Dr. Aleks Zuraw on “People of Pathology” Podcast | Digital Pathology 101 Book

In my recent podcast appearance in the The People of Pathology Podcast by Dennis Strenk, we the importance of digital pathology, its advancements, and the cultural shifts necessary for its widespread adoption. We also discussed how my book “Digital Pathology 101 – all you need to know to start and continue your digital pathology journey” can help increase digital pathology awareness and adoption.

Thriving in digital pathology requires a learning mentality. Regardless of your experience level, keeping an open mind is crucial. This field is evolving rapidly, and even experienced digital pathology users and experts must reassess the best tools and methods. Understanding how different tools and fit together to meet specific needs is key to advancing patient care and pushing the field forward.

It’s been an honor to be invited onto the podcast for the third time. One of the main reasons for my return was the publication of my book, “Digital Pathology 101: All You Need to Know to Start and Continue Your Digital Pathology Journey.” The book is read chapter by chapter on the Digital Pathology Podcast, making it accessible to those who prefer listening to reading.

Writing the book was a seven-month journey filled with challenges and insights. Compiling my experiences and knowledge into a coherent framework required a lot of heavy lifting, but it was a rewarding process. The book aims to serve three types of readers: absolute beginners, intermediate learners, and those looking to teach digital pathology. The goal is to provide a structured resource that helps readers quickly grasp the essentials of digital pathology and apply them effectively.

Writing the book was a seven-month journey filled with challenges and insights. Compiling my experiences and knowledge into a coherent framework required a lot of heavy lifting, but it was a rewarding process. The book aims to serve three types of readers: absolute beginners, intermediate learners, and those looking to teach digital pathology. The goal is to provide a structured resource that helps readers quickly grasp the essentials of digital pathology and apply them effectively.

“Digital Pathology 101” was designed to cater to various levels of expertise. Absolute beginners, people with intermediate knowledge who already know about some aspects of digital pathology, and those who want to teach the subject will all find value in it. Each group can benefit from the book’s comprehensive overview, enabling them to contribute meaningfully to their respective areas of the digital pathology field.

The history of digital pathology is both fascinating and surprising. The field can be traced back to the 1980s, with significant milestones that have shaped its evolution. One of the most notable was the CLIA 88 regulation, which initially set the field back by decades due to a scandal involving remote slide reviews. However, the pandemic provided an opportunity for a significant shift. The Centers for Medicare Services (CMS) temporarily allowed remote sign-outs, paving the way for more permanent changes. Now, digital slides can be used for remote sign-outs, bridging the gap that had existed for years. This is a testament to how advancements in technology, if carefully monitored and applied, can drive fundamental changes in the field.

Interoperability remains a critical issue in digital pathology. Few labs start from scratch; most have existing systems that any new tools must integrate with seamlessly. Ensuring that different software and hardware can work together is crucial for creating a non-disruptive diagnostic workflow. The goal is to enhance diagnostic accuracy and efficiency through seamless integration and not to complicate existing processes.

While interoperability is essential, so is collaborating with vendors to improve their systems. Initially, my role involved helping vendors understand the digital pathology workflow to create better tools. Over time, this relationship has evolved. These days, successful vendor collaborations also involve taking their insights seriously, as they have often worked with multiple labs and have a wealth of knowledge about what works and what doesn’t. This symbiotic relationship helps both parties: vendors can create more user-friendly tools, and pathologists can more effectively integrate these tools into their workflows.

A significant cultural shift is necessary for the widespread adoption of digital pathology. This shift involves a learning mentality—not just for pathologists, but also for IT experts, computer scientists, lab directors, and other stakeholders. These highly specialized experts must be open to learning from one another and adapting to new methodologies and tools. This interdisciplinary approach will help ensure that we stay current with rapid advancements in the field.

Artificial Intelligence (AI) is another game-changer in digital pathology, although it’s essential to understand the distinction between narrow AI and general AI. Narrow AI focuses on specific tasks and is heavily utilized in digital pathology for things like image analysis and natural language processing. This type of AI is not about replacing pathologists but enhancing their capabilities. For example, tools that assist in identifying cancerous regions in tissue samples help pathologists make more accurate diagnoses.

Companion diagnostics is a growing field that ties directly into the broader objectives of personalized medicine. These are tests that provide essential information to ensure the safe and effective use of a corresponding drug. In digital pathology, companion diagnostics often involve biomarker evaluations from pathology images, leading to more effective and personalized treatment plans for patients. The ability to predict molecular properties from images is another promising area of research that could significantly impact patient care.

While often overlooked, toxicologic pathology plays a crucial role in drug development. Veterinary pathologists evaluate the safety and efficacy of new drugs through preclinical testing, which involves examining animal tissues. Understanding the toxicologic profile of a substance is critical before it moves on to human trials. Though not directly linked to digital pathology, this field underscores the interdisciplinary nature of research and development in medicine.

