Welcome to PathPulse: Pathology Innovators in Action. Today, we’re honored to host Dr. Liron Pantanowitz, the Maud L. Menten Professor and Chair of the Department of Pathology at UPMC and the University of Pittsburgh. With a distinguished career spanning over two decades, Dr. Pantanowitz is a pioneer in digital and computational pathology. From co-founding the Journal of Pathology Informatics to advancing groundbreaking guidelines in whole slide imaging and AI applications, his contributions have shaped the field. Beyond his informatics expertise, he’s a globally recognized cytopathologist and past president of both the Digital Pathology Association and the American Society of Cytopathology. He has also been awarded recognition for multiple years on The Pathologist Power List. Get ready for an insightful conversation about shifting mindsets toward digital pathology with one of pathology’s leading voices.
Listen here:
Transcript
James Thackeray: Welcome to the PathPulse: Pathology Innovators in Action podcast. This podcast showcases pioneers, innovators, forward thinkers, and individuals within the digital pathology arena that are making a difference in day-to-day use in this industry. I’m James Thackery, I sit on the executive board for the Digital Diagnostic Summit. And in today’s episode, we will be exploring the topic of shifting mindsets towards digital pathology with Dr. Liron Pantanowitz.
Dr. Pantanowitz was recently highlighted on the Pathologist 2024 Power List. So congratulations on that. I just want to be nominated for such a list in my own family, but I can’t even do that. And you’re in all of pathology. So that’s a really high honor, and we’re really excited to have you on the podcast today. Why don’t we start with just, if you don’t mind, just giving a little bit of your background and then more of what you’re currently working on and just kind of give us that perspective. That’d be great.
Liron Pantanowitz: Yeah. Hi, everyone. Thanks for listening to the podcast. And I really appreciate being invited to share my thoughts about pathologists’ mindset when it comes to going digital. I’m the current Chair of Pathology at the University of Pittsburgh at the University of Pittsburgh Medical Center. And I’ve always seen myself as someone who’s willing to try new technology, you know, and be at the cutting edge. And that’s what I do as the Chair of Pathology.
I am a proponent of digital pathology. This particular institution has a lot of experience in digital pathology, some of it being positive, some of it being negative, but we have learned through our mistakes. And now that I’m here as the Chair of Pathology, I have three priorities as the Chair for my mission. And the first is to take care of people. The second is to improve the process, which helps the people do their job and adds value. And then the third is to apply technology to both of those. Technology to assist with people so that it makes their job easier. This is the place they want to do it so it helps with retention. And then obviously the technology can improve workflow if implemented well.
And so we are going full force at prospectively trying to digitize and connect all of our facilities, capitalize on that infrastructure, and at the same time, we are strengthening our computational pathology group here, and we’re basically developing AI, validating AI, and deploying AI.
James Thackeray: That’s beautiful. Thank you for the introduction. And again, thanks for being here with us. I love the vision as the Chair of Pathology and those three objectives. That’s actually beautiful: “Take care of people.” Then you’ve got the process side of it and then you add technology in to help obviously both one and two. So wonderful. Thank you for that background.
I think, as a way of background for you, in some of the previous podcasts that we’ve had, we’ve talked a lot about barriers to entry in digital pathology. And I think it’s not short of barriers and I think we recognize it. It’s always a balancing act with a technology like digital pathology in general that people know fundamentally is going to be part of the market. But the timing of that depends on overcoming these barriers, of course. And so we’ve talked about major barriers like the complexity of integration, the cost of integration, the lack of reimbursement, just the different barriers that are kind of more technical than anything else. I was really interested in a recent interview that you had with The Pathologist where you talked about the mindset of a pathologist.
And while I think we fundamentally know that that’s part of this, sometimes, and I’m just speaking as a technology provider, we assume that if the technology is good enough, that either adoption will come naturally or that it will be forced upon an organization. I think in theory, I think a lot of people believe that, but I think we don’t spend enough time getting to the actual users of the technology, the pathologist.
So I’d love to just start by asking you kind of the barrier from a pathologist mindset and how some of the things that you’ve done in that regard when you work with your team.
Liron Pantanowitz: You’re right. People have asked me a decade ago, five years ago, and even recently, you know, what are the main barriers? Why are people not going fully digital? Industries have invested, venture capitalists have invested in this technology, pathologists seem eager, and trainees coming into our field all seem eager. But then when you actually take an inventory of how many people have used scanners they’ve purchased to go digital, how many have actually purchased scanners, you know, it’s staggering how few people are, even though there is adoption worldwide, but it’s very low to what we had anticipated back then and now.
