In this episode of PathPulse: Pathology Innovators in Action, James Thackeray, head of the Digital Diagnostic Summit board, interviews Andy Ivie, a digital pathology software expert. Andy Ivie started his career in digital pathology development in 2013, working as Director of Software Research & Development for Leavitt Medical, Inc. The company was later renamed Lumea, and he now serves as its Chief Technology Officer.
Tune in to listen to his insights on significant changes in the last 12 years with digital pathology, the rationale of going digital (other than AI), artificial intelligence, and recent exciting developments in digital pathology — all from a behind-the-scenes software perspective.
James and Andy also discuss Andy’s featured article in the Digital Diagnostic Digest: Top 5 Reasons to Adopt Digital Pathology Today (Spoiler: AI Isn’t One of Them). Subscribe to the free digest today to read the full article and more.
Listen Here:
Watch this episode on YouTube.
Transcript:
James Thackeray: Welcome to the PathPulse: Pathology Innovators in Action podcast. This podcast showcases the pioneers, innovators, and movers and shakers who are transforming digital pathology from theory into practical, implemented, and viable solutions. I am James Thackery, Chairman of the Digital Diagnostic Summit and Chief Commercial Officer at Lumea. And in today’s episode, we will be exploring leading reasons to choose digital pathology aside from artificial intelligence.
So our guest today is Andy Ivie. I’m really excited to have Andy on the show, not only a colleague, but a friend. Andy, why don’t you introduce yourself and give us a little bit of your background.
Andy Ivie: Hi, James. Been a long time.
I started out as a software developer with a focus on security and web applications. And then I met Dr. Matt Leavitt and Dr. Mark Evans through a mutual acquaintance, a friend, and we started talking about things that we could do in the healthcare space. And I got to understand Matt, Dr. Leavitt’s, vision of what he wanted to do.
We built a prototype that turned into what is now Lumea’s Biopsy Link LIS and then Matt bought a very early Aperio scanner that was as big as a room, Kind of reminds me of the early days of computers where they show a forklift bringing your computer in and now we have our phones, kind of similar to the evolution of digital pathology scanners and technology in general.
But we start working with that original scanner, building a web-based viewer that we could use for looking at the whole slide images remotely, and then we brought in a team of people who were focused on AI, the web development, the whole slide image pipeline and just digital pathology in general.
And then kind of where we are today. So that journey’s taken about 12 years to where we are now. A lot of the technologies we’re using today are still based off those kind of original prototypes and early designs that myself and the team put together. But so we’ve got, we’ve had a lot of experience with digital pathology and these whole slide image management systems. But it’s also changing very fast even today. So trying to keep up and get to the forefront of it.
Major Changes in the Last 12 Years of Digital Pathology
James Thackeray: I love it. 12 years. And I was thinking, 12 years in digital pathology – specifically around the software that’s associated with digital pathology – that’s a long time. It’s got to be as long as almost anyone. As much as you look like you’re – 12, it’s crazy that you’ve been doing this that long.
Tell me just, I guess high level, what are some of the major changes you’ve seen in that 12 year evolution of where we are today and like you said, I know it’s changing fast even today but what are some of the highlights if you were to just take us, you know through that journey a little bit. What are some of the highlights that you would emphasize during these last 12 years?
Andy Ivie: Yeah, I mean there’s so many. Maybe I’ll start on the scanner side where the scanners have gotten smaller and more reliable.
Our first scanners, you know, it was a battle to keep them running and they took up a ton of space in the lab. Now there’s scanners that fit on a desk very compactly that can do small numbers of slides and then even the scanners that can do hundreds of slides in a run are much smaller than they used to be and are definitely much more reliable even with all the robotics that’s going on to manage the set of slide images that are being made. So that’s a big one.
The AI side has been interesting to develop over the last decade or so. When we first got started, it was kind of pre-Paige, pre- any of these diagnostic AI algorithms that we have today, which are all very interesting projects and have, I think, a bright future. I think we’ve also, at Lumea, focused on the workflow AI and built processes to make the pathologists more efficient, to help them be a better diagnostician, being able to see more of the slide at one time, being able to navigate complex slides and perform calculations and documentation really efficiently.
And then, you know, in general, the industry has just matured and grown so much, you know, there’s dozens of viewers, there’s, you know, probably dozens of scanners, you know, especially – there’s maybe half dozen really big players and then a lot of new entries who are targeting specific markets. So, yeah, it’s just much more mature, much less of the like old scanner in the corner or in the basement.
