September 2026—What do leaders of the Digital Pathology Association have to say about where digital pathology stands now and what to expect from AI? Six DPA leaders met online July 24 with CAP TODAY publisher Bob McGonnagle to talk about the technology and the tools. Many patients still don’t understand pathology’s role, said Orly Ardon, PhD, MBA, of Memorial Sloan Kettering Cancer Center, “and AI can create additional concerns if it is not explained properly.”
Dr. Ardon is a member of DPA Board of Directors, as are Eric Walk, MD; Bethany Williams, MBBS, PhD; David Lahm, MBA, MS Pharm; and Nathan Buchbinder. Michael Rivers is DPA Foundation president.
The adoption of digital pathology was top of mind in last year’s roundtable. We posited that the adoption rate was about 15 to 20 percent of labs, and we know it has gone up. Laboratories are either at 100 percent digital reading or moving in that direction or making plans for digital pathology. Eric Walk, is that a fair characterization?
Eric Walk, MD, chief medical officer, PathAI: Yes, it is. It’s good to see the market data catch up to the anecdotes we’re collecting, especially those of us who speak to pathologists almost daily. No one I speak to says, “This DP thing’s not going to happen. I’m sitting out.” Those who are sitting out now know they have to go digital. Data from a late 2025 Laboratory Economics poll show that 50 percent of pathology labs are digital in some capacity and another 30 percent plan to go digital in some capacity over the next 12 months. That leaves only 20 percent that are not digital at all, which correlates with what I see day to day. The DPA just did a survey as well that roughly correlates to those numbers.
In addition to the traditional drivers of adoption—workflow efficiency and precision medicine—an interesting dynamic I’m starting to hear consistently is around hiring. When pathology labs look to hire now, a standard question they get from pathologists is if they can work from home using digital pathology. Retaining and hiring talent has become an important driver.
We’re happy to see the adoption headed in this way, and it won’t be long before we can say most labs are fully digital.
Bethany Williams, many people in the United States have noted that a shortage of pathologists is what led to the adoption of digital pathology. It is also true in Europe. Do you agree? And what are the trends now in the U.K.?
Bethany Williams, MBBS, PhD, lead for digital pathology education and training, National Pathology Imaging Co-operative, and PhD fellow in digital pathology, Leeds Teaching Hospitals NHS Trust: Yes, I agree. In terms of digitization, it’s similar to what you said—departments have already procured what they need or are in the process of doing so or are moving from their first digital pathology solution to something that better suits their needs. Now that people seem to be attaining the funds to deploy these systems, the issues are how we make the case for sustainability as the costs of storage and the demands for storage grow.
Digital pathology has had a fantastic impact not only on recruitment to the profession, particularly for hard-to-fill posts and locations, but also on patients’ lives, by speeding up pathways from initial biopsy assessment to specialists, subspecialists, diagnosis, and multidisciplinary team meetings and by making access to secondary and expert opinions more equitable.
Mike Rivers, what are you seeing as you begin to survey the scene from a slightly different perch than you had before?
Michael Rivers, Rivers Strategic Advisors LLC (formerly vice president and lifecycle leader of digital pathology, Roche Tissue Diagnostics): We’re finally seeing adoption begin to tip and move and accelerate in the way we hoped for many years it would. Getting the infrastructure, the digitization layer, in place is critical to unlocking amazing AI opportunities.
We’ve seen an increasing number of FDA breakthrough device designations for exciting tools that are coming, as well as continuing and growing investment from pharma. Labs are recognizing there’s an opportunity to drive value for the patient, themselves, and their businesses. By ensuring the infrastructure is in place, they can leverage, unlock, and provide access to these wonderful tools that are in the pipeline.
David Lahm, talk about Eli Lilly’s view and your own view of precision oncology and how digital pathology and AI will be governing the future.
David Lahm, MBA, MS Pharm, associate vice president, global diagnostics, Eli Lilly and Company: From a pharma perspective, we care about finding biomarkers and developing those biomarkers early. Digital pathology offers us a unique platform to interrogate samples for biomarkers, particularly multimodal data sets that health systems are sitting on, where they have archived digital pathology images connected with genomic profiles connected with clinical outcomes and treatment selection. That’s a trove of valuable data that can be used in biomarker discovery and development.
