FogBank: a single cell segmentation across multiple cell lines and image modalities

Area of Science:

• Cell biology
• Image analysis
• Bioimaging

Background:

• Cell lines in research grow in confluent sheets, making individual cell separation crucial.
• Current segmentation methods struggle with noise and over-segmentation in microscopy images.
• Accurate cell separation is vital for counting, identification, and measurement of individual cells.

Purpose of the Study:

• To develop a novel, automated segmentation method for accurate cell separation in confluent cell sheets.
• To address the limitations of existing watershed-based methods in handling noisy microscopy images.

Main Methods:

• FogBank utilizes morphological watershed principles with histogram binning to reduce noise.
• A geodesic distance mask is employed to accurately detect individual cell shapes.
• The method is applied to phase contrast, bright field, fluorescence, and binary microscopy images.

Main Results:

• FogBank achieved a segmentation accuracy of approximately 0.75.
• The method outperformed all compared segmentation techniques on reference datasets.
• Visual verification confirmed accuracy across 14 cell lines and 3 imaging modalities.

Conclusions:

• FogBank provides highly accurate single-cell segmentation for confluent cell sheets.
• The method is versatile, applicable to diverse cell lines and imaging modalities.
• Open-source code with a GUI is available for user-friendly execution.