Digital pathology is at an exciting juncture, filled with rapid advancements and opportunities for innovation. My book, “Digital Pathology 101: All You Need to Know to Start and Continue Your Digital Pathology Journey,” aims to provide a comprehensive resource for anyone involved in this field. Whether you’re a beginner, an intermediate learner, or someone looking to teach digital pathology, the book offers valuable insights and practical advice.

For those interested, the book is available in multiple formats, including a free PDF, Kindle, Audible, and physical copies via Amazon. Your feedback, particularly in the form of reviews, is invaluable and helps other potential readers gauge the book’s usefulness.

Thank you for taking the time to explore these topics with me. The journey of learning and adapting never ends, and together we can continue to advance the field of digital pathology.

• Digital Pathology 101 .
• Digital Pathology Place’ Newsletter
• Original Episode Podcast episode of The People of Pathology

• Be Part of the Pathology Evolution: Stay informed on the latest in digital pathology innovations. Subscribe for more insights, become a member of the Digital Pathology Club, and get your complimentary copy of “Digital Pathology 101“. Embark on your path to discovery and progress in the fascinating world of pathology.

• Digital Pathology Club Membership
• Free Digital Pathology 101 eBook

• DIGITAL PATHOLOGY APPLICATIONS: From Clinical Practice to Drug Development
• What’s up in Digital Pathology – a crossover podcast with Beyond the Scope
• Digital Pathology for Community Hospitals | Dr. Elizabeth Plocharczyk
• Using AI in Pathology to Benefit Patients w/ Mariano de Socarraz, CorePlus
• Digital Pathology Applications: From Clinical Practice to Drug Development

Introduction to Learning Mentality

Aleks: [00:00:00] Learning mentality is the culture ship that will have to happen and learning mentality for all parties, like regardless of the level of experience, regardless of the discipline you’re an expert in, which is keeping an open mind because this field is changing so fast, like I had to take steps back to understand what would be the best tools to move the project forward, the field forward, patient care forward, and it wasn’t always the old, well-known thing, and it wasn’t always the new, best, most talked about hype. It was a hybrid, and the key to this was understanding how these things pieces work, how they fit together, and how you can design a tool or design a project that responds to the need.

Welcome to the Digital Pathology Podcast

Intro: Learn about the newest digital pathology trends in science and industry. Meet the most interesting people in [00:01:00] the niche and gain insights relevant to your projects. Here is where pathology meets computer science. You are listening to the Digital Pathology Podcast with your host, Dr. Alexandra Zuraw.

Guest Introduction and Book Announcement

Dennis: Dr. Zuraw, so first of all, thanks for coming back on the podcast. I think this is, what, your third time here? 

Aleks: I think so. Thank you so much for inviting me three times. I hope I assume I bring enough value to be your guest for the third time, and it’s an honor to be here. 

Dennis: Oh, I appreciate that. And you absolutely do.

The big reason I wanted to have you back, you just published a book and this is called Digital Pathology 101, all you need to know to start and continue your digital pathology journey. And I read the book. Yeah. I actually listened to the book because you, you read it on your own podcast on the digital pathology podcast.

You read episode is like a chapter at a time, which I thought was a great idea. [00:02:00] So I listened to it there and I’m curious about this book.

The Journey of Writing Digital Pathology 101

Like why did you decide now was the time to write a book? 

Aleks: Yeah. Yeah. I probably decided it should be there. I probably would have benefited from this book.

And yeah, I recorded it as a podcast because actually, I don’t read. Okay. I don’t want to say my book is boring, which is a stupid thing to say at the beginning of a podcast, but it’s a technical book. And I struggle to get through technical books that I know I need to read to advance.

And this book, I wish I had something like this when I was starting my digital pathology career because I had to figure out everything on my own. I had my pathology expertise that I was supposed to contribute. My first job was at an image analysis company. I was supposed to contribute my pathology expertise to improving their image analysis [00:03:00] results.

So I had no idea about it. How images are generated, what the computer vision principles are that guide image analysis, and all these things I had to learn as I went. And I was putting pieces of a puzzle together without an outline and the picture in front of you, how it’s supposed to look. So it was a little bit of a random thing.

And then I figured it out more or less, you always learn in this space because this is a super dynamic space, but I thought, okay I can give this framework that I needed so badly at the beginning to the people who are starting this journey and I can also support others who already have.

And expertise is one area in one area of digital pathology and need all the other pieces. 

Dennis: Okay. That makes sense. And you, in the book, you talk a little bit about your journey and how you had to, there was a story about how [00:04:00] you first discovered what digital pathology even was. 

Aleks: So I’m not going to tell the story because I want people to… 

Dennis: Got to read the book for that.