I work within medical affairs, so we’re thinking later in the journey. Once we’re approaching the approval of a targeted medicine, or have a targeted medicine, we ask how we can streamline the patient journey and pathway to reduce attrition points, or clinical care gaps, to maximize the impact of the medicine we’re bringing to patients in real life. We get excited about reducing wait times through efficient triage, leveraging predictive algorithms that can interrogate H&E images for the likelihood of biomarkers—things that can increase health equity or decrease time to biomarkers, all to maximize the impact of our medicines.
Each year at the ASCO meeting I am reminded that many patients are not getting the proper testing on the basis of the pathology samples that have been analyzed to date. There’s been a great workflow problem of getting those patients into the system and getting proper turnaround times and ultimately the proper care. Nathan Buchbinder, can you comment on this aspect of your work at Proscia?
Nathan Buchbinder, chief strategy officer, Proscia: For us this sits at the crux of why many labs are going digital and why digital pathology will also be so impactful for pharma. The stats are intense: About 40 percent of patients with breast or lung cancer who are qualified to receive precision therapies receive the requisite testing. Even for patients who are tested, they wait weeks or longer to get results. By the time a clinician wants to get them on a targeted therapy, it’s possible the patient has already begun a contraindicated treatment or progressed beyond the point where they can be given the care they need.
Digital pathology and the images it generates provide a rich biological signal, right at the moment of diagnosis, to support timely decision-making. In sitting so far upstream from treatment, these signals can better support patient surfacing for trials and targeted therapies in a manner that is lower cost and less likely to miss the critical window for physicians to make the most informed decisions.
Pharma is often willing to pay for the value that this actionable insight delivers, creating new financial opportunities for the labs that have gone digital. And for those labs, it means serving patients better. They can offer more precise services that also reduce time to diagnosis and, more importantly, time to receiving the best treatment.
Orly Ardon, how are things working at Memorial Sloan Kettering when it comes to improving this workflow, this tight connection of biomarkers, diagnosis, and treatment?
Orly Ardon, PhD, MBA, director of digital pathology operations; associate member, Department of Pathology and Laboratory Medicine; member, Warren Alpert Center for Digital and Computational Pathology, Memorial Sloan Kettering Cancer Center: At MSK, digital pathology has helped connect pathology images, molecular testing, and clinical outcomes. We are fully digital now and the centralized digitization operation allows rapid access to quality images and associated clinical and genomic data, helping support biomarker-driven diagnosis and reduce time for treatment decisions. The infrastructure we’ve built over the years enhances precision oncology and enables research collaborations and AI development using large, well-governed data sets. The ultimate goal is connecting these data sets so we can accelerate research while also improving patient access to biomarker-driven care.
One lesson we’ve learned is that digitization itself is only the beginning. The long-term value comes from governance, metadata, quality systems, and making data accessible for clinical and research use. Institutions are likely to find that whole slide imaging hardware is not good enough. You also need to establish the infrastructure necessary to be able to use the data to build cohorts of patients, de-identify the cohorts, ensure governance systems, and have a seamless, sustainable ecosystem that allows the data to be used for research, which is a prerequisite to improving the workflow.
Dr. Williams, tell us about your experience at the National Pathology Imaging Co-operative with the challenge of organizing workflow and data so it can be useful.
Dr. Williams (NPIC): We have a different setup. The National Pathology Imaging Co-operative started as a single-site deployment at Leeds Teaching Hospitals NHS Trust, a tertiary cancer center, then built up into a regional network of laboratories all providing data into a single, vendor-neutral archive, and is now a national system. We have national specialist reporting networks. We’ve taken a step to start onboarding hospitals across the country and are now serving 12.5 percent of the country’s population.
