Use of Computer Assisted Analysis To Facilitate Tissue Microarray Interpretation
L Goodell, W Chen, P Javidian, M Chekmareva, J Hu, DJ Foran. UMDNJ-Robert Wood Johnson Medical School, New Brunswick, NJ
Background: Tissue microarrays (TMA) have surged in popularity as a tool to conserve human tissue in translational research. Interpretation of TMAs, often with hundreds of tissue cores, is tedious. Our group has developed algorithms and conducted man-machine comparison studies to evaluate the use of computer assisted analysis to facilitate standardization, semi-quantitative interpretation and management of biomarker data using immunohistochemistry (IHC).
Design: Man-machine concordance studies were designed to evaluate IHC staining intensity on 5 breast cancer TMA slides stained with CK18, Her2, Cyclin D1, CK19 and ER digitized using a Trestle MedMicroscopy system at 40x equivalent resolution. Automatic quantification algorithms analyze the color profiles of each specimen, decompose the histospots into constituent staining maps according to the Principle Color Vectors, and generate Effective Staining Intensity (ESI) scores. Using our TMR software, a pathologist first selected representative histospots of different semi-quantitative staining levels for standardization. Three pathologists then used the TMR software [Figure 1] to evaluate each tissue disc by selecting a staining level or rejecting the image based upon the tissue quality and image fidelity using a standard set of controlled lighting/viewing conditions. The ESI scores were mapped back to the same evaluation categories used by the pathologists through linear regression.
Results: Concordance between pathologists and computer scores were measured from 1407 tissue cores using squared-weighted Cohan's kappa to account for the ordinal nature of the grading scale (Table 1).