What AI biomarker scoring means for cancer care
AI biomarker scoring is the use of machine-learning algorithms to detect, quantify, and interpret molecular or protein signals on pathology slides so cancer biomarkers are measured in a consistent, reproducible way across patients, labs, and time. In oncology, those biomarkers decide which patients qualify for targeted drugs and in what sequence they receive therapies, so small differences in scoring can change treatment paths. Until now, most PD-L1 detection algorithms and similar tools have lived in research settings or pharma-sponsored studies, separate from the scanners, stains, and workflow tools that pathologists use every day. That separation limited impact: algorithms were promising, but hard to run at clinical scale. Today, a new class of partnerships between imaging, pathology software, and AI firms aims to close that gap and bring oncology pathology AI into routine practice.
Leica, Indica Labs and Lunit build an end-to-end workflow
Digital pathology adoption accelerated when remote work demands made whole-slide imaging more attractive and temporary FDA enforcement discretion lowered perceived barriers. Leica Biosystems has responded by building a tightly integrated workflow: its Aperio GT 450 DX scanner feeds images into a clinically validated version of Indica Labs’ browser-based Aperio HALO AP DX image-management software, which Leica distributes for clinical use. On top of this infrastructure sits the Aperio AI Store, an app-like layer where third-party oncology pathology AI tools can run inside the same workflow that handles sectioning, staining, and visualization. According to Leica’s Karan Arora, the aim is an open, best-of-breed ecosystem controlled for end-to-end standardization through software development kits and APIs. AI biomarker scoring algorithms from partners such as Lunit, MindPeak, pharma developers, and major labs can plug in without disrupting how pathologists already work.
From research pilots to a live PD-L1 detection algorithm
The clearest sign that AI biomarker scoring is entering production is the first live biomarker algorithm in Leica’s Aperio AI Store. Leica, Indica Labs, and Lunit have released Lunit SCOPE PD-L1 CAL10 NSCLC, a PD-L1 detection algorithm tuned to Leica’s CAL10 primary antibody and optimized for non–small cell lung cancer. The algorithm reads whole-slide images produced by the Aperio GT 450 scanner and delivers a quantitative PD-L1 score directly in the HALO AP DX environment. Pathologists still make the final call, but they start from an AI-generated tumor percentage and cutoff suggestion rather than a manual estimate under the microscope. This move links Leica’s assay chemistry, hardware, and workflow with Lunit’s oncology pathology AI in one clinically oriented pipeline, turning what used to be a research-only asset into a tool pathologists can access through the same interface they use for daily cases.
Reducing variability in PD-L1 scoring and oncology workflows
PD-L1 scoring highlights why AI biomarker scoring matters. Current manual reads show meaningful inter-observer variability: two pathologists may set different cutoffs at clinically important thresholds such as 1% and 50%, leading to different decisions on whether a patient receives chemotherapy first or moves directly to an antibody–drug conjugate. Arora describes the new approach as “read the slide once, the same way everywhere.” The slide is scanned, the AI performs the first analysis, and the pathologist reviews and confirms the result. This reduces variability between readers and between centers, because the PD-L1 detection algorithm does not depend on local experience. In practical terms, oncology teams get more consistent, reproducible biomarker calls to anchor therapy selection, while pathologists can shift from laborious manual counting toward higher-level interpretation and correlation with clinical data.
Clinical AI validation and the coming verification layer
Multi-vendor efforts like Leica–Indica–Lunit underline how clinical AI validation is becoming a shared priority. For oncology pathology AI to scale, labs need not only accurate PD-L1 detection algorithms but also clear proof that each version has been tested on the right samples, under known conditions, and within a regulated workflow. That need is driving attention to verification layers that sit beside clinical systems. Aimwell Partners, the company behind the AimwellBio healthcare intelligence network, is exploring infrastructure that can attach a proof trail to AI-generated signals, showing who submitted an item, when it changed, and what evidence backs it. According to Aimwell, as major platforms race to embed AI across documentation, charting, and decision support, they are “expanding the market for verification.” The same logic applies in pathology: as more biomarker scores come from algorithms, their provenance and integrity must be independently verifiable.
