Split and Merge Watershed: a two-step method for cell segmentation in fluorescence microscopy images.

Area of Science:

• Cellular biology and imaging science.
• Microscopy and image analysis.

Background:

• Cellular heterogeneity is crucial for understanding cell behavior, but identifying individual cells in microscopy images is challenging due to cell clustering.
• Existing cell segmentation methods often suffer from over-segmentation, under-segmentation, or misidentification, limiting their utility in research.

Purpose of the Study:

• To develop an advanced cell segmentation method for fluorescence microscopy images that accurately identifies individual cells while minimizing segmentation artifacts.
• To provide a robust alternative to current methods that require labeled data or machine learning.

Main Methods:

• A novel two-step method combining Marker-Controlled Watershed (MC-Watershed) with a Split and Merge Watershed (SM-Watershed) approach.
• Preprocessing of input images followed by MC-Watershed for initial segmentation.
• A two-step SM-Watershed process to split cell clusters and merge over-segmented regions, utilizing cell size, convexity, and other features.

Main Results:

• The proposed method achieved high performance, with an average visual accuracy of 90% and an F-index exceeding 80%.
• Demonstrated a suitable tradeoff between over-segmentation and under-segmentation, preserving cell shape.
• Outperformed other cell separation techniques and does not require labeled data or machine learning.

Conclusions:

• The developed cell segmentation method offers a significant improvement for analyzing cell populations in fluorescence microscopy.
• Its ability to accurately segment touching cells without labeled data makes it valuable for various applications, including virus infection analysis, drug discovery, and morphometry.
• The method provides a robust and efficient solution for cell identification tasks in biological research.