Automated identification of stained cells in tissue sections using digital image analysis

Area of Science:

• Computational pathology
• Biomedical image analysis
• Immunohistochemistry

Background:

• Accurate quantification of stained cells in tissue sections is crucial for research and diagnostics.
• Manual cell counting and area measurement are time-consuming and prone to inter-observer variability.
• Existing automated methods often struggle with complex cellular structures and background noise.

Purpose of the Study:

• To develop and validate a novel automated image analysis system for precise differentiation of immunohistochemically stained cells from background.
• To improve the accuracy and efficiency of cell segmentation and separation in digital pathology images.

Main Methods:

• Global thresholding algorithms (Otsu's, Kittler's, Kurita's) were applied for initial cell segmentation.
• A novel refinement algorithm incorporating edge pixel erosion was developed to enhance segmentation accuracy.
• A new decomposition method using semantic and relational information was employed to separate overlapping cells.
• The system was evaluated on stained tissue sections, comparing automated results with manual tracings of cell areas and numbers.

Main Results:

• Otsu's and Kurita's algorithms, when combined with edge erosion, showed good agreement with manual measurements of macrophage areas (P = .07 for Otsu's).
• Kittler's algorithm was less successful, even with edge erosion.
• A significant correlation was observed between computed and manually determined cell numbers, with regression analysis supporting accuracy.

Conclusions:

• A combined approach of global thresholding and edge erosion effectively identifies immunohistochemically stained macrophages.
• The developed automated system provides reliable cell area measurements comparable to manual analysis.
• This automated system offers a promising tool for quantitative analysis in immunohistochemistry studies.