Color deconvolution for the analysis of tissue microarrays

Area of Science:

• Digital pathology
• Computational pathology
• Biomedical imaging analysis

Background:

• Tissue microarrays (TMAs) are crucial for high-throughput analysis.
• Color deconvolution separates dye signals in immunohistochemically stained images.
• Observer scoring of TMAs can be time-consuming and subjective.

Purpose of the Study:

• To evaluate the efficiency of color deconvolution for TMA analysis.
• To compare automated TMA analysis using color deconvolution with manual observer scoring.

Main Methods:

• TMAs from 100 autopsies were stained for connective tissue growth factor.
• Regions of interest were defined using binary masks.
• Diaminobenzidine (DAB) and hematoxylin signals were deconvolved.
• DAB signal intensity was measured and compared to manual scores.

Main Results:

• Automated analysis of 1,683 cores required significant annotation time.
• A mean of 31.3 minutes was needed per TMA for evaluation.
• Observer scores strongly correlated with median DAB intensity (Kendall's = 0.71).

Conclusions:

• Color deconvolution provides an efficient method for TMA analysis.
• The automated approach demonstrates high correlation with manual observer scoring.
• This technique enhances the objectivity and throughput of digital pathology image analysis.