Mieke Zwager - Automated Ki67 hot-spot detection and analysis leads to higher Ki67 proliferation indices

Background & objectives

It is suggested that Ki67 proliferation hot-spot scoring is a prognostic and predictive marker in breast cancer. However, visual identification of Ki67 hot-spots is difficult and manual scoring is labour-intensive and prone to inter- and intra-observer variability. Automated detection and scoring of Ki67 hot-spots by digital image analysis (DIA) could aid in a standardised and reproducible assessment of the Ki67 proliferation index. The aim of this study was to compare manual Ki67 hot-spot detection and scoring with DIA Ki67 hot-spot detection and scoring.

Methods

Whole tissue sections of 117 consecutive invasive breast carcinomas were immunohistochemically stained for Ki67. Firstly, Ki67 hot-spots were detected by two independent observers and scored using a validated manual counting protocol. Secondly, manual Ki67 scores were compared with DIA on these manually annotated hot-spots. Thirdly, automated Ki67 hot-spot detection and Ki67 calculation was performed using DIA. Inter-observer agreement between manual scores and DIA in manually annotated hot-spots, and between manual observers was assessed using the coefficient of determination (R2). Means of manual scoring and DIA results were compared.

Results

102 cases were available for assessment. Correlation between both manual observers was suboptimal (R2=0.78). Manual and DIA Ki67 scores in manually annotated hot-spots showed a strong correlation (R2=0.90). Hot-spot detection by DIA (mean: 39.0%) led to higher hotspot scores compared to manual scoring (means: 33.4% and 29.4%).

Conclusion

Automated Ki67 hot-spot detection and analysis is a reliable method that leads to higher hotspot Ki67 proliferation indices.