SAMCell: Generalized label-free biological cell segmentation with segment anything

Area of Science:

• Cell Biology
• Microscopy Imaging
• Computational Biology

Background:

• Assessing cell morphology, confluency, and growth is crucial for cell health analysis in microscopy.
• Manual inspection of cell images is laborious for high-throughput applications.
• Automated cell segmentation methods often require expertise and annotated datasets.

Purpose of the Study:

• To develop an automated cell segmentation technique for microscopy images.
• To reduce the technical expertise and labor required for cell analysis.
• To improve the efficiency and quality of cell segmentation in high-throughput studies.

Main Methods:

• Modified Meta's Segment Anything Model (SAM) into SAMCell.
• Trained SAMCell on a large-scale dataset of diverse microscopy images.
• Developed a user-friendly graphical user interface (UI) for the automated technique.

Main Results:

• SAMCell effectively performs cell segmentation on various microscopy images, including unseen cell types.
• The model demonstrates robustness across different microscopes and image acquisition conditions.
• The user-friendly UI significantly lowers the technical barrier for automated microscopy.

Conclusions:

• SAMCell provides higher quality automated cell segmentation compared to previous methods.
• The graphical user interface streamlines the process, reducing manual labor in cell culturing.
• This automated approach enhances the efficiency of biological research involving cell analysis.