Auto-Analysis for Ki-67 Indices of Breast Cancer Using Specified Computer Software and a Virtual Microscopy

Ki-67 index is one of important markers that is correlated with chemotherapy response and prognosis of breast cancer patients. However, Ki-67 index is not easily provided and are limited by intra-observer error and potentially subjective decision making. We performed this study to develop an objective auto-analysis system to count Ki-67 indices. A total of185 invasive breast cancer cases were used. Immunohistochemical staining was performed using auto-stainer and MIB-1 antibody. The results were stored digitally by virtual microscopy and auto-analyzed by Genie/Aperio software (Vista, CA, USA). As for Ki-67 indices, a good correlation was observed between direct ocular observations and auto-analysis techniques (r = 0.94, p < 0.001). The index examined by auto-analysis was significantly correlated with nuclear atypia, mitotic counts, and nuclear grade of pT1 breast cancers. Auto-analysis of 5 high power fields was better correlated with nuclear grade than that of whole fields. Further, the Ki-67 index was better correlated with mitotic counts than with nuclear atypia. Auto-analysis can provide results concordant with those obtained by direct ocular observation in a short time. Auto-analysis is more likely to result in an objective observation and provide a means by which to standardize methods for immunohistochemical Ki-67 indices of breast cancer.