Quantification of immunohistochemical staining by color translation and automated thresholding.

OBJECTIVE: To develop a method of quantitating immunohistochemical staining using diaminobenzidine (DAB) and hematoxylin by means of color image analysis. STUDY
DESIGN: Red-green-blue (RGB)-filtered grayscale values from images from microscopic slides of mouse jejunum, stained with DAB, hematoxylin or both were analyzed using the public domain program NIH Image. Based on the correlations between the R-G- and B-filtered grayscale values, a simple translation algorithm using the RGB information was developed, providing the option for separation of DAB only- and double-stained areas from hematoxylin only-stained areas by means of automated thresholding. The method was tested by staining mouse jejunum for the growth factors EGF, TGF-alpha and TGF-beta 1-3 using immunohistochemical techniques.
RESULTS: A good separation of DAB- and double-stained pixels from hematoxylin-stained pixels was achieved, with misclassification of only 2.4% of the pixels as compared to 34% misclassification using automated thresholding of the blue component of the RGB image, the untransformed grayscale images with the most contrast for DAB- and non-DAB-stained areas. Significant differences in relative areas stained and mean specific optical density for the growth factors in mouse jejunal crypts and villi were observed.
CONCLUSION: The image analysis method described offers the possibility of objective determination of stained area in histologic slides with the commonly used DAB and hematoxylin chromophores. It shows that reproducible and objective measurements can be made based on RGB true color images acquired using a simple microscope and video camera setup and the public domain program NIH Image.