December 2024—Digital pathology in 2024—what’s the sentiment, the pace, the holdup, the worry? Five people spoke online with CAP TODAY publisher Bob McGonnagle on Oct. 9 about the state of play as the end of the year neared. “It’s an ongoing conversation and foregone conclusion that laboratories will be going digital,” said Lisa-Jean Clifford, president-elect of the Association for Pathology Informatics.
Katie Maloney, in last year’s roundtable on digital pathology and artificial intelligence, we discussed the results of the DeciBio market report on digital and computational pathology, the findings of which were reported in CAP TODAY [https://bit.ly/CT_1023-digipath]. There was a lot of enthusiasm last year for the digital pathology market. What’s the atmosphere now from your perspective?
Katie Maloney, partner, DeciBio: We released the second edition of that report last week. From that research, we have a good sense of what has and hasn’t changed over the last year and a half. There is still excitement and a growing sentiment, which were reflected in the pathologist and lab director interviews we conducted over an 18-month period, around the expectation that this is coming and is inevitable and will play an increasing role in education, even at institutions that haven’t adopted these tools.
It always feels like it’s coming next year. It still hasn’t hit that inflection point where you move into the late majority or even the early majority for the types of tools on the market. Promising AI tools have been developed on the biopharma side, like AstraZeneca’s QCS [quantitative continuous scoring] tool for the TROP2 biomarker. It’s still slow going in terms of adopting AI tools.
Digital pathology and AI are now a foundation of anyone’s plans moving forward. At a recent Clinical Lab 2.0 meeting, it was said to be one of four pillars of establishing the value of laboratories. Lisa-Jean Clifford, how have the atmospherics changed for better or worse in the past year?
Lisa-Jean Clifford, chief operating officer and chief strategy officer, Gestalt Diagnostics, and president-elect and secretary, Association for Pathology Informatics: I agree with Katie. We feel like, “It’s here. No, it’s coming. We’re almost there.” It’s a continual push and expectation, but we are seeing movement. There are several conferences and meetings focused on digital pathology and AI, whether in clinical use, education, research, or all of the above. It’s an ongoing conversation and foregone conclusion that laboratories will be going digital. We’re at about 10 percent adoption, but organizations have this on their plan. Organizations of all types are hiring titles now for digital pathology solution manager or VP of digital pathology implementation.
One hiccup in the past months is that there’s been change in our industry, specifically with two major players. People took a step back and asked what it means. Is digital pathology going away? Are the companies going away? But people have realized that shareholders were driving those changes within those organizations and not the technology itself, so we were able to overcome that hiccup quickly.
Mike Quick, what are your impressions of what you’ve heard? It’s inevitable to have ups and downs in a new field, investors who become disenchanted and others waiting on the sidelines, perhaps ready to pounce. Tell us about the atmospherics around the field today.
Michael Quick, president, Digital Pathology Association, and vice president, research and development and innovation, Hologic: If I were to summarize it in one word for the year, it would be “dynamic.” There’s been a lot of change, positive and interesting. The first half of the year’s mindshare was largely consumed around laboratory-developed test regulation and what it means for digital pathology. There was some hesitation in the field to move forward with that uncertainty. Now there’s a better understanding and clarity about what the regulations mean as we move into 2025.
By and large, the field has taken a huge step forward when you look at regulatory approvals at a pace we had not seen previously, collaboration with the FDA, and a number of new digital pathology societies expanding internationally. We just came back from the European Congress of Pathology but it could have been called a “digital” pathology conference.
A stay of execution of the PAMA cuts has to make everybody in the field feel better about running laboratories and how to plan for the future. Doc de Baca, can you comment on that and give us your impressions of what’s happened in the past year?
M. E. (Doc) de Baca, MD, vice president for medical affairs, Sysmex America; founder of MDPath; hematopathologist, Pacific Pathology Partners, Seattle; member, CAP Board of Governors: The kind of volatility we’re seeing in the AI and digital space of taking two steps forward, one step or half a step back is familiar to us over the arc of history, but we have days that feel less positive than others. The reprieve with PAMA is helpful. The LDT question is monumental. We’re waiting to find out if the final rule changes go through. The Venn diagram of the people who are in the regulatory medical space and those in laboratory medicine in hospital or clinic settings looks like two separate circles. We at the CAP need to start working hard on getting people ready for what creating a union in those circles would look like. Every topic you’ve asked about is its own hurricane, if you will, and we are living in the eye of not one but many storms right now.