Aleks: Yeah. But what I can say is that if I had the research like this, I would have avoided a couple of conflicts that were disturbing experiences at the beginning of my pathology career that were like people not being on the same page regarding knowledge about digital pathology and how different pieces of this of the discipline fit together, totally avoidable.

Conflicts that cost me some sleepless nights and stressful weeks at work. 

Dennis: I can imagine. I can imagine. How long were you working on the book? How long did it take to write it? It took me seven months, I think. It was an on and off project. It grew, of course, as I was writing it. So I naively thought, Oh, let me put like all the [00:05:00] presentations that I gave.

They make sense together and let me transcribe and edit it and make it into a book. And it’s not that simple, there are pieces missing and it required some heavy lifting when it comes to editing reading it several times by myself by others. So yeah, seven months it took me to finish it.

Balancing Content for Different Readers

Dennis: Ok, ok, So the book is meant as an introduction to digital pathology and it’s, but it’s also useful for those who already have some knowledge of the field. Like when I was reading it. I feel like I have a basic knowledge of digital pathology and I felt like I still got quite a bit out of the book.

How did you balance that, those two areas, like the beginner and kind of the intermediate learner? 

Aleks: That’s great to hear because yeah, that, that’s super useful feedback for me because, you, by the nature of what you’re doing, [00:06:00] you also talk to many different digital pathology experts and you have had several guests that I had on my podcast.

You’re definitely familiar with a lot of concepts and it’s great to hear that it’s, it was useful for you. So it is meant for two actually three types of readers. One, the first is an absolute beginner, somebody who joins the digital pathology world. Often these are people who have domain expertise in terms of product development.

Like in the tech space, they join the world of digital pathology in a digital pathology company, and they need to learn about this landscape. What are the components and what’s relevant, right? Absolute beginner with an expertise in a different domain, joining the digital pathology work. So they get the framework.

And enough information from this book to start acting and contributing and working in the digital [00:07:00] pathology space without having to spend a year learning this stuff. They can basically read the book, listen to the podcast and go and contribute in their next meeting. Then there is the second type of reader is, let’s call them intermediate and why do I call them intermediate?

It’s usually people who have expertise in one area of digital pathology discipline. For example, I come from an image analysis background. I started in an image analysis company. So I knew. Things that were related to image analysis. I knew the computer vision terms. I knew how to talk to computer scientists, and what computer vision principles they’re using to quantify and classify objects.

And I was able to translate that into how this corresponds to what’s in the pathology image. But I didn’t know all the other stuff, like how our images were generated and what was the history of digital pathology. How is it relevant to the clinical [00:08:00] space? I was working in more in the pharma space from the beginning, so I had expertise in one area, but then I was still figuring out all the other pieces as the information was coming to me from some projects, from conversations with partners, and things like that.

And then there’s a third type of reader that will benefit from this, somebody who wants to teach digital pathology. I’ve had people reach out to me and ask, Hey, can I use the diagrams from your book in a presentation at the conference, I’m going to be teaching this in this topic, and I thought this was a really good overview of what I want to talk about.

It is put in a way that it can be taught. And the spoiler alert, I will be teaching the contents of the book. In the webinar series, I don’t have it scheduled yet, but maybe by the time we publish this podcast, it’s going to be scheduled and it’s going to be linked if you’d like to invite your listeners to that.

Dennis: I [00:09:00] definitely will. So the people that are asking for these images to use for teaching, they’re talking about the giraffes, right? 

Aleks: Yeah, actually the giraffes. You’re right. Exactly. There is an image of the giraffes, so go and check. 

Dennis: Gotta read the book. Gotta read the book to understand these jokes. But the things you were saying, that’s a big part of what you do.

Like you’re not just teaching digital pathology to pathologists. You’re including people in computer vision, people in image analysis, all these other different areas are trying to I think it even says at the beginning of your podcast, you’re bridging the gap between pathology and computer science.

And so this book continues that theme, would you say? 

Aleks: It does, I don’t want to, I don’t want it to sound that, Oh, it’s not for pathologists because that’s not the case. If you’re new to digital pathology and you’re a pathologist, it’s totally going to give you the framework you need. You will just have the pathology piece of it.

So yes, it’s very much structured as a [00:10:00] holistic starter slash refresher resource, also showing. What are the different areas of digital pathology are. Yeah, I want to keep bridging the gap between pathology and computer science. And I also included a chapter that is very specific to my pathology specialty that’s not that well known in the pathology world.

Dennis: Yeah. I want to touch on that a little bit later.

Digital Pathology Milestones and Surprises

So like any good kind of manual or kind of one on one sort of course, you start with the history of the field or what you call digital pathology milestones. And it’s interesting that I don’t think a lot of people know this field, the digital pathology field actually.

