Novel Image Processing Technique Allows Immunofluorescence Signals To Resemble Traditional H&E Stain and Enable Multiple Biomarker Visualizations on a Single Slide
Mike Lazare, Kevin Kenny, Natalia Jun, Ainura Kyshtoobayeva, Denise Hollman, Dave Henderson, Alex Corwin, Tommy Ha, Connie Chow, Kenneth Bloom. GE Global Research, Niskayuna, NY; Clarient Diagnostics, a GE Healthcare Company, Aliso Viejo, CA
Background: The current protocol for examining a sample via immunohistochemistry involves using one section of tissue for a chromogenic stain to visualize a single biomarker of interest and another serial section for hematoxylin and eosin (H&E) staining to visualize the morphology of the tissue, identify tumor, and determine regions of interest (ROI). Some of the limitations of using two slides are that 1) tissue can be limited and cutting a second slide for an H&E may exhaust available tissue, precluding further chromogenic tests, and 2) since the stains are on separate slides it can be difficult to correlate fields of view of the two stains. Here, we describe a method to use immunofluorescence (IF) staining and autofluorescence (AF) for morphology determination and ROI selection which will allow the same slide to be used for target biomarker staining as well. We have termed this technique “molecular H&E” (mH&E).
Design: Slides were stained with DAPI, a cy3 conjugated pan-cytokeratin antibody, and a cy5 conjugated Her2 antibody. IF whole slide imaging was then performed using filters for DAPI, FITC, Cy3, and Cy5. Image processing techniques were used to blend the DAPI (nucleus), FITC (AF), and Cy3 (epithelium) images to closely resemble an image of a true H&E stain. A board-certified pathologist reviewed 21 breast cancer cases using mH&E images initially, followed by true H&E images of serial sections of the same case series. The pathologist assessed the images based on two main scoring categories: the ability to 1) identify normal tissue, stroma, ductal carcinoma in situ (DCIS), and invasive tumor, and 2) to assess tissue percentage of invasive tumor.
Results: The pathologist was able to identify the stroma and invasive components in all samples using either type of H&E. Of the 18 cases where normal tissue was observed in the H&E, 15 cases had normal tissue detectable via mH&E. Of the 13 cases where DCIS was observed in the H&E, 11 DCIS cases were detected via mH&E. There was consistently ≤ 20% difference between the two tumor percentage scores.
Conclusions: Variability of tissue between serial sections likely contributed to some observed discrepancies. Overall the pathologist was satisfied with the mH&E as compared to the traditional H&E. Importantly, Her2 was also stained in Cy5, which facilitated side-by-side display of mH&E images and Her2 to visualize these two distinct pieces of information on the same slide.
Wednesday, March 6, 2013 9:30 AM
Poster Session V # 267, Wednesday Morning