Digital Pathology Association (DPA) held its annual conference in October 2025. I took all the presentation titles from the online agenda, fed them to Chat GPT 5, and asked for a conference summary.
Online, https://digitalpathologyassociation.org/pv25-agenda/show/session/162
At bottom, I clip DPA's abstract on Reimbursement.
###
DPA25: From Digitization to Intelligence — How Pathology Became Agentic
By [Your Name], San Diego — October 10, 2025
At the 2025 Fall Meeting of the Digital Pathology Association (DPA25), held October 5–7 in San Diego, digital pathology officially crossed a threshold. After years of focusing on scanners, workflows, and FDA readiness, this year’s agenda was about something else entirely: intelligence, agency, and impact. The theme, evident in dozens of talks and more than sixty posters, was that digital pathology has matured from an infrastructure challenge into an ecosystem of reasoning systems — AI models, multimodal agents, and even patient-facing platforms that are starting to reshape the identity of pathology itself.
From Image to Insight
The opening sessions framed the field’s dual trajectory — one technical, one organizational.
AstraZeneca presented its Quantitative Continue Scoring (QCS) platform, a supervised AI system that measures target expression heterogeneity and subcellular localization to predict therapy response, particularly for antibody–drug conjugates. The company’s language was unmistakable: digital pathology is now a biomarker discovery engine, not just a diagnostic aid.
Meanwhile, the Society for Imaging Informatics in Medicine (SIIM) highlighted the practical realities of “going digital,” describing the difficulties of harmonizing systems through the DICOM standard and the educational infrastructure needed to bring community hospitals along. Taken together, these sessions illustrated a field operating on two speeds — one racing ahead into molecular-level precision, and one still laying the digital foundation for interoperability and trust.
PathPresenter’s talk, “Into the PathoVerse,” captured that tension perfectly. The company described an integrated environment where storage, compute, AI, and workflow coexist — a “universe” of connected pathology. But the metaphor came with a warning: without new skill sets and business models, many institutions will struggle to navigate the PathoVerse they are now entering.
AI Moves From Concept to Practice
If 2023 was the year of validation studies and 2024 was about deployment pilots, 2025 was the year of AI in production.
The API session “AI in the Trenches” offered rare candor about quality control in live environments, as laboratories discussed monitoring model drift, auditing outliers, and adjusting QA/QC plans to account for algorithmic behavior. Leica Biosystems, in turn, described how its Aperio ecosystem is being aligned with companion diagnostic development — a theme echoed across the exhibit hall as vendors positioned themselves as translational partners rather than device suppliers.
A highlight came from Pramana, Techcyte, and BD, which jointly described Austria’s Tyrolpath laboratory and its now fully digital cervical cytology workflow — an industrial-scale deployment that uses AI-assisted screening across thousands of cases each week. For many attendees, it was the first glimpse of what routine digital cytology looks like when it actually works.
Regulation and Reimbursement Take Center Stage
The meeting’s most policy-relevant moment came during the DPA’s Regulatory and Standards Task Force update, where co-chairs John Groth and Laura Chang outlined rapid progress in regulatory collaboration with the FDA. The working groups — covering AI/ML oversight, interoperability, and validation — reported a new level of maturity. Draft frameworks for plug-and-play interoperability and structured validation are expected to circulate by early 2026, potentially providing the field with the clarity it has long lacked.
For an audience keenly aware of the FDA’s April 2024 Lab-Developed Test (LDT) rule, this session carried a reassuring message: digital pathology is one of the few sectors in diagnostics where policy is keeping pace with innovation. Roche’s later session on companion diagnostics added a business dimension, connecting regulatory readiness directly to reimbursement models and clinical adoption — a discussion that felt overdue but deeply pragmatic.
The Patient Enters the Picture
A striking shift at DPA25 was how often patients appeared — not as data points but as participants.
Michigan Medicine presented its Breast Pathology Consultation Program, where patients review their own digitized slides alongside pathologists, improving understanding and trust. Pathologist Rouba Ali-Fehmi and advocate Michele Mitchell described how visualizing her own tumor transformed Mitchell’s comprehension of her disease and her confidence in treatment decisions.
That theme continued at Moffitt Cancer Center’s “Patient-Centered Pathology” session, which merged personal narrative with AI implementation. One speaker — both a cancer survivor and a pathology business leader — described digital pathology and AI not as tools of efficiency, but as “bridges between data and the human story behind every slide.” Such remarks suggested that the technology’s social implications are finally being taken as seriously as its technical ones.
Foundation Models and the Augmented Pathologist
The technical depth of the conference was remarkable. A presentation describing a hematopathology foundation model trained on 27,000 slides demonstrated that narrower, domain-specific models can outperform massive generalist ones — a rebuke to “one-model-fits-all” thinking. A Vision Transformer–CNN comparison in lymphoma reached perfect accuracy, while a study titled “Beyond Black-Box AI” evaluated large language models on pathology board-style questions, concluding that newer systems such as Gemini and DeepSeek show expert-like diagnostic reasoning but still falter in heuristic and pattern-recognition tasks.
