Comparison of parameter-adapted segmentation methods for fluorescence micrographs.

Area of Science:

• Cell Biology
• Biomedical Imaging
• Image Analysis

Background:

• Fluorescence microscopy image interpretation is time-consuming and lacks reproducibility.
• Accurate cell segmentation is crucial for reliable automatic image analysis.
• Developing robust segmentation algorithms for diverse cell types is essential for biological research.

Purpose of the Study:

• To compare and evaluate image segmentation methods for fluorescence-stained cells.
• To identify robust segmentation schemes with minimal parameterization for various cell types.
• To enhance segmentation performance for cultured epithelial cells.

Main Methods:

• Investigated and compared four segmentation methods: maximum-intensity linking (MIL), improved MIL, watershed, and improved watershed with morphological reconstruction.
• Utilized three manually annotated datasets (261, 817, 1,333 cells) of HeLa or L929 cells.
• Evaluated segmentation accuracy and robustness across different cell types and staining patterns.

Main Results:

• Watershed-based methods demonstrated significantly superior segmentation performance compared to the MIL method.
• Improved watershed methods, particularly with morphological opening by reconstruction, enhanced segmentation accuracy.
• The findings are applicable to cells with markers exhibiting dotted surface patterns.

Conclusions:

• Watershed transform-based segmentation is a more robust and accurate approach for fluorescence microscopy cell analysis.
• Morphological reconstruction offers a valuable enhancement for segmenting challenging cell surface features.
• The study provides practical insights for improving automated cell image analysis in biological and biomedical applications.