Colorimetric Detection of Prostate Cancer in Grid Regions Utilizing a Ratio of Racemase to High Molecular Weight Keratin (HMWK) and P63
Benjamin M Brasseur, Andrew D Johnson, Jonathan C Henrikson, Joseph S Koopmeiners, Karin B Daniels, Greg J Metzger, Stephen C Schmechel. University of Minnesota, Minneapolis, MN
Background: For prostate cancer(Pca) laboratory studies, correlating preoperative prostate magnetic resonance(MR) spectroscopy features with postoperative reconstructed whole organs(from pathology slides) and quantifying relative proportions of carcinoma, benign epithelium, stroma and gland spaces would allow weighting of each tissue component during statistical analysis. We developed SigMap software that aligns pathology whole slide images(WSI) and generates grid regions in which tiled protein expression data using immunohistochemistry(IHC) may be generated across tissues(PMID:22438942). We are investigating whether IHC stains for racemase and basal cell markers may facilitate quantifying relative amounts of carcinoma(racemase developed with Fast Red chromagen), benign epithelium(p63 and HMWK basal cell markers developed with diaminobenzidine), stroma(blue hematoxylin counterstain) and gland lumens(white) areas within grid regions.
Design: Ten PCa-containing blocks were randomly selected. One section was stained with H&E; an adjacent section was stained with racemase, HMWK, and p63(and hematoxylin counterstain) using an automated platform(Ventana). Slides were digitized(ScanScope XT, Aperio) and PCa areas were annotated by pathologists. SigMap software aligned the sections and generated virtual 0.5 x 0.5 mm grid squares across tissues; color deconvolution(Aperio) software was used to quantify IHC stains. Fractional percentages of red, brown, blue and white, as well as red:brown ratios were calculated.
Results: Of an average 1080 total grid regions per slide, on average 27% of the regions contained PCa(range=0.9-54%). Red-stained pixels comprised a higher percentage of PCa-annotated grids(23.5%, SD= 14.6) than in non-PCa grids(2.75%, SD=0.91, p=0.0015). Brown-stained pixels comprised a lower percentage in PCa grids(2.18%, SD=1.29) vs non-PCa grids(4.84%, SD=1.58%, p=0.00069). The average red:brown ratio was 21.6-fold higher in PCa grids vs non-PCa grids(p= 0.0002). There was less blue-stained area in PCa grids(57.4%, SD=13.2) vs non-PCa grids(68.2%, SD=2.6, p=0.031), indicating a higher proportion of stroma in benign areas. There was less white area(16.9%, SD=6.1) in PCa grids vs non-PCa grids(24.2%, SD=2.1, p=0.0042), indicating a higher proportion of luminal area in benign tissue.
Conclusions: Racemase and basal cell IHC stains within computer-generated grid regions may be useful to determine the fractional composition of carcinoma, benign epithelium, stroma, and gland lumens. Also, a ratio of racemase:basal cell markers may be useful in automatically annotating PCa.
Wednesday, March 6, 2013 9:30 AM
Poster Session V # 295, Wednesday Morning