Really started in the 1980s and some of the stories you were telling about, I know I’ve had other people on this podcast that have been involved in those things. So that was interesting for me to make those connections to the stories that you were telling in the book. As you were going through this and writing this sort of [00:11:00] timeline, did any of those milestones surprise you?

Aleks: Yes.  Well…Yes, surprised, or I felt strongly about them when I was researching this, I had during my, so to say, digital pathology career, several of those milestones happened and I was happy for different things to happen in advanced digital pathology. And when I did it, I was doing the research on this.

I didn’t think I was going to like it. feel anything about the regulatory or historical milestones. I thought, okay, it’s like any discipline. It has a history, right? But I think it starts with the first no, the first milestone is telepathology and there is a full timeline in the book. But the one that I was surprised when I discovered it happened was the CLIA 88 creation, so the CLIA regulation the 88 following the CLIA 67 [00:12:00] happened or this regulation appeared when a scandal happened.

A scandal where people or pathologists and cytotechnologists were reviewing slides at home. And it resulted in a lot of false results. And I’m specifically referring to the pap mill scandal with cervical cytology slides. And when I learned about it, I was like, Oh no, why did, why were they so sloppy?

Why did they do that? It resulted, it resulted in like their actions resulted in pathologists not being able to work remotely for decades. So I was like, annoyed at, something that happened in 88. I was four at that time. But when I was doing this research I was like, Oh no, if only this didn’t happen, we would be in such a more advanced place.

And then at the end of the timeline, I think I have the end of the public [00:13:00] health emergency. So basically the end of the public pandemic, sorry, the end of the COVID pandemic, And the CMS the Centers for Medicare Services in 2020. So when the pandemic started, they issued a memorandum and temporarily allowed for remote sign-out again.

So pathologists were able to sign out remotely. And throughout the pandemic, I was like, Okay, they allowed. Will they disallow? What’s going to happen? Will we be good enough? And, we, I don’t even work in a CLIA environment. I work in a different regulatory framework, GLP, Good Laboratory Practice.

But basically, I was rooting for everybody during the pandemic. Please do a great job. Please do a great job. So they allow us to keep doing this. And in my last year, this public health emergency ended and actually they did allow CMS to allow the remote [00:14:00] sign out for pathology. But the super cool thing is you can only do it on digital slides.

So they only allowed remote sign-out powered by digital pathology. You cannot take slides home and sign up remotely, but you can have a secure compliant digital setup and sign out remotely. So these two framings, two bracket milestones were something that first I was annoyed, and then throughout the whole pandemic, I was rooting for every single pathologist signing out remotely.

Let’s do a good job so that they allow us to keep it. 

Dennis: I didn’t pick that. The beginning scandal, I remember hearing about that a long time ago and, but I didn’t realize that it had such an impact on, it basically, yeah, it set the field behind for a couple of decades. It’s it’s really sad, how far we would be with digital pathology now if that had to happen. 

Aleks: Maybe we wouldn’t, you never know what kind of other hurdles would have happened, but just in this [00:15:00] context, I was like, we could have like you say, we could have been so much further, but that’s okay. We now can. I know we have shown as a community. As a profession it’s okay, we can handle that and handle this remote sign-up responsibly.

Aleks: Okay.

Interoperability in Digital Pathology

I’d like to touch on a couple of, themes or major kinds of subjects that are in the book without, without giving away too much of it, of course, but one of the first ones is interoperability. And I know we’ve talked about this before. So talk about this, like why, especially at the beginning of digital pathology and even, now, what, why is this an important issue?

Aleks: So nobody starts their digital pathology journey. With an empty lab maybe some do they just start, okay, we’re going to start a digital pathology lab and we’re starting from scratch and we can pick whatever we want, but still, you will need more than one system. If you’re an existing lab, you already have systems [00:16:00] and whichever systems, tools, hardware, software you’re bringing in, they need to work with each other because it’s a huge, Pipeline of dependencies in any laboratory with all the software tools, all the hardware tools and everything that people are using to provide care.

So if you have a tool that doesn’t work with other tools meaning if you have a tool that’s not interoperable you’re gonna, you’re going to have trouble using this tool or you will just not use it because it doesn’t fit into what you already have. So to be able to. increase access to digital pathology and with that increase access to patient care with pathology diagnosis you need tools that fit into the tools that you already have, and that’s interoperability.

I think it’s crucial because we are not the only discipline that’s evolving rapidly. There are other aspects of patient care that fit into this diagnostic ecosystem. And [00:17:00] they should all work seamlessly, maybe it’s wishful thinking, but it should work well enough together so that the so that using them is not disruptive to a diagnostic workflow.

Dennis: Okay. That kind of ties into the next thing I wanted to ask you about, which is, and this is something you’re a big proponent of, working with vendors to help them improve their systems. That this kind of fits in with the interoperability aspect? I think so you tell me about that.