And you’re right when people look at what are the main barriers. And they often say that, well, these are the barriers. But when you look at those barriers, we’ve overcome them. So it always has been things like regulations. But that’s not true anymore. We figured out with the FDA, who are well informed now, they are proving a lot of these products, and they’re pretty good devices now even to work from.
People have spoken about the cost, but there’s a good return on investment and there are even billing codes coming around and for some practices it’s vital to actually go digital for their survival, to be able to connect regionally. Yes, people have complained about interoperability, but more and more of the industry partners are agreeing to DICOM and are working together. And if they haven’t connected you, they will work with you to do that. So that’s not really a barrier either.
But I’ve heard pathologists complain a lot about the workflow, saying things like “This is terrible, It’s too destructive.” But the minute they figure out that the workflow helps them work from home, well, you know, that’s a different story. So that cannot be true either. Which leaves us with the final barrier, which is the topic of this podcast, which is their mindset.
The Mindset of Pathologists is Often the Biggest Barrier to Digital Pathology
It’s the mindset of pathologists that I often think is the biggest barrier, which is what I told The Pathologist magazine, that we need to change the mindset of pathologists. Yes, part of the problem is that they were never trained to go digital. So we have to address our education curricula and start training people and get the right mindset. But what do we do? Because the majority of pathologists that are practicing today were never trained in a digital world. And so we have to address that with them. When you actually talk to them one-on-one and you explain to them, and I often, you know, bring this conversation to a real-world experience. If I ask them about their world around them, how do you listen to your music? How do you communicate with your family, family abroad, etc. It’s all digital. They listen to your music.
They use email, they text, but everything in their world outside of work is digital, everything at work is not digital. I like to use the mobile phone analogy and equate that to their microscope. And I’ve had this conversation many times with pathologists, and then sort of the penny drops, and they buy in. So I tell them that it is true that my mobile phone, my iPhone, may not provide as clear a reception when I’m talking to someone if I’m in an area where I don’t have good reception, if I’m going through a tunnel, yeah, I lose signal sometimes. But I would not give up the portability of my iPhone to have to go to my office to use my landline every time I have to make a telephone call. It’s true that using my landline will give me a very clear conversation, but I always have to be in my office to do that.
And it’s not just the portability. What’s more important to me on my iPhone is not to have conversations but all the applications that I have on it, right? And so when you explain that to pathologists, that one, you gain the portability, yes, at the expense sometimes of, you know, having a perfect conversation or a perfect microscopic image, but it’s all the applications that you will have in the future that you can’t give up now. And so they get that.
And so I think that’s important to have those real-world conversations. And the other thing, James, if you don’t mind me saying, a lot of pathologists have been terrified about digital pathology. And honestly, that’s because of other people’s horror stories and miscommunications. And one of the things that always upsets me is that when people talk about going digital, a lot of people out there give presentations, write articles, but they don’t really do it. It’s one thing to talk the talk, but it’s another to walk the walk. And, in my opinion, that leads to maybe some fabrication and it doesn’t account for a learning curve and so forth. And so you don’t tell the truth, and then it’s much harder, and people don’t know how to connect the dots. We need to tell people the truth and we need to tell them how to connect the dots. And so that’s where change management becomes important and helps people overcome the mindset.
If you remember, I don’t know if you know Peter Drucker, he’s that famous businessman who came up with this quote that clinical informaticists love to use all the time, which is a culture eats strategy for breakfasts. And so, you know, you can have this amazing strategy, and we’re going to deploy it. And this is the go live date. This is when you’ll be trained, and we’ll have the CAP validation. But what about the culture there, like engaging the employees to motivate them and allowing some flexibility, and getting them excited about innovation. That’s the culture you have to create and I often see that doesn’t happen.
James Thackeray: That’s such great insight. I love that. I love it so much. I think, you know, again, and I, I’ll speak from the perspective of the technology industry. I may be too bold, but I believe we often underestimate how good a pathologist is under the microscope, right? I mean, their entire careers, depending on where they are in their career, have been basically figuring out the way to be as efficient as they can be with the current system that they’re in.
