Applying watershed algorithms to the segmentation of clustered nuclei

Area of Science:

• Cytology
• Image Analysis
• Biomedical Imaging

Background:

• Accurate segmentation of clustered nuclei is essential in fluorescence microscopy-based analytical cytology.
• Inadequate separation of objects can lead to increased analysis time and affect statistical validity.
• Existing segmentation algorithms often fail with specific sample types or single-strategy approaches.

Purpose of the Study:

• To develop and validate a robust algorithm for automatic segmentation of clustered nuclei.
• To provide a tool adaptable for gradient-based, domain-based, and mixed segmentation strategies.
• To improve the accuracy and reliability of nuclei cluster analysis in cytological preparations.

Main Methods:

• Implementation of a novel algorithm based on morphological watersheds.
• Testing the algorithm on microscopic nuclei clusters from peripheral blood and bone marrow preparations.
• Utilizing appropriate markers and transformations for diverse sample types.

Main Results:

• The morphological watershed algorithm demonstrated high performance in segmenting nuclei clusters.
• Nearly 90% of test clusters were correctly segmented across different sample types.
• The algorithm proved effective for both gradient- and shape-based segmentation approaches.

Conclusions:

• The developed morphological watershed algorithm offers a versatile and effective solution for segmenting clustered nuclei.
• This tool enhances the accuracy of analytical cytology by improving nuclei cluster segmentation.
• The algorithm's adaptability and high success rate validate its utility in biomedical imaging and cytology.