Assessment of Automated Image Analysis of Breast Cancer Tissue Microarrays for Epidemiologic Studies

摘要: Background: A major challenge in studies of etiologic heterogeneity breast cancer has been the limited throughput, accuracy, and reproducibility measuring tissue markers. Computerized image analysis systems may help address these concerns, but published reports their use are limited. We assessed agreement between automated pathologist scores a diverse set immunohistochemical assays done on microarrays (TMA). Methods: TMAs 440 cancers previously stained for estrogen receptor (ER)-α, progesterone (PR), human epidermal growth factor 2 (HER2), ER-β, aromatase were independently scored by two pathologists three (TMALab II, TMAx, Ariol). Agreement negative/positive was measured using area under receiver operating characteristics curve (AUC) weighted κ statistics categorical scores. also investigated correlation mRNA expression levels. Results: excellent ER-α PR (AUC range = 0.98-0.99; 0.86-0.91). Lower levels seen ER-β 0.99-1.0; 0.80-0.86) both 0.85-0.96; 0.41-0.67) HER2 0.94-0.97; 0.53-0.72). For PR, there strong ( ρ 0.67-0.74) 0.67-0.77). more strongly correlated with 0.63) than 0.41-0.49). Conclusions: Automated markers is promising approach scoring large numbers tissues epidemiologic investigations. This would facilitate heterogeneity, which ultimately allow improved risk prediction better prevention approaches. Cancer Epidemiol Biomarkers Prev; 19(4); 992–9. ©2010 AACR.