I find it exciting that for once we don’t have a London- or Oxbridge-centric data system. We have something housed in the Northern Powerhouse of the U.K., but we also have a system in which we’re gathering data not just from tertiary cancer centers but also from small, single- and dual-handed pathology practices. We’re able to collect data from across geographical boundaries, different academic and clinical settings, laboratories with different quality issues, and we finally have a repository of data that is truly representative of U.K. pathology as it is practiced, which makes the perfect fodder for AI development, training, and evaluation. We are poised with something unique to go an extra way toward solving the problem of creating AI that is transferable across multiple settings.
Given the U.K. has a national system, you don’t have as many of the proprietary and even competitive elements that you might have in the United States from the big centers.
Dr. Williams (NPIC): We don’t have the same worries about competition, but it’s still not an easy task. Feudalism is still evident. Institutions often have decided they still want to silo it; they want to have control of their data and patients, and it takes careful negotiation and human effort to reach an understanding, even when there is a single provider.
Nathan Buchbinder, would you like to make a comment?
Nathan Buchbinder (Proscia): The common framing is that digital pathology allows labs to go from glass to data-driven workflows. Dr. Williams highlighted the effort that it’s taken for the data generated to be aggregated and assembled in a way that tears down silos. Now that we are centralizing data and running workflows, my prediction is that the challenge will shift to how we make sense of everything happening across this new data-based ecosystem.
AI is going to be critical in addressing that challenge, but not in the way we often view it today. It will be less a matter of which indication or biomarker-specific application to build and more of a rethink about what AI can do to enable pathologists and scientists. How can it help them to answer questions they didn’t even consider asking in the most fundamental aspects of their work?
In the same way that our day-to-day routines were upended, reshaped, and enhanced by the internet in the late ’90s and by LLMs like Claude and ChatGPT more recently, the same trend will happen in pathology, with monumental impact.
Dr. Ardon, a comment from you?
Dr. Ardon (Memorial Sloan Kettering): We’re understandably optimistic about the future, but the availability of data also introduces new responsibilities. One lesson we’ve learned when building the digital pathology infrastructure at MSK is that digitization is only the first step, as I said. Governance, metadata, quality systems, and accessibility ultimately determine the data’s value.
We should remember that there are patients behind those whole slide images, and we need to think of what can be done with the images to improve health care delivery. How do we prioritize research projects? How do we make sure the burning questions are being answered? We must think of what would make the most impact for patients as well as offer a return on the investment to the institution.
In our institution, we see patients from around the world. We have diverse data with multiple WSI systems. We’re seeing differences in the capturing of the data and the way in which we generate it, and there aren’t as many standards as we’d like to see. We are realizing that some of the AI tools being developed and implemented may not work for some populations or scanner systems or even images that were prepared in a certain way.
Solving one issue—putting the infrastructure together—exposed us to additional considerations that had to be resolved, and we have been actively addressing those.
Dr. Walk (PathAI): Digital pathology adoption is well underway; the next wave is AI adoption and what I’d like to call pathologist decision support. We are close to leveraging generative AI, vision language models, and agentic systems to assist pathologists. Academic center pathologists have the privilege of having residents look at every case before sign-out. In the future, every pathologist will have an agent to assist in the primary diagnosis and workup of a case. How much of that will be fully automated remains to be seen. For the indefinite future, there will be a human in the loop.
All of us in the community will have to be fluent in the language of AI, deep learning, and machine learning because if you’re going to adopt and use these tools, you need to know at least fundamentally how they work. You don’t need to code a large language model, but you need to know how a cell segmentation model or an H&E predictor works.
Dr. Williams (NPIC): To expand on the challenges we’ll be confronting around AI and its adoption and implementation, the big issue will be the human element. We need to do a lot around professional education and have pathologists reach the stage where they have the fluency to at least know what they don’t know about AI before they add to the myths that surround its use.
Patient and public acceptability are most important. We cannot take this road on our own. We have a mandate from the public to be as involved as we as pathology professionals can be in ensuring these technologies are properly developed and monitored and that the right ethical framework has been applied. We need to advocate for patients. Whether we are skeptics or proponents of AI and technologies, we need to demonstrate it and follow through on our professional curiosity. We are the people who should influence how this field progresses.