So, that microscope is the way they do it. And I think that change management starts there. Creating from a technology perspective, creating a product that actually gives the pathologist an experience that is a good experience, right? I think it’s hard when we try to push anybody in any industry into a more clunky process, maybe to get some remote usability or something like that.
So I think from the technology side, one of the things that we’re always focused on, and I think you’ve brought out, is that that experience for the pathologist has to be good, or it’s tough to get adoption.
Improving Change Management
Liron Pantanowitz: Yeah, you know, there’s so many more nuances to going digital that people either experience on their own or they’re not told about this, and it just creates a negative situation and then lands up with passive-aggressive behavior.
So I’ll give you some examples. And these, again, fall under this change management, like really taking care of the people when you’re doing the adoption. People are going to change the way they practice for sure.
You know, if you’ve trained in a digital environment, that’s one thing, but if you haven’t, you’re going to change a lot of things. So, for example, I, like a lot of pathologists, and I’ll say so am I, we are addicted to paper sometimes. But you have to go paperless when you go fully digital. But I like having some paper around to remind me about things when I get disrupted, or I can go back to that paper and make some notes on it, etc.
But once you go fully digital, you’re in a paperless environment. And it’s one thing to go slide-less, but it’s another thing to go paperless. And I’ve also seen that pathologists are not used to a “pull” environment. We’re used to a very “push” environment. What I mean is our workload. So when I have a tray of slides in my mailbox, I know that’s my trigger to start working and how much work I have to do and I know that I’d better prioritize the next four hours to signing out, et cetera. But in a digital world, I don’t see any of that.
I need to go pull that information for myself and so forth. And so this makes it very new, as does the ergonomics. And you’re right, pathologists love their microscopes. They have great muscle memory and they’re very quick at doing that. And so when you change the ergonomics to computer screens, and then you tell them, do you want it vertical or landscape? Now you’re changing things and just maybe put it in place and tell them and give them great input devices. So we sort of lead them all to that.
And then if we make false promises, that causes problems. I’m all about telling the truth and telling people that there are problems with technology. They should know that sometimes the system is going to go down, or sometimes it’s going to freeze, etc. And so they should just be aware of that, rather than think that it will be perfect every time for every case, which is not true.
James Thackeray: I love that. I think that addressing education needs up front and what the pros and cons—the real cons, is important. Letting them know some of the things that will happen without undervaluing all the good that it can do. I really love your mobile versus landline analogy; I think that’s very appropriate. So sticking with this kind of change management topic, I was invited to attend and sat on a panel for a large organization, a big healthcare company, that is trying to figure out their AI strategy.
And I was fascinated by the discussion because often, again I’m going back to something I said earlier, there’s the idea that once AI is really integrated in, it will force us out—nobody will have a choice. That may or may not be true, I have no idea.
And maybe that’s not so far down the road, but I wanted to get your perspective more on AI and I know there’s different phases, but let’s just talk about diagnostic AI and how that plays into the pathologist mindset and how you deal with that. Is that looked at as, okay, my career could be short-lived? How do you deal with AI in this whole mindset situation?
How Diagnostic AI Affects the Pathologist Mindset
Liron Pantanowitz: Yeah, so there’s those pathologists that have not used AI yet, but have seen it and are anticipating using it. And one thing is there’s certainly a lot of hype around that, which I, again, I wish we wouldn’t hype it up, but rather tell people that these are the things that are required for deployment. We’re not quite sure how to validate it, but this is what we’re thinking. And that builds trust.
So I think one of the things pathologists are concerned more about is trusting the AI because, at the end of the day, it’s their name that’s going at the end of the report. Whether there’s AI completely signing out the case, there’s still a medical director overseeing the lab, or they’re going to actually intervene and look at the AI output and then agree or disagree or modify it, it’s still their name on the report. So, building up that trust will take time.
The other thing is, at our institution, we have two AI algorithms running in anatomical pathology every day, and we’re putting in a third one now. And when I go back and I talk to the pathologists who are using the AI every day, there are some concerns. It hasn’t necessarily been grandiose and made life great, because of some of the ways we’ve designed the AI.
Initially, we had instituted AI after the pathologist looks at the case because they wanted to build some sort of a learning curve and trust it and use it more as a QC measure, et cetera. But when they realized that, wow, this could actually help with triage, could actually expedite what they do and get them home quicker, we switched from a second to a first read and put the AI upfront. And that was great.
