Use of automated image analysis in evaluation of Mesothelioma Tissue Microarray (TMA) from National Mesothelioma Virtual Bank

The National Mesothelioma Virtual Bank (NMVB) was established to provide annotated biospecimens to the mesothelioma research community. The resource provides tissue microarrays (TMA) to evaluate the biomarkers along with a variety of other resected mesothelioma specimens. In this manuscript, we describe the immunohistochemical evaluation of the mesothelioma TMA with three key antibodies that are used in making the diagnosis of mesothelioma, and compared the immunohistochemical assessment between manual scoring and image analysis. The TMA was assessed for the immunohistochemical expression of calretinin (N = 39), cytokeratin (CK) 5/6 (N = 33), and D2-40 (N = 37). Immunohistochemistry was evaluated by semi-quantitative (manual) scoring using light microscope (MS) and by automated image analysis (AS). Calretinin staining was seen in both cytoplasmic and nuclear locations. CK5/6 stain was localized to the cytoplasm. D2-40 stain showed only membranous expression in our cases. • Based on the pathologist's scores, calretinin was positive in 31 of the 39 cases (80%), CK 5/6 in 15 of the 33 cases (46%) and D2-40 in 18 of the 37 cases (49%). • The percent-positive agreement between manual scores and image analysis was 90% (35/39), 94% (31/33), and 95% (35/37) for calretinin, CK 5/6, and D2-40, respectively. There was a substantial agreement between manual and automated scores for calretinin (kappa = 0.614) and an almost perfect agreement for CK5/6 (kappa = 0.879) and D2-40 (kappa = 0.892). Our study confirms that the immunohistochemical staining pattern of mesotheliomas in the NMVB UPMC TMA is similar to other studies. Our findings also show that automated image analysis provides similar results to manual scoring by pathologists, and provides a reproducible, objective, and accurate platform for immunohistochemical assessment of biomarker expression.