I know from reading the British press that the U.K. population is not fond of the NHS, but there is also discontent with our system in the U.S. I often think in this connection about the telecommunications industry. Its final problem was what they referred to as the last mile. They could have great systems, but at some point they had to connect to a house, a television, and so on, and that has proved to be difficult. I would think this patient access and accessibility that’s truly gratifying for both patient and physician is this last-mile problem as we see it in digital pathology and AI. Mike Rivers, can you comment on that?
Michael Rivers (Rivers Strategic Advisors): There are challenges that come with this technology that we have to embrace. Having been in digital pathology for the past 10-plus years, it seems like every few years we talk about the coming demise of the pathologist. For me, it has further reinforced how important the pathologist is. As these AI solutions have become more exciting and embedded in our daily lives, and raised more concern from some quarters, it’s critical to have not only a human in the loop but a highly educated human who has taken the Hippocratic oath and is willing to own the judgment of the diagnosis. I don’t think pathologists fear being obsoleted anytime in the future; they will be critical to this technology.
We have to embrace the technology as well, not only the digitization but also the lab automation that is coming to bear, enabling the AI—it’s an exciting, total transformation that will make a huge difference for patients. We have to manage it correctly, and it’s important we keep the pathologist front and center with their hands on the reins. That will help keep the trust as we go forward.
David Lahm, do you have a comment to share?
David Lahm (Eli Lilly): The momentum of digital pathology reminds me of the book Crossing the Chasm, about the adoption of disruptive technology. It talks about how once you pass the early adopter gap and make it into the mainstream or head of the distribution, the implementation and adoption are not a fait accompli, but it picks up pace and increases the likelihood you’ll make your way into the mainstream in the market. It feels like we’re there now with digital pathology. We’ve built the infrastructure and worked out procedural kinks, and now we’re starting to see broad adoption. It feels like a tipping point.
The fear of losing one’s job, of the machines taking over, never quite goes away. To some degree we’re there, and the younger cohort is more comfortable in taking on enlarged roles for the pathologist running diagnostics and making treatment decisions. Dr. Walk, do you agree?
Dr. Walk (PathAI): Yes, and it connects to a comment I want to make on precision medicine. Just like there’s a transformation on the clinical anatomic pathology side, there’s also a transformation in precision medicine, in pharma drug development. The transformation is about being able to use machine learning to structure biomarker data based on H&E and immunohistochemistry slides in a completely different way.
We’re seeing a strong synergy now between the clinical anatomic pathology space and market and the pharma space and market, whereas in the past these communities rarely interacted with each other. Because of funding challenges on the clinical and academic sides, many of the pathology organizations that never considered partnering or collaborating with pharma are interested in doing so now. Pharma companies are interested in working with community centers even because of this idea of screening with an algorithm for an H&E slide to identify rare, biomarker-defined subsets to accelerate trial enrollment or to identify patients in the real world. We’re seeing this so strongly that we’ve created a mechanism—the Precision Pathology Network—to be the matchmaker between these two communities for specific programs.
As the precision medicine/AI wave progresses, the role of pathologists will be directed more toward diagnostic decisions that directly impact patient care and drug treatment. Precision medicine was the first wave, and there’s the upcoming wave of AI companion diagnostics, like the AstraZeneca TROP2 QCS NMR assay. This is exciting and yet another reason pathologists have to embrace this new mindset and technology so they will be able to correctly identify the right biomarker tests, including AI biomarker tests, for different disease entities and patients so they can provide the best information to the patient and treating physician.
Nathan Buchbinder (Proscia): Eric said it well. In talking with the biopharma community, the first question we get isn’t necessarily about the platform we offer to support routine image-based research and development. Instead, it’s, “Tell us about what your diagnostic lab network looks like.” And when we go to diagnostic labs, it’s, “Tell me more about what opportunities there are to stay in the loop and be engaged in the advances in drug discovery, development, and diagnostics.” This is a big change from where this space was a few years ago.
Dr. Ardon (Memorial Sloan Kettering): Our incoming fellows are consistently enthusiastic about digital pathology and AI. They see these tools as natural extensions of modern pathology and are eager to adopt new technologies. Yet, while pathology departments are embracing these technologies, public understanding remains a challenge. Many patients still do not understand the role of pathology, and AI can create additional concerns if it is not explained properly.