Collaborating with Vendors

Like as far as getting involved with the vendors, why is that important?

Aleks: It actually has evolved since I started in the digital pathology space. I started in 2016, I think. 15, 16. It’s in the book. . 

Dennis: Yes it is. 

Aleks: The exact date is in the book. So at the beginning, it was more Okay. There were the vendors who had the technical solutions to a workflow that they would not thoroughly understand.[00:18:00]

And I saw my role as more about giving them the information that they needed to help me be successful in my digital pathology work. Now we are at the point where. Yes, there are still vendors that are just starting in this space, but there are several vendors that are very experienced. Vendors still can work from us, but we can also work from the vendors.

And for example, if we are starting the digitization project in the lab or In our practice, or I don’t know, we want a camera that is for our personal microscope, right? And the vendors probably already helped people who had the same needs. So capitalizing on this, getting involved in that way, letting them help you as well.

is super beneficial. So developing this relationship can help both parties to be more [00:19:00] successful in the digital pathology space. 

Dennis: Okay. I like that. It’s what you’re giving the vendors. Feedback on what’s working and what’s not working with their product. And then they can change it to fit your needs.

Is that, does that sound right? 

Aleks: So I see it very much in the Toxpath space, my digital pathology specialty. I still see businesses trying to help us with our Toxpath workflow and they don’t know what this workflow is. And I’m like, your tool is not going to help me because I work with slides differently.

It’s not a diagnostic workflow. So things like that in the clinical space, in the medical pathology, anatomic pathology space, I think this is better explored. Also several vendors now either already got their clearance for medical devices, be it software or hardware, or are on the path to getting this clearance.

[00:20:00] So they need to run trials, they need to run validation. So they know the workflow inside out. Because ToxPath is in a slightly different regulatory framework. It’s. There is no like prescribed path on how to do it well, but definitely, the first step is to understand how slides are being evaluated, and what’s needed, and what’s important for the user.

So it’s less than what I saw some years ago, but still there are areas that, that will benefit from pathologist feedback, especially from pathologists with non-standard workflows. 

Dennis: Ok, that’s good that you’ve seen progress in this area that I think that’s hopeful for the future.

Aleks: Yeah, it’s recent. And I think we’re also not that as pathologists, we’re not that used to taking feedback from not our colleagues. It’s standard to ask. For opinion of a more [00:21:00] experienced colleague to have a second opinion on the case to, consult the study concept, whatever you’re doing, but taking this advice from vendors is I know best how I’m working.

Why are you trying to steer me in a different direction? Are you trying to sell me more stuff? I just want this one piece. But if pathologists or if laboratories are not familiar with the space, and don’t know what the hurdles are they definitely can capitalize on the knowledge of the vendor.

And obviously there is a money exchange in the process in every capitalistic transaction, but they do want to help you. They do want you to be. a satisfied customer who can provide better care for their patients. 

Dennis: Okay. That’s, it’s a different way of thinking for the pathology.

The Cultural Shift in Digital Pathology

And actually, this sort of ties into the you mentioned in the book that a cultural shift that needs to happen for the, for digital pathology to really catch on. And I want to talk about what you mean by this cultural shift. 

Aleks: A couple [00:22:00] of aspects to this. One, one of them is what we just talked about.

But basically. A couple of aspects that I’m going to mention that can be on, can be summarized under one umbrella even though it’s called digital pathology, it’s a very interdisciplinary field. And it’s a field of highly specialized experts and in general, highly specialized experts have, worked hard and long in developing their expertise.

And like, when you are very specialized, this learning mentality you’re more than in a teaching mode, imagine a professor with like very long tenure. They’re more in the teaching mode than in the learning mode. And what I refer to as the cultural shift, we constantly need to go back to the learning mode to be able to learn from the experts from different disciplines.

[00:23:00] And it applies to the pathologist, it applies to IT experts, computer scientists, all the lab directors, everybody was highly specialized, valuable expertise, very deep knowledge. We need to take a step back and step into the learning mode, be open. Keep figuring out the best way to provide patient care.

Dennis: Okay. I like that. That’s a good message. I feel like though that might be difficult to do. Especially for, some of us who have been in the field for decades, like changing that way of thinking and going back through the learning mode, like you just said, that, that might be tough to convince people to do.

Aleks: Yeah. I think it’s. A little bit against, I don’t know, I don’t know, can you say against societal grain? Because it’s usually okay, parents teach kids, but now the teenagers probably know so much more about the new tools coming out [00:24:00] that they can totally teach their parents and just it’s a larger cultural shift, I think, to stay open.

I think we’re starting there. Part of this is the heritage of CLIA 88 and people who lived through that and who were basically terrified by this scandal because it was bad. No one wants this to happen again. So there is skepticism, right? That’s a baggage that, that people have.

