An Automated Clinical Workflow Integrating an AI-Driven PD-L1 22C3 Scoring Algorithm Into a Browser-Based Digital Pathology Solution
Michael J. Cascio, Wei Song, Lauren Lawrence, Wei-Zhong He, Kyunghyun Paeng, Jeongseok Kang, Jongseok Choi, Sooick Cho, Dylan Cairns, Hallie Rane, Eric Runde
CAP, 2023
Context: Quantitative image analysis of histopathologic tissue sections provides a promising approach for advancing precision medicine in clinical practice. However, there is limited literature on integrating digital pathology–based artificial intelligence algorithms and workflows in clinical practice. Here, we demonstrate a real-world integration for Lunit SCOPE PD-L1 (Lunit, Inc, Seoul, South Korea) for non-small cell lung carcinoma (NSCLC) currently deployed for clinical use at Guardant Health.
Design: A 3-way integration was built by implementing Lunit SCOPE PD-L1, an AI-powered PD-L1 scoring algorithm, in HALO AP, a browser-based image management system (IMS) (Indica Labs, Albuquerque, New Mexico), which was integrated with the Guardant Health (Redwood City, California) laboratory information management system (LIMS).
Results: PD-L1 22C3 immunohistochemistry is performed on NSCLC tissue sections and whole slide image (WSI) files are generated using a Leica Aperio GT450 whole slide scanner (Leica Biosystems). After WSIs are received by the LIMS and available in the IMS, WSIs are automatically transferred to a cloud instance for AI analysis by Lunit SCOPE PD-L1. Within minutes, QC information, heat maps, and tumor proportion scores are returned to the IMS, and a scoring report is prefilled. Pathologists perform review of the NSCLC WSIs and algorithm results within the IMS, and either approve the AI-generated results or reflex to manual scoring. After result approval, tumor proportion scores and heat maps are automatically returned to the LIMS for clinical report generation and release.
Conclusions: Automated clinical workflows integrating AI algorithms with IMS and LIMS can be used in clinical practice to simplify and strengthen pathology workflows.