Image Analysis of Ki-67 IHC Stain for Breast Carcinoma – A Study Comparing Two Image Analysis Systems
Amy M Plagge, Taofic Mounajjed, Jesse S Voss, Douglas M Minot, Delores (Lorie) J Tuve, Benjamin R Kipp. Mayo Clinic, Rochester, MN
Background: Ki-67 immunohistochemisty (IHC) is often used to evaluate the proliferative activity of paraffin-embedded breast cancer and may provide useful prognostic information. The goal of this study was to evaluate the reproducibility and analysis time of different digital Ki-67 scoring methodologies on breast carcinoma specimens.
Design: Ten cases containing invasive or metastatic breast carcinoma from specimens previously analyzed clinically for Ki-67 IHC were selected for this study. The slides were scanned with our current analysis method using the Dako ACIS III (Automated Cellular Imaging System) and the Aperio ScanScope instrument to create digital images. The images were independently analyzed by three Cytotechnologists using three different scoring methods. ACIS analysis was performed by selecting 8-10, 40X sized fields of view representative of the overall staining within the tumor. The specimens analyzed using Aperio software were circled to encompass at least 75-100% of the invasive tumor by two methods: 1)a Wacom Cintiq 21UX tablet monitor (21.3 inch screen) with an interactive pen (tablet tracing) and 2)a standard 19 inch computer monitor with mouse (mouse tracing). All analyses were timed with results recorded.
Results: Data analysis revealed that the three Ki-67 quantitation methods produced very similar results. Ki-67 scores obtained by both Aperio tumor tracing methods correlated with those obtained by the ACIS method, with R2 values of 0.934 and 0.930 for the mouse and tablet tracing methods, respectively. The mouse and tablet tracing methods also had excellent correlation of Ki-67 scores (R2=0.993). While the three scoring methods produced similar Ki-67 results, the two Aperio tumor tracing methods tended to produce more consistent Ki-67 scores for each case by different technologists (SD=2.88 and 1.93 for tablet and mouse methods) than the ACIS method (SD=5.75). However, the average analysis time per case was lowest using the ACIS instrument (3 min, 8 sec) compared to 11 minutes, 23 seconds for the mouse tracing method and 10 minutes, 26 seconds for the tablet tracing method.
Conclusions: While Ki-67 quantitation using the ACIS instrument can be performed considerably quicker than the Aperio tumor tracing methods, it is not as reproducible as the Aperio tracing methods. Aperio analysis using a tablet monitor and pen reduces analysis time when compared to a standard computer monitor and mouse. Further studies are needed to find ways to reduce Aperio analysis time while maintaining the high reproducibility of tumor tracing techniques.
Wednesday, March 6, 2013 9:30 AM
Poster Session V # 269, Wednesday Morning