Then AI revolution AI being used in all aspects of medicine, including pathology, people are afraid, they distrust the tools, right? There’s going to be a big dose of skepticism. But think we can summarize it through that. Learning mentality is the culture ship. That will have to happen and learning mentality for all parties, like regardless of the level of experience, regardless of the discipline you’re an expert in, just basically keeping an open mind because this [00:25:00] thing, this field is changing so fast.

Like I would witness some image analysis attempts of, I don’t know. It was like let’s say it was random forest. I was first. So happy that it’s gonna be so much better than the handcrafted features and I saw it in action I’m like, this is so bad. I don’t want to see it ever again And then deep learning came in and I saw it in one use case that was really nice.

And I was like, Oh, this is the new thing. We should not touch anything else that was underperforming before. Let’s use this one. And then I saw it in action and it wasn’t that always that perfect. So I went back and say, Hey, how about we combine things? So I myself was also like overly and I know I’m digressing, but I had to take steps back to understand what would be the best tools to move whatever I was doing, the project forward, the field forward, [00:26:00] patient care forward.

And it wasn’t always the old well-known thing, and it wasn’t always that new, best, most talked about hype. It was a hybrid. And the key to this was understanding how these pieces work, how they fit together and how you can design a tool or design a project that actually responds to the need. 

Dennis: It’s exciting though, like seeing some of these new products or technologies or whatever come out and yeah, maybe they don’t work the best at first.

But it’s exciting to see you can see the potential of eventually this is where it’s going to go. And that’s one aspect that I really like a lot. And they already went there in radiology. So they have all the frameworks that help them be compliant, but at the same time, innovative enough to take advantage of the new tools.

So I’m, I think it’s just a matter of time. A matter of us [00:27:00] figuring out the right frameworks and figuring a balance before, between regulators, putting the brakes on things and those who are super enthusiastic, trying to push things forward prematurely. When we find that balance, we’re going to be on a good trajectory to advancing this field.

Dennis: Yeah, for sure.

AI in Pathology: Narrow vs. General AI

You mentioned AI a little bit ago and I want to talk about that for a bit. Because that’s another one of those things like. There’s skepticism about it and people are like, this is going to replace our jobs and things like that. And I think something that you said in the book was For it to replace people, you’re talking about general AI, which is not where this is not the AI that we have.

You you mentioned, I think, narrow AI and what is the difference between those two? 

Aleks: These concepts are a little bit like discipline concepts in the AI space, in the computer science space. And narrow AI is artificial intelligence, a computer tool can be [00:28:00] algorithmic or can be, different types of computer tools that can take care of one task.

Whereas general AI is something that would function like a human could do everything all the time. So talk on the phone and text and check emails on the computer and drive. Which is not recommended, but this is what people are capable of without, too much effort. Somebody can even drink coffee at the same time.

So that would be an, like extreme example of general AI. No, in today’s world, you would have you would need one tool to answer the tags and another one to answer the emails then a self driving car to drive. And you can drink the coffee yourself, right? And you can maybe steer all these things with another tool that is voice command.

So you would have how many tools did we already mentioned? I don’t know, five, six, to take care of each [00:29:00] aspects of. Your drive to work. So this would be general AI. We are not there yet. There is research and there are efforts trying to have more and more powerful AI tools combined with automation, with robots and like really science fiction type of stuff.

That is being worked on in our daily use and in medicine as well, but also just everyday life, the tools that we have are narrow AI. And even though for example, chat GPT is sometimes being referred to as general AI. It’s a general language model, but it’s still a narrow tool that just handles length.

Now it is combined with other tools that generate images. So we have two tools combined in one. So we broadened the scope of this narrow AI tool, but it’s still in the category of narrow. 

Dennis: Okay. Okay. That makes sense. So then [00:30:00] as far as applying AI to pathology, like how does that work? So that’s still narrow AI.

How is it applied? How does that, how does it happen? 

Aleks: Yeah, so there there was actually on February 29th there was a Digital Pathology Association webinar about generative AI. So I would recommend that as a learning resource, but to put the framework around it, there are different things that AI can be applied to in pathology.

And where I started and when, where this field started was computer vision and image analysis. And now the trend and the fastest developments are natural language processing. It’s in pathology. It started with deep learning for image analysis. So recognizing things and images. Now we are trying and we’re working on leveraging the natural language processing, so the large language models where where reports can be [00:31:00] generated and now there is heavy work happening on merging the two.

So like generating words from images, generating or searching for images for words, so these are like the three main categories where AI can be applied. Images. Words and the combination of the two. 

Dennis: Okay, I see. So the AI things that we see now, this is really just the beginning, right? 