And now, you know, these machines that are very efficient and these software products that are very refined, that are driving high volume workflows and really replacing a lot of the glass process for most anatomic pathology.
James Thackeray: That’s great. What about – a lot of it – it seemed like early on, and there still is, I don’t want to minimize it, but there was a lot of complexity around tying all these things together. So kind of the integration side of it. What integration and storage and uploading and what have you seen as an evolution in that? I know intimately what your team has figured out, but maybe just highlight on a few of those, that evolution.
Andy Ivie: Yeah. And there’s still some of this today, like most technologies, you kind of get something that maybe works for a particular use case and stick with it, even if the industry has advanced beyond it. But most of the early viewers and use cases were built around having the whole slide image available on the computer that you’re looking at it on.
So your viewer is actually just an application that’s running on your computer looking at a local directory or one that’s connected over a local network at image files that are, you know, a gigabyte plus and a lot was done to move those files around a hospital or large group network to try and make that efficient. But you’re still kind of practicing digital pathology at your desk on your local network and using, a lot of times, a desktop computer to view those images.
Where now most digital pathology, I feel like, is done through cloud-based web applications that are streaming sections of the image over the internet very efficiently, which allows pathologists to go in and work from their desk if they want to, or are able to, and also be very efficient working from home or on the road if that better fits their work that day, or maybe even full-time. So that’s a pretty big difference in the way the expectations of a pathologist who’s operating digitally.
No longer are they necessarily tethered to a desk. If they have the right solution, they can go anywhere.
Rationale for Going Digital Other Than AI Applications
James Thackeray: I love it. I recently read an article that you wrote about, you know, the rationale of going digital, but maybe it’s not, you know, what are the other factors of going digital outside of artificial intelligence use? Because I think –
I do think when we think of digital pathology, often a very high primary driver is to get to artificial intelligence at some point in this evolution, right? But I loved – you highlighted a number of things outside of that that I think are equally as important that maybe sometimes get forgotten, you know, in this quest to go digital. Tell us about that. Tell us about some of that.
Andy Ivie: Sure. Yeah, AI obviously has a very strong voice in the digital pathology community right now, and is doing some really interesting things. At the same time, there are people who are hesitant about incorporating AI into their workflow for many reasons and are kind of lumping that together with maybe a hesitancy to go digital.
And digital really has its own set of value propositions that are maybe enhanced by AI but are also independent. We were talking about being able to diagnose anywhere. That’s a huge component of it. There’s several others that I go through in that article where even being able to view more of the slide at one time.
Also, a lot of pathology residents are trained digitally, have a lot of experience being digital. And so there’s this new wave of pathologists that are digital-first in their approach to pathology. And for groups that wanna have the ability to hire a new pathologist, having support for that is is really important.
I don’t want to kind of recreate the whole article right now, but I’ll give you something to look forward to if you have a chance to read it, but yeah logistically in the lab a lot of things are simpler if you have a digital process.
But yes, there’s the expense side of getting a scanner and an IMS and viewer partner But it can really accelerate the time, the turnaround times, it can really reduce your costs around logistics and moving glass around
People, I think, naturally know you’re not going to be shipping slides around to pathologists, but also just the logistics of filing and tracking your blocks and slides can be much more simple if they never leave the building, never leave the lab, and can be filed immediately without having to be re-collected and sorted down the road and waiting for stuff to come back and trying to get them all lined up linearly. You can file immediately for compliance and for later retrieval and know that the digital is already out and distributed. So those are some of the reasons that I highlight in that article.
James Thackeray: That’s great.
I was thinking – I know you highlight this in the article and I don’t want to just rehash the entire article – but I do think people often are wondering “okay, with the cost of going digital (because there obviously is an upfront cost I don’t think anybody’s denying that it takes an investment into it.)” But the workflow for the pathologist changes so significantly and I was thinking I saw some recent enhancements your team has made to pathology workflow of cases. I don’t know if that’s the best way to term and term it but – talk about that just a little bit Would you just – like what that does if you’re a big pathology group and what that could potentially do for that group and efficiencies there.
Andy Ivie: Yeah, yeah, one of the things we focused on is building these flexible worklists that pathologists can use and configure their own way.