But we didn’t take care of other things. For example, I wanted to make sure that the workstation and the environment and how quickly we transmit digital images, all of that provides a great experience too. It’s like putting a Ferrari engine in a little VW. That’s what we did to them. Give them a Ferrari, but we wouldn’t let them get on the Autobahn and drive. That doesn’t work well.
And we didn’t fully integrate the AI algorithms into their workflow. It still remained a bit of a standalone system. And even today, where vendors who are making AI models will tell you that they’re compatible with certain image management systems, you still don’t get all the bells and whistles—you lose some of the functionality. So, it’s not really fully interoperable.
And so pathologists know that. And you know, it’s extra work, extra systems, extra clicks. And we’re telling them you should use AI because it’s going to make you efficient and automated. And the truth is that they don’t experience that because of all of these extra, you know, problems that they find from technology. I think, yeah, in reality, it’s not as great as predicted.
You know, I always think of when you see an advertisement to go to this beautiful island and there’s this beautiful beach and they tell you, you should come to this, know, Club Med, for example.
And then when you show up there, there’s hundreds of people on the beach and there isn’t even a spot for you to sit down on the beach, right? So it’s like, wow, that’s not what they advertise. And that’s kind of what we are selling to some pathologists. Go digital and then you can add AI. That’s the justification. And then they showed up and it was not like what was advertised.
James Thackeray: Yeah, I love the statement you used: “walk the walk, talk the talk.” Because I think, a lot of times you’re right. It is coming from those that are not utilizing digital pathology day to day, where a lot of this comes from.
You know, we’ve learned a lot because I think when we talk about AI in general, most people are thinking diagnostic AI, but I think from a technology standpoint, if we’re thinking about the pathologist experience, the first place we start is actually AI built in the viewer that’s going to make the pathologist more efficient.
And you talked about triage or the ability to use even, I guess, diagnostic AI to triage is one thing, but anything that actually makes or helps the pathologist be more efficient. That’s certainly where we’ve had a lot of focus. I think diagnostic AI in general is coming fast, but it’s going to take longer to integrate into a day-to-day process.
I just think from the technology perspective, we should focus on almost the workflow AI that makes someone more efficient. At Lumea, we’ve focused on prostate because that’s what we started with and we wanted to make that workflow for prostate so that actually a pathologist without even diagnostic AI is more efficient than they would be under the microscope, which we’ve been able to do. Now it’s like, okay, now do we do it in GI and how do we do it in derm.
And it almost is specialty specific in some regard. And I don’t know, I guess I wanted to get your feedback too on that. Like, how do you see it evolving as far as a platform goes? Do you see it going into more specialty? I mean, because there’s so many nuances in specialty or like we use this foundational model that can try to accommodate every specialty. That’s a little off topic, but I’m curious.
Specialty vs Broad AI
Liron Pantanowitz: Yeah. Okay. No, well, so talking about AI coupled with digital pathology, there are many ways you can think about it. I’m going to break it up in terms of functionality. Well, how will it help a pathologist and how will it help a pathology lab and therefore how will it help, you know, healthcare in general, reduce costs and better integrate systems.
So the way I think about it is, and just so that you know, I wish more people were making AI models to do all of these, but they’re not. So for those of you listening to our podcast, if there isn’t an AI model, please go and make one for us. So James, number one is we’ve realized by going digital and especially ramping up a digital pathology operation with high volume and high throughput, that there are so many images that the QC measures we had in place before to check the quality of the images, we cannot do manually. It’s sort of counterproductive. You’re trying to automate and make it efficient. You cannot have humans check all those images.
We need AI to do those quality checks for us and send back the image for a rescan or whatever the issue is, get it fixed before it even goes to the end user or gets analyzed by another AI algorithm. And I’ve seen some nifty products. I’ve seen AI vendors put that into their whole slide scanners and sort of edge computing right there and it takes care of it.
And others that have standalone software that can do it in an image management system before it even goes forward to whatever your downstream uses. So that’s number one.
Then there’s the traditional AI that we’re talking about, which is diagnostic. Can you help a pathologist make that diagnosis either as a first read before they even get the case or after they get the case? And there are a lot of uses before they get the case. Triage it, work up the case, prepare the report, for example. Or afterwards, you know, do a quality check, help them with the differential and so forth.