Aleks: It is the beginning.  And we have very few. Few tools that are mature enough to be used. So let’s take image analysis. There is one cleared diagnostic tool. It is called computer aided. It is the category of this tool is computer-aided diagnostics. So something, this is for the prostate cancer and the algorithm developed by PaigeAI.

And it’s cleared by the FDA. It was cleared by the FDA in 2021, and it [00:32:00] shows pathologists the prediction of cancerous region in the image. And then the pathologist goes in, checks, Oh yes, it is malignant. Let’s classify it like that. It is not malignant. Let’s classify it in a different way. And of course now it’s even more advanced and can suggest a Gleason grade and Gleason scoring.

And, but that’s one tool. There are a couple of other ones that were used for image analysis for IHC quantification, for immunohistochemistry quantification that have been there for a long time. They were rule based they didn’t hit the wall of AI explainability, the AI black box, that is a kind of a reserving factor for regulators and as an as a factor, risk factor in when we use deep learning.

But these are like a couple of tools. When I recently checked the list of all AI powered cleared medical devices, including software as a medical device [00:33:00] there was a large list of not yet 200 devices, I think 157. We can fact check it later, but do you know how many of these were for pathology?

Dennis: How many? 

Aleks: Three. And one of them was the Paige tool that I just mentioned. The other ones was, were for molecular pathology, so not really image related. So this is something that’s cleared, right? It doesn’t mean that we cannot use AI tools for enhancing the workflow and for working in different regulatory framework, right?

This FDA clearance for a device is just one category, but it basically gives the picture what’s the situation in the, in this discipline. There are other ways of using AI tools that can be used for image analysis in the Lab developed test setting as a test specific to the lab, workflow enhancements, [00:34:00] including the auto QC of the pathology images when they’re being produced and scanned and new tools are popping up, which I’m super excited about because it was being talked about for a long time and I didn’t see it in reality, was the molecular predictions from images.

The prediction of molecular properties of tissue and the way I saw it being used or read about it there was a case report from Israel. There was an algorithm used on an image. It predicted a certain mutation and then another test was used to confirm this mutation. A quick PCR was used to confirm this mutation.

And then to be able to apply treatment, start the treatment a lot earlier, if the classical workflow of next gen sequencing would be used. So they were able to shorten the time to treatment by, I think, two weeks. 

Dennis: That’s fascinating. And then they, [00:35:00] this is all image analysis of a digital slide.

Aleks: Yes. Tissue based diagnostic tests. And this particular thing, I was so excited when I first heard about those molecular predictions then I think one year passed, two years, three years, and everybody was saying, Oh no, it’s just research. It’s just research. And I stopped believing that it will ever make it to the clinic.

And then last year I read this paper and I was like, yes, finally. 

Dennis: That’s amazing. 

Aleks: It takes time, but it’s happening. For sure. Okay.

Companion Diagnostics and Personalized Medicine

I guess this sort of ties into the next thing I wanted to ask you about, which is companion diagnostics. You touch on this a little bit in the book and it’s, this is a new upcoming field and it has quite a bit to do with this, these molecular, the molecular testing and the personalized medicine that we keep hearing about.

But how is all of that relevant to digital pathology? 

Aleks: I very much like the [00:36:00] concept of companion diagnostics because it ties the diagnostic and drug development and the drug development industry, which is where I’m personally involved in. So a companion diagnostic is a medical device. Often it’s an in vitro diagnostic.

So something that can be done in vitro in contrast to in vivo. And it provides information that is essential for safe and effective use of a drug. Or a biological product. And it often involves a biomarker evaluation. So in our case, in the pathology case, The first companion diagnostic that I learned about were based on IHC quantification.

So basically the test was the IHC, and then a pathologist had to quantify it. So the IHC was the test, and then a pathologist would need to evaluate. So it was just on a slide that would be a companion diagnostic. And if they said, okay, it’s a certain threshold of positivity we can give the drug. Or no, it’s [00:37:00] below.

This person cannot get the drug because they’re gonna be harmed by this more than they will benefit. Or it has been shown that they will not benefit at all if the expression of this biomarker is not high. So now we can apply this to images. The IHC quantification is still a super valid biomarker because we have a lot of IHC markers and a lot of those proteins can be biomarkers that when you quantify, you can separate the people who respond to the drug from those who don’t.

But also, we have the molecular predictions from images, and this field is constantly evolving. So the mere fact that we can digitize images and we can apply computer tools to them. And gives us leverage. We still have to figure out exactly how we can like, what biomarkers are relevant. It’s not that, okay, we have the method and now we’re going to go and discover all that biology.

The biology has to be there. [00:38:00] So there is still this discovery discovery element to this. There’s no guarantee that those new methods are going to reveal all the new biomarkers that we’re counting on, and, but we definitely have better tools to personalize medicine even more, to only give the drugs to people who will actually benefit from them.

