Segmenting overlapping cell nuclei in digital histopathology images

Area of Science:

• Digital Pathology
• Computational Biology
• Medical Image Analysis

Background:

• Accurate cell nuclei quantification is crucial for pathological diagnosis.
• Segmenting severely clustered and overlapping nuclei presents a significant challenge in image analysis.

Purpose of the Study:

• To develop and validate a new automated approach for segmenting clustered overlapping cell nuclei.
• To improve the accuracy of nuclei segmentation in immunostained pathological images.

Main Methods:

• A combined global and local thresholding method was used to extract foreground regions.
• Seed markers were generated using morphological filtering and region growing for nuclei separation.
• Seeded watershed algorithm was applied for segmenting clustered nuclei.
• A post-processing step was implemented to eliminate false positive pixels from cytoplasm.

Main Results:

• The proposed method successfully segmented severely clustered and overlapping nuclei.
• Experimental results demonstrated superior performance compared to existing state-of-the-art nuclei segmentation methods.
• The approach was validated on manually labeled Tissue Microarray (TMA) and Whole Slide Images (WSI) of colorectal cancers stained for P53.

Conclusions:

• The developed approach offers a robust solution for automated nuclei segmentation in challenging pathological images.
• This method has the potential to enhance diagnostic capabilities in digital pathology.
• The technique shows significant promise for analyzing P53 biomarker expression in colorectal cancer.