There are people who continue to move forward in some spaces despite not knowing what will happen in another space. There are people who see that digital pathology is going to move forward anyway and need to be prepared, so they invest and will figure out how to deal with what happens to the LDT rule, even though that will affect their bottom line. The number of balls that need to be juggled make balancing and feeling like you’re making rational decisions rather difficult. Pathologists tend to be rational people, administrators try to be rational people, and the data sets they use to make decisions tend to be different. How we all play these simultaneous but different games using volatile environments as our data input makes for interesting talks when we go to these meetings.
Michael Rivers, tell us what the past year has seemed like to you, what the atmospherics are, and where you’re settling in.
Michael Rivers, vice president and lifecycle leader of digital pathology, Roche Diagnostics: I remain optimistic and excited about the potential for digital pathology. The past year has been a dynamic year of change. We continue to see tremendous movement with the FDA in terms of the number of systems approved and cleared for use. That gives labs more flexibility in the choices they make. I’ve seen nice depictions of the interoperability and connectedness in those approvals. That’s one gauge of progress. The growth of membership in the Digital Pathology Association and in exhibitors at the Pathology Visions meeting is an indicator too.
We need more clearances on algorithms and more movement in terms of reimbursement. There are still barriers to clear, but there are many exciting advancements and the technology continues to move in a positive direction. I remain positive and Roche continues to invest in this space.
Several people have mentioned how digital pathology and AI are taking off internationally. It’s my perception that many European meetings are offering a richer environment than some of the meetings stateside. I think of infectious disease—no one does a better job than the European Society of Clinical Microbiology and Infectious Diseases, whose meeting attendees are increasingly Americans. Digital pathology is following in the same footsteps. There’s something in the water that’s not just digital pathology; we’ve seen shifts in the entire field. Katie Maloney, does that make sense to you from your perch in England?
Katie Maloney (DeciBio): Yes. There is a lot of innovation, excitement, and adoption of these tools in specific European countries at levels we don’t see in the U.S. There’s a different set of drivers and barriers. There’s variation at the country level, taking a step back from looking at Europe as a single market. There’s a number of good conferences in Europe and a good ecosystem of conferences globally. It’s exciting to see the momentum.
Lisa-Jean Clifford, what are your impressions of this ecosystem of learning and of companies demonstrating products and figuring out how to do business in various parts of the world?
Lisa-Jean Clifford (Gestalt and API): It’s a dynamic space from a vendor perspective but also from an adoption perspective. Each country has its own use cases, rules, and regulatory requirements. At the Association for Pathology Informatics, we’re looking at what it means per country. Even in the Americas, there are different rules and uses in Canada, the U.S., and South America.
Being able to drive the education, knowledge, and conversation forward at these meetings and understanding which meetings are growing and are meaningful is important, and it changes annually. Even within the year, you can see changes in the associations and organizations, from membership growing to vendors being more present and not only demonstrating but also speaking and having their customers speak about their adoption process.
Judging by the studies being published in reputable journals, people are grappling with what the products are offering and are weighing and studying them seriously. Doc de Baca, do you agree?
Dr. de Baca (Sysmex): I do. During the arc of my career, there was expansion and contraction of the immunohistochemistry markets and companies that exploded and then there were mergers. Now we’re down to fewer numbers. As things become integrated into the practice, they become just pathology. It’s another tool. When we lose the moniker “digital” on digital pathology, or “AI pathology” turns into “AI” or “AI” is dropped completely, we understand that the novel technology has arrived—it has been included in the status quo of the practice of pathology.
For the person who comes up with an emerging technology, that’s their hammer, so everything looks like a nail. You use that same Gaussian curve of first adopters to deniers and everyone in the middle. People start to look at things with different eyes, and you get questions about capacity, capability, specificity, error, ethical implementations, and how it works with other things we do. That’s where we’re starting to see shifts in the applications of AI for pathology specifically and for medicine in general. We’ve been looking at relatively large hammers and nails, but those are blunt tools. Now we’re trying to understand when to use what tool. Which hammer do you use in this case, what size nail, what angle do you use it for? When would it be wrong to use it? Those are the questions people are starting to grapple with.