I’ve been in labs that have these very complicated sets of documentation of how Dr. X and Dr. Y and Dr. Z want their slide trays laid out how they want them stacked, what preferences they have… And then even if they follow that perfectly when the pathologist receives the stack of slide trays they often have to kind of sort through them separately and figure out what’s going on.
Digitally, we can create many different visualizations of the same set of work for different people, different times of the day, different specialties and specializations and focuses.
And so this is actually something that people often overlook when they’re looking to move from a glass process to a digital process is how do you know what’s ready to work on now? Like everything gets ready faster, but I still wanna know which cases have all the slide images available that I need to look at, which ones are rushed, which ones are on hold, which ones are stat.
And we can create views in our software, either for groups that are using us as their LIS or those where we’re integrated into an existing LIS that they already had in the lab, and be able to sort by AI availability and also human screener availability, different statuses within the case and different tissue types, different rule-outs.
Like you can build these lists and work through them throughout your day and get notified when new cases are ready for you.
So you’re not getting a batch or batches of stacks of glass. You’re just constantly being notified when the image comes off the scanner.
James Thackeray: Yeah, I love it. It’s so it’s such a – it really is a paradigm shift, right? Because it’s just… You could have pathologists obviously anywhere but to have that kind of workflow built in and however that practice wants to build it right – like you said by specialty or by urgent, you know, or cases or whatever it could be. I mean there’s so many different ways you could build that from a workflow perspective. It’s pretty exciting.
It gets exciting to me when you look at it and you kind of start seeing how that can be can be built out and customized to the practice and how they would want to work it.
Andy Ivie: I’ll just mention that a lot of these things are what happens when a lab is fully digital. And so we hear a lot, and there’s press releases of groups that have adopted digital pathology. Often, especially larger groups adopt digital pathology at a small scale initially, and are doing it for tumor boards or other like small subsets of their work.
But when a lab goes and a pathology group goes fully digital, you get all these interesting time and efficiency savings where no longer, like no longer do you have to go and digitize a case in order to send it to somebody, it’s already digital, and so you just send it to them. No longer do you have to digitize your tumor boards. They’re just already digital and you can track them inside of your system and recall them when you want to do that quality check. Same with just general QCs and over-reads. You don’t have to ship those, you don’t have to digitize them because they’re already digital. You just link them around and share them with others whenever you need to.
James Thackeray: It’s such a good point and it’s one of those somewhat intangible sides of it that absolutely has a cost-saving and efficiency perspective to it. But it’s sometimes hard to kind of put actual cost savings to it, but it’s true. And I think there’s so many of those little things once you’re fully digital, like you said, that actually if you were to do a full evaluation, it would be recognized as cost savings in the end, you know?
AI In Digital Pathology
Okay. I’m going to shift just a minute. I know we were talking not a lot about AI. I want to actually get your perspective because I think it is somewhat unique with your history with being in this field. What, I mean, as you kind of oversee the technology side of Lumea, what is your perspective on AI? And how does Lumea work with AI vendors? And just give us that. You briefly touched upon workflow AI, but give us your perspective and kind of what Lumea is doing on that end.
Andy Ivie: Sure. Lumea is definitely committed to working with the AI that our shared customers want to use. And so we see ourselves as a hub, maybe a platform for access to these images and as an image management system that can integrate through our API and through others, connecting to others APIs. We can share the image data, we can receive the results.
We can make them available to the pathologist either upfront or retrospective studies or specific cases that are identified later on. I think all of that is a valuable tool.
I think it’s similar to self-driving cars right now, which is an application of AI that has legitimate safety concerns and is getting better and better and better. Where we are right now with AI in medicine is actually pretty similar to where we are right now with AI in self-driving cars, where it needs to be supervised. And that changes some of the dynamics. If it didn’t need to be supervised, there’s a lot of very interesting things that could happen, and hopefully for the good of humanity, AI progresses and we can do many things unsupervised with AI and have it be an incredible tool. Current state and for the near future it needs to be supervised, which introduces some economic challenges.
In general, these algorithms aren’t reimbursed and take some time to review. And so pathologists have to evaluate that and weigh it against, you know, the potential benefits to the patient and to the practice and to the liability.
So there’s a lot of careful consideration that has to go into adopting AI as a tool for a pathologist. And there’s plenty of cases where it’s a great addition and where people are using it very effectively.