Dennis: Okay. I like it. And this is a fascinating new field. And I’m interested to see how this grows and how quickly, yeah. 

Aleks: Yeah. I. I don’t see huge breakthroughs recently, but it doesn’t mean that there are none to come. It’s don’t want to say marginal, but it’s it’s difficult to quantify.

Dennis: Sure. Sure. I think it’s probably too early at this point. All right.

The Role of Toxicologic Pathology

So last thing, and you mentioned this a little bit earlier, you’re if you want to call it day job, so you were, you’re in toxicologic pathology. When you’re not talking about digital [00:39:00] pathology, you’re working in toxicologic pathology.

And I want to finish with this. So what is the role of toxicologic pathology in all of this? 

Aleks: Toxicologic pathology is a very unknown specialty of veterinary pathology. I think when people hear veterinary pathology, they might think of this as parallel to what MD pathologists are doing.

They’re doing diagnosing cases. So toxicologic pathology workflow is not a diagnostic workflow. It’s a research workflow. And you need a veterinary pathologist for that because this research is being done on animals. So all the drugs that ever make it into clinical trials and later are approved and make it to the pharmacy shelves needed the preclinical component of testing. So the preclinical component of testing includes testing them on animals. And then veterinary pathologists who are [00:40:00] evaluating the tissues of these animals to check if the substance that’s being given to them is safe and are toxicologic pathologists. So the questions that they answer is, okay, is it safe?

Does it cause any toxicity? The other questions can be, is it efficacious? But basically they’re checking can we move the drugs further away in the drug development pipeline? So when there’s often this picture of a drug development funnel being presented when in talks about drug development.

So the top of the funnel is very broad where the chemists are figuring out what compounds will work. And then they work together with biologists to see, okay, will they even bind to the cells where they bind to the target? So this part of the funnel is very broad and then it narrows down. And then at the end, it’s very narrow.

Where maybe I don’t know, 10,000 candidates, you will have three [00:41:00] that will make it to a clinical trial or five that will make it to clinical trial. Clinical trial, and then later approval is. The like super thin part of the funnel, the chemical manipulation and creation of the drug is at the top and the toxicologic pathology is more or less in the middle of the funnel.

The example I would give, okay, 99 of those 10,000 that start at the top of the funnel will make it to the toxpath evaluation. And then, only one comes out at the end of the funnel. 

Dennis: Okay. I see. So it. It’s actually pretty important and sounds like relevant to companion diagnostics, which we just talked about.  Is that accurate? 

Aleks: So companion diagnostics, not really. So because companion diagnostics are already applied to the particular human patient so the diagnostic test does not have too much to do with toxicological pathology. The drug that is being [00:42:00] used. Later, if the diagnostic test says, yes, it’s beneficial, this definitely had a lot to do with toxicologic pathology because it had to be tested on animals.

Dennis: Okay. I see. So it’s an important part of the whole kind of process. That’s good. Okay. All right. 

Aleks: It’s an integral part that I think people don’t know exists. 

Dennis: Yeah. Yeah. That’s for sure. That whole section reading that I was like, okay, I didn’t know any of this stuff. Good. 

Aleks: Okay. That, I did a good job. I educated at least one person, one, one toxicological pathologist. 

Dennis: Yes. I will have links in the show notes to the book.

Conclusion and Final Thoughts

Everyone, please, you can listen to the book like I did. You can download it as a digital copy and you can get a physical copy too, right? 

Aleks: Yes. There is a physical copy on Amazon in Kindle format, all the formats I could publish it in, it’s there.

If you’re an Audible user, it’s on Audible hard copy, soft copy Kindle, whatever, and the free [00:43:00] PDF that you can download from the link that you’ll put in the show notes. 

Dennis: Wonderful. I love it. Okay. Dr. Zuraw thanks for, thanks again for coming on the on the podcast again. Talk about your book.

I, I think this is going to be a great resource for a lot of people in the field. So Dr. Alexandra Zuraw. thank you very much. 

Aleks: Thank you so much. It has been a pleasure and an honor to be a guest of the show. People of Pathology for the third time. Thank you so much. 

Thank you so much for listening. It was funny to be the guest instead of the host.

And I don’t think I was paying attention to my interviewer. I was just talking, and didn’t have to make sure that I’m not interrupting in the middle of the sentence because I was the person who would need it to be interrupted to keep asking questions. If you don’t have the book, please get the book.

And especially if you’re just beginning your digital pathology journey, it’s going to help you speed up the learning curve, shorten it, make it [00:44:00] less steep. And if you already have a book and maybe you bought it on Amazon, if you could leave a review, a five-star review so that other people know that it’s a good book and it’s useful for all digital pathology trailblazers, I will very much appreciate and I talk to you in the next episode.