Towards Automatic Protein Co-Expression Quantification in Immunohistochemical TMA Slides.

Area of Science:

• Computational pathology
• Biomedical image analysis
• Cancer research

Background:

• Immunohistochemical (IHC) analysis is crucial for cancer protein expression assessment.
• Large-scale image data necessitates computational pathology for integrative analysis.
• Studying protein co-expression, not just individual proteins, may yield deeper clinical and therapeutic insights.

Purpose of the Study:

• To develop a computational pipeline for spatial alignment and protein quantification in tissue microarray (TMA) slides.
• To explore protein co-expression patterns in gastric cancer (GC) using IHC.
• To compare automated quantification with pathologist assessments for improved diagnostic accuracy.

Main Methods:

• Construction of a modular, open-source image analysis pipeline for gigapixel slides.
• Spatial alignment of TMA slides, quality evaluation, and tumor region definition.
• Quantification of E-cadherin and CD44v6 protein expression before and after tumor segmentation in 142 GC cases.

Main Results:

• Automated quantification within defined tumor regions enhanced agreement with pathologist classifications.
• A co-expression map successfully identified tissue cores expressing both E-cadherin and CD44v6.
• The pipeline effectively integrated data for potential use in learning-based pathology approaches.

Conclusions:

• The developed computational pipeline facilitates exploration of protein co-expression in cancer tissue.
• Automated analysis within tumor regions improves concordance with expert pathologist evaluation.
• This framework supports data integration for advanced, learning-based computational pathology.