Qualitopix: Artificial intelligence-based quantitative quality assurance of immunohistochemistry stain-ing-The Henry Ford Health experience
Recommended Citation
Baba O. Qualitopix: Artificial intelligence-based quantitative quality assurance of immunohistochemistry stain-ing-The Henry Ford Health experience. Am J Clin Pathol 2023; 160:S101-S102.
Document Type
Conference Proceeding
Publication Date
11-1-2023
Publication Title
Am J Clin Pathol
Abstract
Introduction/Objective: Immunohistochemistry (IHC) offers crucial patient data for diagnosis, prognosis, and treatment. Pathologists currently assess stain quality subjectively, comparing control sections to patient tissue. Qualitopix (Visiopharm, Denmark) is a cloud-based platform that enables objective, quantitative analysis of staining consistency through external cell line controls. Methods/Case Report: Over six months, we tested 1121 slides from five cell-line blocks with epitopes of different intensities for ER, PR, Ki-67, Her-2 Neu, and PD-L1. Slides were stained using Ventana Benchmark Ultra (Roche, Basel, Switzerland) for ER, PR, and Her-2 and Dako Omnis (Agilent, Santa Clara, California, United states) for Ki-67 and PD-L1. They were scanned with DP 200 scanner (Roche) and uploaded to Qualitopix for image analysis. Cells were detected using artificial intelligence and classified based on diaminobenzidine (DAB) staining intensity, reported as H-scores (0-100). Outliers were determined at 1 standard deviation. a selection of outliers was rescanned. Inter- scanner comparisons and repeatability studies were performed for ER and 50 within and between the DP 200 and HT scanners (Roche). Results (if a Case Study enter NA): Our analysis revealed a 26-34% rate of outliers at 1 SD for the five stains, 10 times higher than the currently reported number from our lab. Most outliers (70-100%) remained out of range at rescanning, suggesting pre-analytical technical issues. 5-33% of slides failed analysis due to technical errors. Poor discrimination between sequential cores of varying intensity for ER and Ki-67 was noted. Inter-scanner comparison and repeatability studies demonstrated precise and consistent H-scores within and across both scanning platforms for ER and PR. Conclusion: Qualitopix offers automated, objective, and quantitative quality assurance for IHC stain quality assessment, a hitherto manual, subjective and qualitative-driven process. Our findings emphasize the need for continuous IHC quality monitoring. The origin of outliers, technical failures, and their correlation with pathologist's subjective findings require further investigation.
Volume
160
First Page
S101
Last Page
S102
