellSAM's performance on mammalian cells, yeast, and bacteria across various imaging modalities.

Main Results:

• CellSAM achieved human-level performance in cell segmentation across diverse datasets.
• Demonstrated strong zero-shot generalization capabilities.
• Showcased improved performance with few-shot learning and applicability in various bioimage analysis tasks.

Conclusions:

• CellSAM offers a universal and scalable solution for cell segmentation.
• The model generalizes effectively across different cell types and imaging conditions.
• CellSAM represents a significant advancement in automated bioimage analysis.