Perhaps the most forward-looking talk was “The Augmented Pathologist: Agentic Workflows in Practice.” Here, AI was no longer a plugin but a collaborator. Using multimodal data — from eye-tracking and voice dictation to LIS integration — developers described “agentic” systems that can perceive, plan, and act under the supervision of a human pathologist. One pilot even allowed users to narrate findings aloud as the system generated draft synoptic reports in real time. The result was an image of pathology as a co-creative process between human expertise and algorithmic assistance.
Federal and Global Infrastructure
While the private sector pushed ahead, the public sector quietly revealed one of the largest data collaborations in U.S. healthcare. A joint presentation from the Department of Veterans Affairs (VA) and the Department of Defense (DoD) detailed a project to digitize millions of archival slides, linking them to longitudinal electronic records through the Joint Pathology Center. With a throughput of up to 160,000 slides per month, this may soon become the world’s largest annotated training dataset for pathology AI — a federal-scale answer to the data bottleneck that has constrained commercial development.
Across the Atlantic, the UK’s National Pathology Imaging Cooperative (NPIC) reported similar ambitions: 2.5 million slides scanned, 20 hospitals networked, and a petabyte-scale research infrastructure supporting both industry and academia. NPIC’s emphasis on patient engagement and governance showed how digital pathology can scale nationally without sacrificing ethics or transparency.
Research Horizons: Spatial, Predictive, and Three-Dimensional
This year’s scientific track illustrated how digital pathology is converging with other “omic” sciences.
A session on AI-inferred spatial transcriptomics demonstrated that image-based models can reconstruct gene-expression maps directly from H&E slides, producing spatial biomarkers predictive of tumor stage and purity. Another on AI-based recurrence prediction in lung adenocarcinoma showed that integrating cellular, clinical, and morphological features yields significantly higher prognostic accuracy than conventional methods.
Perhaps the most futuristic talk came from a group pioneering magnetic resonance histology, a rapid, non-destructive 3D imaging method capable of rendering whole-organ volumes in about an hour without interfering with standard staining. The technique, they argued, could finally unite radiology and pathology in a single volumetric workflow.
Posters and Grassroots Innovation
The poster sessions, dense and energetic, provided a glimpse of the field’s grassroots creativity. Titles ranged from “AI Identification of Invasive Breast Carcinoma with Precise Tumor Cell Counting” and “DermPath Mind: Real-Time GPT-Powered Intelligence” to “Prototype Virtual Pathologist — A Multimodal Large Language Model for Pathology Assistance.” The message was clear: the LLM revolution has reached the slide. Many projects explored generative or self-supervised learning, real-time QA, and cross-modality data fusion — all signs of a maturing but still exuberant discipline.
A Profession Redefined
By the close of DPA25, a consensus had emerged: digital pathology’s future is not about replacing humans, but about augmenting them. Ground truth itself, several speakers noted, is no longer a static label but a negotiation between expert opinion, biological evidence, and machine inference.
If earlier DPA meetings were about AI adoption, this one was about AI collaboration — between pathologists, engineers, regulators, and even patients. The scanners hum faster, the models reason deeper, and the profession, at last, seems ready to inhabit the digital universe it built.
As one presenter remarked, “We’ve digitized the slides. Now we’re digitizing how we think.”
##
From the agenda - Reimbursement.
Reimbursement Task Force Update - Navigating the US
Reimbursement Landscape
Tue, Oct 7 11:05AM - 11:35AM PT Seaport Ballroom A - D
Digital pathology adoption and AI image analysis show significant potential, but return on investment (ROI) remains a challenge. The lack of reimbursement for FDA-cleared digital pathology (DP) solutions is a major barrier to widespread adoption.
The DPA Reimbursement Task Force has a mission to 'To define and shape the pathway to reimbursement for digital pathology solutions enabling broad market access and delivery of value to patients and stakeholders."
This session will provide an update on the key work of
the DPA Reimbursement Task Force and provide a detailed look into the framework
for DP AI reimbursement that the task force is pursuing. In addition, we will
hear from experts on key changes to the reimbursement landscape in the US and
important trends that are shaping how these solutions are being brought to the
market.
Learning Objectives:
- Update
DPA members on the progress and priorities of the Reimbursement Task Force
- Provide
an overview of the current Digital Pathology reimbursement landscape and
identify key trends that will impact both WSI and DP AI algorithms
- Discuss
the implications of US reimbursement trends for the Digital Pathology
industry with a panel of experts from the DPA Reimbursement Task Force,
along with select outside consultants