I would say digital pathology and moving away from the glass and the microscope is much further along from a safety and efficacy and equivalence to the predated process where AI, I don’t think there’s anybody who would say diagnostic AI is ready to replace a pathologist. That may be a long-term goal there. I think it’s ready and available to help a pathologist.
Where digital pathology, I think, is ready to replace the microscope. There’s maybe some specific cases of different polarities and things that the current scanners can’t do effectively in a one pass and then save it for later, things that are maybe a little bit more fiddly that require different solutions, maybe even a virtual microscope where it’s being accessed remotely, but you still don’t have to have a microscope directly on your desk. So there’s some like edge cases that maybe there’s still a case for glass at the desk, but for the vast majority of anatomic pathology, digital pathology is there to replace the microscope and is ready for it.
And it takes some investment and some planning, but it works and it’s effective. In many ways it’s better and in some ways, in many ways it’s equivalent and some ways it has some drawbacks that have to be understood and worked around, but it’s a mature technology that’s ready for pathologists to use.
Recent Exciting Developments in Digital Pathology
James Thackeray: That’s great perspective. I appreciate that. Okay, I have one last question. So if you were to, without divulging IP, which I know you wouldn’t, if you just highlight, this is a chance to brag for a minute and maybe it’s more bragging for your team, but what are some of the recent developments in the last few years that you’re like, this is really cool? Like this is, and you can be specific to Lumea or it could just be in general in the industry.
Andy Ivie: Sure. Well, I’m excited about our auto registration side-by-side viewer. I think that’s going to be really interesting how that applies both to standard H&E initial reads and to the immunostains. And to describe this for those watching or listening to this, it’s side-by-side viewers that are locked in the zooming and panning and rotating, but we use an in-block fiducial that gets cut and is consistent between different sections of the block on the slide that allows us to very accurately align. We call this DxAlign. It aligns these images side-by-side so that you can zoom in to a particular part of the tissue that you’re looking at and see it across multiple levels of the H&E and also across multiple levels of additional stains and I think that’s going to have really interesting impact on pathologist efficiency and their ability to, instead of having to like remember all the tissue they saw on the other level be able to see it side-by-side or on the other stain be able to see it side-by-side.
I also think I have some interesting applications in the development of AI and the improvement of AI being able to kind of have a human annotate once and get many copies of that data across multiple levels and across multiple stains for training, I think will be an interesting application of that.
But yeah, that one’s pretty interesting from a new technology perspective. And then while maybe keeping some of the secret sauce secret, I would say some of the advancements that are kind of maybe unsexy on the back end to make the web-based cloud-enabled digital pathology to be as fast as having the whole slide image on your computer and being able to pan around directly.
We do comparative studies too on these native applications that are running on computers against what we can do over the internet and a web browser. And I feel like that is really cool technology leverages a lot of the big infrastructure that’s been made to deliver media over the internet and makes it so that we can help pathologists navigate very quickly across slides over a decent internet connection it doesn’t have to be the world’s best internet connection on a decent computer or even on a tablet and make it really efficient and really fast.
And like I said, maybe not the sexiest feature, but I get like special satisfaction when we shave off dozens of milliseconds on tile delivery because it feels like we’re, you expand that out across all of the tiles and all the images that we deliver to customers every day and it really makes their life better. They can get through their cases more quickly, they can have more confidence in the system, they can provide quicker turnaround time and better diagnoses for the patients. So it’s all fulfilling.
James Thackeray: I love it. This is great.
Well, thanks Andy. This has been, this has been a lot of fun. I hope we get to do it again. I mean, we get to choose if we do it again, so I guess we will probably do it again.
I know that the two things that you like to talk about most: there’s digital pathology and BYU Cougar Football so if there’s any last second, you know BYU shoutouts, you’re welcome to do that. We’ll probably edit them out, but go for it.
Andy Ivie: Sure. I mean, I know that prior to this year you used to really like to talk about Utah football. I don’t know when this is going out but BYU and you are playing the holy war football game a week a week or so. I hope
I hope you’re sad then, but if you need a shoulder to cry on, you can come cry on my shoulder. I’ll probably be wearing this shirt then also.
And hey, look at that. What good timing. Happy to rep the Y and go Cougs.
James Thackeray: Yeah. All right. Hey, thanks, Andy. Appreciate it.
Andy Ivie: No problem. Thanks, James.
