Abstract
Immunohistochemical data (IHC) plays an important role in clinical practice, and is typically gathered in a semi-quantitative fashion that relies on some degree of visual scoring. However, visual scoring by a pathologist is inherently subjective and manifests both intra-observer and inter-observer variability. In this study, we introduce a novel computer-aided quantification methodology for immunohistochemical scoring that uses the algebraic concept of persistent homology. Using 8 bit grayscale image data derived from 90 specimens of invasive ductal carcinoma of the breast, stained for the replicative marker Ki-67, we computed homology classes. These were then compared to nuclear grades and the Ki-67 labeling indices obtained by visual scoring. Three metrics for IHC staining were newly defined: Persistent Homology Index (PHI), center coordinates of positive and negative groups, and the sum of squares within groups (WSS). This study demonstrates that PHI, a novel index for immunohistochemical labeling using persistent homology, can produce highly similar data to that generated by a pathologist using visual evaluation. The potential benefits associated with our novel technology include both improved quantification and reproducibility. Since our method reflects cellularity and nuclear atypia, it carries a greater quantity of biologic data compared to conventional evaluation using Ki-67.
Highlights
Breast cancer is the most commonly diagnosed cancer in Japanese women[1], for which invasive ductal carcinoma is the most frequent type of invasive carcinoma
Breast cancers are classified as low, intermediate, or highly proliferative according to their Ki-67 labeling index, with these categories linked to labeling indices of under 15%, 16–30%, or over 30%, respectively[7]
We propose a novel quantitative evaluation method for immunohistochemical labeling based on persistent homology
Summary
Breast cancer is the most commonly diagnosed cancer in Japanese women[1], for which invasive ductal carcinoma is the most frequent type of invasive carcinoma. Gene expression profiling enabled us to classify breast tumors under five intrinsic subtypes, i.e., luminal A, luminal B, HER2 over-expression, basal and normal-like tumors[8,9]. This classification is approximated by information of ER, PgR, HER2, and Ki-6710,11. In terms of shape recognition in image analysis, the classical tools for capturing the characteristics of shape from black and white images are sensitive to noise in data sets Coping with this difficulty, persistent homology enables us to measure stable topological features in a meaningful manner. We investigated the correlation between the newly defined persistent homology-derived index, nuclear grade, and Ki-67 labeling index, as obtained by traditional visual scoring
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
