Color deconvolution for the anaylsis of tissue microarrays
Marc K. Halushka None, MD, DSc, DVM, CRNP, MA, BS, AOCN, ACSW, ART, AS, ASN, BA, JD, BDS, BVSc, CCRC, CMA, CTR, DMS, DrPH, FACP, FACS, FRCP, DDS, FRCS, LDS, LPN, LSW, MBA, AAS, AND, ATC, LDN, CS, FACC, ASCP, DABR, DMD, DO, BSN; The Johns Hopkins Medical Institutions; Toby Cornish MD; The Johns Hopkins Medical Institutions;
Content:
Tissue microarrays (TMAs) are a powerful tool for the analysis of multiple tissue specimens. TMAs conserve tissue, facilitate uniform experimental conditions, and, once constructed, allow high-throughput in slide preparation and analysis. Most TMAs are stained with chromogenic dyes, and are often analyzed manually using traditional grading systems. Increasingly, digital imaging is used to analyze the staining in TMA-based experiments. In the case of chromogenic immunohistochemistry, a number of analysis methods are available. One of these, color deconvolution, effectively separates a single color digital image into multiple grayscale images that represent the individual contribution of each dye. Here, we use widely-available software tools to automate the color deconvolution and analysis of TMA image datasets.
Technology:
The markup and analysis software was implemented in the macro language of Java-based ImageJ image processing package (Wayne Rasband, NIH). The analysis macro uses a color deconvolution plugin (Gabriel Landini, University of Birmingham, England).
Design:
TMAs were constructed from 100 adult autopsies and included tissue from renal cortex and mesenteric arteries. Immunohistochemistry for tissue inhibitor of metalloproteinase 3, a protein known to be highly expressed in renal tissue, was performed on the TMAs using an Anti-TIMP3 antibody. A methylene blue counterstain was used. The TMAs were imaged using the Chromavision ACIS II system (ChromaVision Medical Systems, San Juan Capistrano, California), and images of individual spots were isolated for analysis. The markup macro was run on the image dataset, allowing the user to create three binary masks: tissue area, inclusion area, and exclusion area. The masks were then combined to create an analysis region of interest. The analysis macro then performed color deconvolution on each image, separating the brown signal and the blue counterstain. The brown signal was measured for each region of interest.
Results:
Analysis of TIMP3 staining reveals a significantly greater intensity of staining in the renal cortex arterioles when compared to the mesenteric artery. Results from the color deconvolution analysis are compared to other TMA color image analysis methods, including hue-saturation-value color space anaylsis and manual scoring.
Conclusion:
Color deconvolution offers an alternative to other methods of analyzing color TMA images.
