Physician, pathologist, and innovator focused on advancing precision medicine in cancer care
We’re entering an era where tissue biomarkers won’t be measured by eye. Even when they’re IHC-based, the next generation of assays won’t rely on pathologists estimating intensity or proportion scores like we do for HER2 or PD-L1. That’s one clear takeaway from Pathology Visions 2025 (Digital Pathology Association) AstraZeneca’s QCS system for TROP2 was a standout. The FDA now recognizes its “normalized membrane ratio” (NMR) as a novel biomarker—a computational measurement that quantifies membrane and cytoplasmic expression patterns linked to drug internalization and efficacy. It’s literally impossible to score an NMR by eye. This marks a turning point. In AstraZeneca’s model, deep-learning systems (ART and SSTC) handle the region and subcellular segmentation. The locked, reproducible NMR outputs are the biomarker. The pathologist’s role shifts to quality assurance—verifying the image quality, overlays, and workflow integrity—not subjective interpretation. This is what a “human-in-the-loop” system really means. We also saw new data from Tamara Jamaspishvili using PathAI‘s AIM-PD-L1 algorithm in NSCLC: AI-derived PD-L1 quantification (AIM-PD-L1) stratified survival better than human pathologists. In fact, AI reclassified nearly 30% of cases—and those reclassifications correlated more closely with objective response and overall survival. We’ve known that PD-L1 interpretation by pathologists is inconsistent. Now we can measure how inconsistent. More biomarkers will follow—HER2 may be next. The most powerful biomarkers ahead may not be visible at all. They’ll be defined by quantitative patterns that correlate with mechanism of action and clinical benefit, not by thresholds that a human pathologist can eyeball. The pathologist’s role isn’t disappearing—it’s evolving. From measuring by eye to governing the quality, validity, and interpretability of algorithm-driven diagnostics. As AI learns to quantify what the human eye can’t perceive, diagnostic pathology is shifting—from seeing to measuring. #DigitalPathology #Pathology #ArtificialIntelligence #PrecisionOncology #PathVisions25 #AIMedicine #Cancer #biomarkers
This is for sure happening—-but only if pathologists move to a digital workflow FIRST. Without that—the trust and seamless integration will never arrive.
David R. Braxton M.D. This novel marker represents the first shot over the bow of what lies ahead for pathology. CorePlus transitioned to primary digital reading from whole slide images (WSI) on January 1, 2020. Shortly thereafter, the organization integrated AI algorithms across prostate, breast, and cytology, among other domains. These include multimodal prognostic and predictive models capable of generating genomic-level insights directly from the same primary images on which the original diagnosis was rendered. The precision and efficiency enabled by this transformation have ushered in a new era for pathology, one that redefines the very role of the pathologist. In the not-too-distant future, we may witness an evolution from “pathologist” to “diagnostician”, a professional seamlessly integrating morphology, molecular data, and machine intelligence to deliver deeper, faster, and more personalized answers for patients. Juan Carlos Santa-Rosario MD Maria Ayme Ramos MD Ibex Medical Analytics Epredia Techcyte AIxMed, Inc. ArteraAI Ataraxis AI Joe Lennerz
100% agree David R. Braxton M.D. it’s time to shift the gears towards quantitative AI-powered assessments for CDx and all tissue-based biomarker testing! The community needs to generate more clinically driven evidence for regulatory bodies! So glad that this very important topic was heavily emphasized by Digital Pathology Association at this year’s meeting 👏
But!!! Let’s make sure sample sufficiency and fit for purpose is there. Carolyn Compton had been telling you this for years. You must have tissue stewardship. Great to have these things but if it’s garbage in its garbage out. https://thepathologist.com/issues/2018/articles/mar/garbage-in-garbage-out
Thank you for these valuable insights ! Radiology is also changing but less in the “quantitative way”. Its next step for radiology and AI. I like these quotes from you as they reflect the major trend : “It’s literally impossible to score an NMR by eye. This marks a turning point.” “The pathologist’s role shifts to quality assurance—verifying the image quality, overlays, and workflow integrity—not subjective interpretation. This is what a “human-in-the-loop” system really means.” That’s why I identified pathology as the next discipline to be deeply transformed by AI after radiology, and invested in some companies, including Primaa. Looping in the SOCKEEL Family behind Primaa: Stephane Sockeel , Marie SOCKEEL and Fanny SOCKEEL MASSART 🚀
Multiplex is way to go with DP and maybe AI. Pathologists are not ELISA machines lol
Completely agree with this take. Moving from measuring by eye to algorithm-driven quantification isn't just more precise, it's capturing biology we couldn't reliably detect before. HER2 may well be next.
Excellent points, David. I think this is a major next step in practical clinical adoption and benefit of AI in cooperation with pathologist oversight.
This has been my thoughts over the past few years and I’m confident pathology will get to the point where AI algorithms will be used to detect tissue anomalies/cancers with WSI.
Helping clinicians bring clarity to cancer using NGS and Digital Pathology.
2wThis!!!! I’m starting to feel like Digital Pathology is the new NGS. Having been at Illumina early days and seeing the technology grow into the clinic was incredible. Digital Path is next.