Please select Print from the file menu or right mouse click and choose print from the menu to print this page.

[147] Measurement of Her-2/neu Membrane Receptor Immunoreactivity by Fully Automated Digital Image Analysis

RE Jimenez, AS Basu, S Mahooti, AS Gholap, SH Barsky. The Ohio State University College of Medicine, Columbus, OH; BioImagene, Inc, Cupertino, CA

Background: Her-2 / neu staining produces colorimeteric differences in membrane staining which are interpreted by pathologists and expressed on an ordinal scale of relative staining from 0 3+. The Her-2 / neu status represents important prognostic / predictive information of human breast cancer but its subjective interpretation results in interobserver, intraobserver and fatigue variability. Semi-automated image analysis systems have been used with limited success to eliminate this variability but still require the selection of the regions of interest by a pathologist. In addition, because of a significantly high percentage of false negatives and false positives by the subjective immunocytochemical method of Her-2 / neu assessment, Her-2 / neu is also measured by fluorescence in situ hybridization (FISH), which is considered the gold standard.
Design: To address all of these limitations, we created fully automated and pathologist-independent algorithms to analyze pixels of virtual slides. These algorithms could successfully divide a virtual slide into a grid of hundreds of fields of view (FOV), analyze each field and identify breast cancer cells in a background of stroma, successfully calculate the epithelial percentage in each field, determine the FOV with the highest density of cancer cells and select those FOV for further study. Image acquisition utilized scanners capable of producing images with a resolution of 20 pixels /10. Selected FOV were processed by jpg conversion of scanned virtual slides, screened for quality, enhanced and processed.
Results: Membrane algorithms utilizing both colorimetric (RGB) as well as intensity (gray scale) determinations were used to quantitate Her-2 / neu immunoreactivity. The algorithm-based ordinal values for Her-2 / neu strongly correlated with the subjective measurements (Intraclass Correlation: 0.88; 95% Confidence Interval: 0.79, 0.94) yet exhibited no interobserver, intraobserver or fatigue variability. In addition the algorithms provided measurements of Her-2 / neu on an expanded continuous scale that subjective review could not assess. These algorithm continuous scale Her-2 / neu measurements more strongly correlated with Her-2 / neu FISH than the subjective measurements.
Conclusions: Because of all these properties, fully automated digital image analysis of Her-2 / neu immunoreactivity adds value as a means of measuring this important membrane biomarker of human breast cancer.
Category: Breast

Wednesday, March 28, 2007

Poster # 2, Wednesday Morning