We should emphasize that these technologies are decision-support tools that help pathologists deliver better patient care and allow better access to subspecialty expertise. This is something professional societies and pathologists can explain to the patient community. We could use more publicity and messaging on what we’re doing for patients.
Dr. Williams, is there a point you wish to make?
Dr. Williams (NPIC): One thing that has helped us amplify our voice on an institutional level is that all the images are digital in the tumor board, or multidisciplinary team meetings. When we are with our professional colleagues, we can educate them on exactly what we’re doing. We can discuss in real time whether or not a margin is a real surgical margin.
Anything that shines a light on what we do is fantastic. Anything that makes it accessible for a surgeon, for example, to understand what we’re looking at and what the clinical question is is fantastic. But it’s important, as Orly said, to get the message out about the importance of pathology to patients. Patients and members of the public are the best possible advocates and have the most power with our politicians, funders, and decision-makers to ensure appropriate funds are put into deploying these technologies, not just in the academic medical centers but in smaller practices that are struggling to deliver health care to patients.
Mike Rivers, you know many pathologists and have been in many labs and to many meetings. What is the progression of the specialty that you’ve observed over the past 10 years?
Michael Rivers (Rivers Strategic Advisors): I’ll answer in the context of what we’re discussing here. We’re moving to precision medicine. Highly skilled and trained pathologists are able now to use their skills and training to focus on the highest and best value for the patient, rather than get bogged down in the more mundane or less useful aspects. That’s the opportunity—we can free up pathologists to do what they’re trained to do, give them the tools and support through AI to see new insights and pull new information together. The advent of multimodal decision support, the opportunity to pull in sequencing data and clinical factors with pathology data, is exciting. Pathologists are ready for this, if we could just unlock it and provide them with it.
Nathan Buchbinder, leaders in pathology say it will be essential to the survival of the health system, not just pathology, to have every pathologist working at the top of their license, bringing to bear the highest levels of expertise, experience, and use of technology. What have you been observing at Proscia?
Nathan Buchbinder (Proscia): We’ve been seeing exactly that. One of the values digitization is providing for labs is the capacity to remove bottlenecks and inefficiencies that keep experts from being fully able to leverage their subspecialist knowledge and expertise. We work with groups that are spread out across many states that have subspecialists in oral pathology, for example, who are located hundreds, if not thousands, of miles away. Digitization allows them to immediately access that subspecialist, who, instead of spending 40 or 50 percent of their time on the cases that need to go to them, can now spend 80 to 90 percent of their time on those cases.
It’s also led to a rethink of how the operations of the laboratory work to reroute cases. It’s shifted the thinking about the role AI plays, where AI can serve as a phenomenal screen and QC tool. It can help provide additional insight that elevates the role of every pathologist and serves as another mechanism to ensure cases are being allocated appropriately among pathologists.
Dr. Walk (PathAI): We recently went live at a large health network in Brazil. Previously, the specialists existed in a silo because they were in different parts of Brazil than the general pathologists. Within weeks of going live, they saw a dramatic increase in collaboration and sharing of digital cases live that improved patient care and diagnostic accuracy because there was a democratization of skill and more collaboration between specialists and general pathologists.
Dr. Williams, how would you sum up today’s conversation?
Dr. Williams (NPIC): Absolutely fantastic. It’s wonderful to see we have many of the same themes that I’m sure have been addressed over the years in terms of the basic piece of deployment. We know there are pockets of the world where we’re still doing a hard sell and answering pragmatic questions about getting equipment into hospitals, but look how far we’ve evolved. Now every department is not thinking just about digital pathology but also about AI. How do we as a community best share our stories and experience? There is a wealth of experience between the people on the DPA board and those who attend the DPA meetings—how do we ensure we’re getting all the information to the right people, whether they are taking their first steps or are confident users looking at the next shiny thing? Collaborating and sharing can only lead to a better outcome for all. Digital is a fantastic example of industry working in partnership with academics and health care providers for genuine patient and public benefit.