But we’re now, especially as we’re getting into generative AI, we’re able to do much better predictive analytics. And so, while we haven’t seen too many of those algorithms, they’ve been demonstrated to be feasible. People have published it, proof of concept. So, you know, look, here’s an H&E image. This is what the stain will look like if you actually did the stain. Here’s the H&E image.
And from this phenotype, we’ll predict a particular molecular genotype, so maybe you don’t even have to do the molecular testing, or we can triage that there’s a low likelihood there’s a genetic abnormality. So don’t spend all the money on this small piece of tissue that you want to conserve for other reasons. But no one is really doing AI for the background QA stuff that we do in labs all the time. We spend a lot of time on trying to get data out of these siloed information systems, trying to match them up, and we have a lot of people doing that work.
So examples would be cytology-histology correlation. If someone has a PAP test and then they have a follow-up biopsy of the cervix, who does that correlation? It’s up to the pathologist to do that. If someone has a frozen section and then they have a follow-up surgical resection, who has to do the frozen versus permanent correlation?
It’s a human that has to do that. If someone has a mammogram or even a prostate MRI, and then there’s a biopsy that follows, who does the path red correlation? It’s a whole lot of humans. And now you’re putting two different departments. Oh my goodness. Radiology, and pathology, to even get that coordinated. And believe it or not, some of this is mandatory for these centers to remain accredited. So I would love for people to be building those AI algorithms to take care of that important, but, you know, very human, you know, heavy tasks to undertake.
James Thackeray: So insightful, and that all kind of goes under that. I know we kind of categorize everything under workflow AI, but that’s so good. And fortunately, I think we do have a lot of very good AI partners that I hope listen to our podcast. They’re good partners for us. If they’re good, they should be listening and they should be listening to you because these are super good insights. I’m sure you may have some pretty good AI partners reaching out. I’m sure they’re already doing that regularly with you anyway, but great insight.
Conclusion: A Message to Pathologists
Okay, in summary, if you were to just give a closing statement to pathologists in general, maybe we should narrow it down. Like a pathologist that’s in the middle of their career, not at the end and not at the beginning, as a plug for what going digital potentially could do. And I know that this is kind of a blanket statement and that each, if you’re in a hospital setting versus, you know, outside of the hospital setting, there’s all sorts of different scenarios.
But what would be your pitch that you would give them on why digital should be in, you know, in their minds and in the future?
Liron Pantanowitz: Yeah, you know, I would tell them and if they’re listening, I’ll tell them now that medicine changes. Everything I learned in med school I probably do not need anymore because I’ve had to continually learn and change the way that I’ve practiced, the knowledge that I’ve had to acquire, and how I have to apply that every day. And we’ve reached a point now where there’s just so much information, it’s so complicated, and a lot of the stuff that we have to do is something that cannot be done, I’m sorry to say, around the microscope.
My microscope is no longer a crystal ball that I can predict the things I have to get done. And my microscope is not the place to quantitatively do stuff that really patient management depends on a lot. There’s a lot of criticism about HER2. How low can you go?
But they want us to go very low because it makes a difference for even one patient. We’re in this era of precision care, so we need to be precise about that. And I don’t think we can do that anymore.
So, the point is we need to adapt. We’ve always adapted as physicians and pathologists to the change in medicine. And so the best way to do that today is with technology and tools.
And that’s what Darwin basically said to us: it’s not the most intelligent of the species, right? Or the strongest or the toughest. It’s the one that’s going to be willing to adapt and change and undergo evolution. That’s the species that will survive.
So you can be the pathologist that you think you’re smart enough that you don’t need it, or you can be the pathologist that’s, you know, tough and brutal and walk into your frozen section room and want to do it all by hand, or you can be willing to adapt and undergo some evolution, and you’ll be the pathologist that will most likely be able to survive in the future. So that’s the message I would tell folks.
James Thackeray: So good. What a great summary. Well, thank you, Dr. Pantanowitz. You’re so great to come on our small but growing rapidly podcast and be a part of this. We’ve been following you. I shouldn’t say we because Dr. Leavitt and Lumea, and others around have been following you for some time and are grateful for all the work that you’re doing in this space.
So, thanks for joining us, and we hope to have you on a future podcast. If not, at the very minimum, we need to have you at our digital diagnostic summit in Park City, Utah, in September. But we’ll talk about that after this podcast.
Liron Pantanowitz: Goodbye, everyone. Thanks for listening to us.