In the AI space, many people understand computation and data, but many people who make AI tools don’t understand the context in which these data arise and how the use of these data, if not absolutely correct, can have huge downstream effects on lives, on epidemiologic decision-making at government or insurer levels, or on population health. There’s a lot to think about.
A participant in another AI roundtable last year said none of this will work unless the expert pathologist is collaborating with the expert people in digital pathology and artificial intelligence. Mike Quick, can you comment on that?
Michael Quick (Hologic and DPA): That collaboration is necessary for the field to move forward. There’s often hype and concern about AI replacing the role of the pathologist. How will that change over time? In the past year we feel more solid that it’s an integrated role. Bringing AI tools into clinical practice makes the pathologist or cytologist better at what they do.
Michael Rivers, can you comment on that same question?
Michael Rivers (Roche): AI image analysis will fundamentally change the practice of pathology. It will be central and critical to what is done. The pathologist will continue to own the diagnosis and pull the information together in a meaningful way for the patient.
Exciting tools are being developed and data are being presented at conferences, but getting them into a clinical workflow, into the hands of a pathologist to use them on a patient, is another story. We’ve been focusing on enabling the integration of AI tools into an efficient clinical workflow so they can be used effectively, so a pathologist doesn’t have to serve whole slide images to multiple platforms to get information and data. Providing an integrated solution for pathologists to allow them to work in an efficient, meaningful way is important. We’re bullish on the future of AI. It is going to be part of pathology and the pathologist will remain central to the process.
Lisa-Jean Clifford, pathologists have anxiety about their future, and it’s easy to imagine how anyone would be anxious without clarity on how AI tools can be integrated into a workflow. Can you speak to that question?
Lisa-Jean Clifford (Gestalt and API): There’s two parts to that. The first is the development of the algorithm. In many cases nonmedical people—academics or programmers—are designing the specifications and the algorithm. Algorithm development must involve pathologists, with continual input and feedback.
The second part is the integration of the algorithm for clinical, research, or educational use. Once the algorithm is developed, how do you put it into play in a workflow that is real, usable, valuable, and validated by pathologists? Pathologists must be involved on both ends of the spectrum. Integration in a single platform and single viewer is also key because requiring a pathologist to go to multiple windows, screens, or viewers to look at outcomes from algorithms across multiple vendors is fraught with possibilities for human error.
Katie Maloney, can you comment on this issue? I’m sure this must arise in the interviews you do for your reports.
Katie Maloney (DeciBio): It’s integral to driving the industry forward. It is important not only from a clinical perspective but also for growing the market or expanding into new labs that don’t have the expertise, time, or resources to work with multiple platforms and jump between windows. There might be tolerance for that in research institutions where they have time or want that flexibility, but an integrated workflow will be necessary to see widespread adoption of these tools across clinical labs.
Everyone is floating in an ocean of Epic EHR and Beaker installations, certainly in the United States. IT systems are a backbone of success here. What do your interviewees say about that?
Katie Maloney (DeciBio): IT barriers are a big reason we see hospital labs with a scanner sitting in a corner but that haven’t fully digitized their workflow. The IT lift needed to integrate it with the rest of what’s going on in the lab and institution is significant. If you don’t have buy-in across the board, it’s going to take a long time for it to happen. We’ve heard from interviewees that the adoption of these tools, such as Epic Beaker, can help streamline these things because of the integration between the EHR and LIS, but there’s still the need for the IMS [image management system] to play within that and to make sure the data flow is seamless.
Lisa-Jean Clifford, laboratory IT can’t be separated from digital pathology and AI. Can you comment on that?
Lisa-Jean Clifford (Gestalt and API): Being able to integrate technology is key. Most modern applications have the ability to integrate. Whether the vendors are open to having those conversations and play nice with each other is another matter.
With some legacy systems, it is more difficult to have deeper integration or a true bidirectional, real-time integration. There are ways around that, especially with people who are more focused on having the technical background to drive interoperability forward. There’s middleware. There are many ways to solve those problems. A true bidirectional, real-time data exchange is key not only to having a streamlined and efficient workflow but also to supporting patient safety, which is my focus.
Health care IT departments have many of the same stressors that pathology departments have. Budgets are tight, and there are great shortages of people. Can you comment on this, Doc de Baca?
