| Abstract [eng] |
Treatment decisions for breast cancer are significantly influenced by an accurate assessment of the tumor proliferation rate measured by Ki67 immunohistochemistry. A visual assessment of Ki67 which is commonly used in current clinical practice, has serious limitations due to a low reproducibility among the pathologists and is obscured by low cut-off values, intra-tumoral heterogeneity and a lack of standardized recommendations for manual Ki67 evaluation. In this thesis we prove that Ki67 labelling index obtained by digital image analysis outperforms visual estimates, taking manual stereological counts as a reference value. A novel methodology for comprehensive Ki67 proliferative activity quantification and heterogeneity assessment, based on the systematic subsampling of digital image analysis generated data into a hexagonal tiles was developed. This approach enables the computation of texture and spatial distribution indicators for Ki67 intra-tumor variability and to visualize Ki67 LI intra-tissue heterogeneity in the whole slide image along with an automated detection and quantitative evaluation of Ki67 hotspots. Hexagonal tiling data provide a useful model for establishing tissue sampling requirements for biomarker studies and visual estimations, which depend on intra-tissue heterogeneity and must be determined on a peruse basis. The spatial heterogeneity indicators (the bimodality status in particular) of proliferative tumor activity, measured by the digital image analysis of Ki67 expression and analyzed by the hexagonal tiling approach, can serve as an independent prognostic indicator of overall survival in breast cancer patients and outperform the prognostic power of the level of proliferative activity. |
