Leica Biosystems, Indica Labs, and Lunit announced on May 26, 2026, a collaboration to develop image-analysis algorithms for IHC, with an initial focus on PD-L1 and emerging biomarkers in oncology research. The first announced product is Lunit SCOPE PD-L1 CAL10 NSCLC. It is designed to work with the Leica Biosystems PD-L1 CAL10 primary antibody and is available through the Aperio AI Store.
The news matters for pathology practice because the announcement is not about a separate model added on top of the laboratory after the work is complete. The practical idea is that the algorithm enters a known pathway: Leica IHC staining, slide scanning on Aperio GT 450, image management through Indica’s Aperio HALO AP, then quantitative analysis from Lunit SCOPE. This type of connection gets close to the question that matters to clinical-research laboratories: can quantitative analysis enter the daily pathway without rebuilding the system around a single tool?
Details of the announcement
According to Leica’s statement, the collaboration will focus on supporting pharmaceutical partners in bringing AI-supported digital pathology tools into routine clinical research. The reason is clear to anyone who has worked on PD-L1 or semi-quantitative IHC markers: variation between readers, borderline cases, and study sites remains a source of pressure in multicenter trials. Measurement tools do not remove the pathologist’s judgment, but they may make measurement more reviewable, especially when therapeutic or analytical thresholds sit close to gray zones.
The first product, Lunit SCOPE PD-L1 CAL10 NSCLC, targets non-small cell lung cancer and is used with PD-L1 CAL10. Availability through the Aperio AI Store means the proposed adoption point goes beyond downloading a file and running it outside the system. It moves toward adding an algorithm inside an environment already used by some laboratories and pharmaceutical companies. That operational difference should not be underestimated. Many image-analysis solutions fail in the laboratory not only because the model is weak, but because of friction around file transfer, slide matching, user permissions, and producing results in a format study teams can accept.
Why does this matter to pathologists?
Manual PD-L1 reading is familiar, but it becomes tiring when the work moves from a single diagnostic case to a research project or large drug trial. In a research setting, the requirement usually goes beyond one decision and moves toward auditable data: who read the slide, which algorithm version was used, how non-tumor areas were handled, and whether the result can be reproduced during later review.
Having an algorithm inside the Aperio pathway may help make these questions part of the work system instead of leaving them scattered between scanner, image server, analysis software, and external spreadsheets. For the pathologist, the value is not an automated number on the screen. The value appears when the team can compare human reading with quantitative analysis, open areas of disagreement, and identify cases that need second review instead of handling every slide the same way.
In PD-L1 specifically, the problem is not that every case is difficult. Many are clear. The real challenge is in cases close to the threshold, small samples, intratumoral heterogeneity, and separating tumor cells from stained inflammatory or stromal cells. Any tool claiming to help here must show how it handles these details, not just produce a final percentage.
The strength: connecting stain, image, and analysis
Leica on the staining and scanning side, Indica on image management, and Lunit on the algorithm side give the collaboration a practical direction. Computational pathology does not work in isolation. Staining quality, preparation stability, scan settings, data management, and slide-acceptance criteria all affect algorithm performance. Linking the tool to a defined pathway may reduce some sources of variation that appear when a model is trained in one environment and used in a completely different one.
But this connection also brings a professional question. The closer an algorithm comes to a closed or semi-closed workflow, the more transparency matters: what are the limits of use? Which specimens are acceptable? How are versions documented? Can the laboratory export and review data outside the platform? Is there a clear way to handle rejected slides or cases where the model gives an unstable result?
Research use, not diagnostic use
The statement clearly says that the Lunit SCOPE PD-L1 CAL10 NSCLC algorithm, as well as specific elements of the system, are intended for research use and not for use in diagnostic procedures. It also explains that Aperio HALO AP has the CE-IVDR mark for diagnostic use in Europe, the United Kingdom, and Switzerland, while it remains research use only in the United States and does not have FDA clearance for clinical diagnosis there.
These details are not a legal footnote. They define the correct way to read the news. The announcement does not mean a laboratory can use the algorithm tomorrow to issue a clinical PD-L1 result for a patient across all markets. The closer meaning is that there is a pathway aimed at drug research and biomarker development, and it may test what quantitative IHC reading looks like when connected to a known image and data-management platform.
What should be watched next?
For pathologists, the coming questions are more important than the announcement itself. We need published or partner-accessible validation data showing algorithm performance across different specimen types, staining gradients, and multiple centers. We also need to know how tumor cells are defined in NSCLC, how necrosis, folds, and non-tissue areas are excluded, and how heatmaps or counting regions are shown to the human reviewer.
Another area to follow is the link to emerging biomarkers. PD-L1 is the announced starting point, but the collaboration refers to other biomarkers. Here the pathologist’s role becomes more sensitive, because some markers do not yet have a standardized reading language or stable thresholds like established tests. Early quantitative analysis may help pharmaceutical companies understand spatial distribution and intratumoral heterogeneity, but it may also create excess confidence if the statistical and pathologic design is not clear.
A professional reading of the news
This collaboration reflects an expected market direction: IHC algorithms will not remain separate tools that display isolated numbers. They will gradually enter staining, scanning, and image-management pathways. Pathologists will need to evaluate them as part of a complete work system, not only by model accuracy.
The practical position is cautious welcome. An algorithm inside a known pathway may reduce friction and make use in drug research easier. At the same time, professional judgment should remain tied to three points: scope of use, validation data, and how the result is incorporated into a report or study file that can be reviewed. If the companies answer these points clearly, the news will be more than a partnership announcement. It will be an important test of how quantitative IHC analysis moves from technical demonstrations to daily work that pathologists can trust.