Proliferation assessment in breast carcinomas using digital image analysis based on virtual Ki67/cytokeratin double staining

Abstract

Manual estimation of Ki67 Proliferation Index (PI) in breast carcinoma classification is labor intensive and prone to intra- and interobserver variation. Standard Digital Image Analysis (DIA) has limitations due to issues with tumor cell identification. Recently, a computer algorithm, DIA based on Virtual Double Staining (VDS), segmenting Ki67-positive and -negative tumor cells using digitally fused parallel cytokeratin (CK) and Ki67-stained slides has been introduced. In this study, we compare VDS with manual stereological counting of Ki67-positive and -negative cells and examine the impact of the physical distance of the parallel slides on the alignment of slides. TMAs, containing 140 cores of consecutively obtained breast carcinomas, were stained for CK and Ki67 using optimized staining protocols. By means of stereological principles, Ki67-positive and -negative cell profiles were counted in sampled areas and used for the estimation of PIs of the whole tissue core. The VDS principle was applied to both the same sampled areas and the whole tissue core. Additionally, five neighboring slides were stained for CK in order to examine the alignment algorithm. Correlation between manual counting and VDS in both sampled areas and whole core was almost perfect (correlation coefficients above 0.97). Bland-Altman plots did not reveal any skewness in any data ranges. There was a good agreement in alignment (>85 %) in neighboring slides, whereas agreement decreased in non-neighboring slides. VDS gave similar results compared with manual counting using stereological principles. Introduction of this method in clinical and research practice may improve accuracy and reproducibility of Ki67 PI.

Digital image analysis of Ki67 in hot spots is superior to both manual Ki67 and mitotic counts in breast cancer

“…The scoring of cells as either Ki67‐positive or Ki67‐negative was performed with the Visiopharm Ki67 Breast Cancer APP, whereas the scoring of PHH3 was performed with a PHH3 APP that was custom‐made for this project. This system has previously been shown to have excellent reproducibility and correlation with manual assessments …”

“…This system has previously been shown to have excellent reproducibility and correlation with manual assessments. 16,17,33,34 We evaluated the percentage of cells stained by Ki67 and PHH3 in the following regions: (i) the tumour's invasive margin; (ii) the highest-intensity hot spots; and (iii) across the full tumour section.…”

Quality assessment of Ki67 staining using cell line proliferation index and stain intensity features

“…Medical image analysis methods can enable objective stain quality control by reproducible quantification of staining characteristics . These techniques for Ki67 quantification have been applied in tissues but never as surrogate of IHC staining quality in cell lines . Extracting features such as Ki67 PI and stain intensity of the cell nuclei within the cell line may be used to serve as a more robust and objective assessment of the stain quality .…”

Digital image analysis in breast pathology—from image processing techniques to artificial intelligence

“…Visiopharm (Hoersholm, Denmark) Virtual Double Staining techniques is using a pancytokeratin-stained tissue section, which is aligned to the IHC-stained biomarker of interest and enables automated detection of tumor regions (Fig 2). [91][92][93] The abovementioned commercial platforms operate on input from a WSI scanner, whereas the Aperio Digital Pathology (Leica Biosystems, Nussloch, Germany) platform operates by integrating a digital microscope with the image software. The Aperio Digital Pathology usersupervised platform has been compared with the automated Definiens Tissue Studio platform for classification of ER and PR IHC positivity.…”

“…DIA has shown excellent reproducibility, although limited to subsets with individual biomarkers or smaller cohorts. 93,[95][96][97] Hartman et al have recently shown the advantages in congruence to gene expression assays and the prognostic power of digital image analysis compared with current manual methods of biomarker assessments. 91 Automated image analysis has been applied to analyze several biomarkers, including HER2 expression, holding promise to reduce the need for additional ISH analysis in HER2 equivocal cases 95,98 and significantly reducing inter-and intraobserver variability.…”