From research prototype to clinical AI imaging workflow
AI biomarker detection is the use of machine-learning algorithms to identify, quantify, and score disease-related biomarkers on medical images or digital slides so that pathologists and radiologists can make faster, more consistent treatment decisions within routine clinical workflows. Over the last few years, these tools have shifted from experimental pilots to clinical AI imaging systems that plug into scanners, image-management software, and laboratory platforms. This transition is clearest in oncology, where pathology AI algorithms now assist with PD-L1 cancer scoring, a key biomarker that helps decide which patients receive immunotherapy and in what sequence. Instead of remaining stand-alone research models, these algorithms are being wrapped in regulated products, integrated with picture archiving and communication systems, and exposed through app-like stores so hospitals can deploy them at scale with minimal disruption.
Leica, Indica Labs and Lunit push PD-L1 cancer scoring toward scale
Leica Biosystems, Indica Labs, and Lunit are turning PD-L1 cancer scoring into a production-ready service inside Leica’s digital pathology stack. Their first joint product, Lunit SCOPE PD-L1 CAL10 NSCLC, is an AI biomarker detection algorithm tuned to Leica’s PD-L1 primary antibody and distributed through the Aperio AI Store. It sits on top of a workflow that combines Leica’s Aperio GT 450 scanner and Indica Labs’ Aperio HALO AP DX image-management software, cleared for clinical digital pathology. In this model, the AI performs the first slide read and provides a quantitative PD-L1 score for the pathologist to confirm. According to Leica Biosystems, current PD-L1 scoring suffers from “meaningful disagreement” around 1% and 50% cutoffs that guide therapy choice, and automating the initial read aims to reduce inter-observer variability while keeping the pathologist in control of the final diagnosis.
Standardized pathology AI algorithms through an open, app-store model
Leica’s strategy shows how medical imaging AI deployment is becoming more standardized. By owning the end-to-end workflow—from tissue processing to scanning to image management—the company can guarantee that pathology AI algorithms see similarly prepared slides and consistent image quality. The Aperio AI Store then acts as a distribution layer, where third-party tools from Lunit, MindPeak, Indica Labs or pharma partners connect through software development kits and APIs. This open-but-controlled model lets pathologists add PD-L1 cancer scoring, additional immunohistochemistry biomarkers, or future computational companion diagnostics without rebuilding infrastructure. It also supports a revenue-share model so algorithm developers reach Leica’s installed base through the same interface. The result is that hospitals can deploy AI biomarker detection like any other software plug-in, while preserving a single, traceable workflow that supports regulatory review and future reimbursement.
4DMedical and contextflow extend clinical AI imaging into thoracic care
In thoracic imaging, 4DMedical’s planned acquisition of Vienna-based contextflow shows the same shift from isolated algorithms to integrated clinical AI imaging services. 4DMedical is known for its CT:VQ platform and functional lung imaging, while contextflow develops AI software that helps radiologists identify and characterize lung diseases and cancer. By bringing contextflow’s CE-marked products, established customer base, and local team under one roof, 4DMedical gains an immediate clinical and commercial platform across European respiratory and thoracic imaging markets. The deal expands its offering beyond functional imaging into AI-assisted lung cancer screening and routine thoracic reads. For hospitals, that means lung AI will arrive as part of a combined portfolio, tied into existing scanners and workflows rather than as a research add-on, and backed by reimbursement pathways that have already been explored by contextflow.
What it takes to move AI biomarker detection into routine care
The path from promising model to everyday tool depends on partnerships rather than stand-alone AI. Clinical-grade biomarker detection requires digital pathology platforms such as Leica’s, imaging specialists like Indica Labs, AI vendors such as Lunit and contextflow, and clinical users to collaborate on validation, regulatory submissions, and integration. Adoption is still early: the College of American Pathologists reported that 28% of practice leaders digitize slides with whole-slide imaging, and only 10% use remote sign-out for primary diagnosis. Yet emerging computational companion diagnostics and reimbursement pathways are starting to change the economics. As Leica’s and 4DMedical’s deployments show, the next phase is less about novel algorithms and more about delivering consistent, reproducible PD-L1 cancer scoring and lung assessments inside the systems pathologists and radiologists already trust.