Dr. de Baca (Sysmex): We talk about this being new and inevitable, and that’s true. Pathologists are good at generating their own anxiety—that’s also true. Consider that we are three percent of the medical dollar but we create between 70 and 80 percent of the actionable data in medicine. We are anxious in part because we realize that since we’re not seen, because we’re not big in the budget, we can take hits that will have huge impacts. And since we’re not in everyone’s face, there’s not an understanding of how important what we do is.
We were talking about pathologists needing to be integral in the development of AI tools. I’d like to add to that: It is AI tools that use any kind of laboratory data. People who are making AI tools for cardiology, for example, are using laboratory data. Pathologists understand the test and the result that are used and then look at things and say, “You can use these data together or you cannot use these data together.” If you’re talking about decisions that will lead people into surgery or will give them drugs that will anticoagulate them or change their blood chemistry forever, that will have effects on morbidity and mortality, and we need to be part of that. But if you’re the tiny tail on the back of the elephant, are you noticed? Even though your role in the ecosystem is huge, will you be noticed?
Then again, we need tools that are integrated so we can get our work done, and we have trouble finding IS support for that, for the reasons you state. They’re trying to keep cyberattackers away and triage other demands. The pathology department is trying to get the data out to physicians. Those things must happen with not enough resources. At the same time, we need to implement digital pathology so we can integrate with AI tools that will be helpful. We need to have digitization to do the AI work that we know is possible. All of that requires time, money, and other resources.
Then we have the government asking for interoperability and real-world evidence. Agencies are asking for different things, all of which require our time, expertise, and money. I’m not new to the game of innovation, creativity, and thinking about where we can go, but I am new to the game of thinking about money. We need to do all these things; I get it. I agree with the requests from the government and with many of the things they see as the end goal. But if you’re only three percent of the medical dollar and you’re creating 70 to 75 percent of the goods, someone needs to help us pony up even more of those goods.
The CAP’s role in advocating for pathologists cannot be understated. If we can educate people from around the house of medicine to understand how pivotal this is and how much they depend on us so they can help advocate for money for these projects, so they can get the necessary data, everyone would be better served. Right now it feels like we’re fighting as hard as we can and there are few of us. The implications of the fight and our effect in that fight will be long felt.
The Project Santa Fe Foundation with its Clinical Lab 2.0 as well as many vendors and health systems in our country are recognizing some outline of this major problem. Michael Rivers, this must be recognized internally within Roche Diagnostics, correct?
Michael Rivers (Roche): Yes. We are convinced that the pathology lab will go fully digital, and if we’re talking about a fully digital workflow, it has to be integrated. It requires bringing together many stakeholders and complex systems that aren’t necessarily made to talk with each other, but solving that complexity and giving options to customers is a key role for us all.
We also talk about interoperability. It’s critical that we give labs the flexibility to put together a solution and give them options to be able to digitize their workflow in a way that’s meaningful to them. There are many components and a lot of complexity, and it’s a barrier to adoption and scaling the adoption. But there’s going to be enough energy from the community at large to overcome this.
A tremendous role for organizations like the Digital Pathology Association and Association for Pathology Informatics is to bring groups together and talk constructively about how to minimize the pain for labs as they’re trying to adopt these new technologies. That will be crucial as we go forward, as will introducing things like DICOM. We need integration at the scanner level as well as at the IMS level and so forth.
Roche has built an information solutions division focused on these types of digital solutions, not just for digital pathology but across the health care continuum, and on the integration that needs to happen to enable multimodal decision support.
Mike Quick, I will close this conversation by asking when Epic will be invited to take a seat on the board of the DPA. I ask only because that’s how important IT is, as I see it. What would you most like to see happen as you’re ending your term as president of the DPA?
Michael Quick (Hologic and DPA): We have a standing invitation for any LIS vendor to be a part of DPA. In many ways, it’s the missing link that will move us forward. The LIS business model is different from most partners we have within DPA. There is much more customization, rather than standardization. Once you’ve studied one lab’s workflow, you’ve studied one lab’s workflow.
DPA met with the FDA this summer to discuss interoperability and collaboration. It was clear in that conversation that we had three entities, between the CAP, DPA, and FDA, and the need for understanding where the governance lies around interoperability, who sets those standards, and how we move that field forward, particularly around things like DICOM and validations. After our meeting, we walked away saying we have to get this right and have clear roles and responsibilities and a path forward so we can move the interoperability conversation ahead. That will then provide the backdrop of what we need for the IT infrastructure that’s necessary to advance digital pathology.