Veterinary Vertex

AI in Veterinary Diagnostic Imaging: Ethics and Challenges

AVMA Journals

Send us a text

Artificial intelligence is rapidly transforming veterinary diagnostic imaging, but are we ready? In this illuminating conversation with veterinary radiologist Dr. Ryan Appleby, we explore the joint position statement from the American College of Veterinary Radiology and European College of Veterinary Diagnostic Imaging on AI technologies.

The eye-opening discussion reveals that none of the currently available AI products for veterinary diagnostic imaging meet established standards for transparency, validation, or safety. Ryan walks us through the extensive collaborative process behind developing the position statement and outlines what veterinarians should expect from AI companies regarding product information. He emphasizes the critical need to separate marketing claims from scientific evidence when evaluating these emerging technologies.

Beyond just technical considerations, we dive into the ethical dimensions of AI implementation, with Ryan powerfully arguing that improved diagnostic accuracy alone isn't sufficient justification for adoption. "We really need to point out and show that leads to a better health outcome," he explains. "Otherwise, we have no business charging our clients for deploying that piece of technology." This conversation offers practical guidance for veterinarians navigating AI tools, including what clients should know about their use and privacy implications.

Whether you're already incorporating AI into your practice or simply curious about its future applications, this episode provides essential context for understanding how these powerful tools should be evaluated, implemented, and regulated in veterinary medicine. Subscribe to Veterinary Vertex for more conversations at the cutting edge of veterinary medicine and practice.

JAVMA article: https://doi.org/10.2460/javma.25.01.0027

INTERESTED IN SUBMITTING YOUR MANUSCRIPT TO JAVMA ® OR AJVR ® ?

JAVMA ® : https://avma.org/JAVMAAuthors

AJVR ® : https://avma.org/AJVRAuthors

FOLLOW US:

JAVMA ® :

Facebook: Journal of the American Veterinary Medical Association - JAVMA | Facebook

Instagram: JAVMA (@avma_javma) • Instagram photos and videos

Twitter: JAVMA (@AVMAJAVMA) / Twitter

AJVR ® :

Facebook: American Journal of Veterinary Research - AJVR | Facebook

Instagram: AJVR (@ajvroa) • Instagram photos and videos

Twitter: AJVR (@AJVROA) / Twitter

JAVMA ® and AJVR ® LinkedIn: https://linkedin.com/company/avma-journals

Speaker 1:

Mealtime is more than a bowl of food. It's the foundation for a long and healthy life. At Wellness, dry recipes are crafted with the highest quality natural ingredients for a taste that pets love and scientifically proven through AFCO feeding trials to ensure that every complete and balanced meal supports the five signs of wellbeing. With Wellness Pet Foods, pet parents can share a life of well-being together with their pet.

Speaker 2:

This is Veterinary Vertex, a podcast of the AVMA Journals. In this episode, we chat about the American College of Veterinary Radiology and the European College of Veterinary Diagnostic Imaging. Position Statement on artificial intelligence with our guest Ryan Eppelbe.

Speaker 3:

Welcome listeners. I'm Editor-in-Chief Lisa Fortier, and I'm joined by Associate Editor Sarah Wright. Today we have Repeat Ryan, who's been a fantastic author for us for JABMA and AJVR, joining us here today. Thanks, ryan, for taking time out of your busy schedule to join us.

Speaker 4:

It's a pleasure to be here. Thanks for having me again. Nice to see you both.

Speaker 2:

All right, let's dive right in. So, ryan, your Javma article discusses the ACVR and the ECVDI's physician statement that outlines the guiding principles for the ethical development and integration of AI technologies to ensure patient safety and clinical effectiveness. Please share with our listeners the background on this article. Yeah, absolutely, thank you so much.

Speaker 4:

Please share with our listeners the background on this article Absolutely. Thank you so much.

Speaker 4:

So the article came together as a part of the work of the Joint Committee on Artificial Intelligence Education and Development, which is a joint committee of both the ACVR and the ECVDI have collected volunteers with expertise in the field of artificial intelligence, and we've been working on a number of initiatives over the past few years.

Speaker 4:

We've had some great success putting together our special issue on AI, which was available in vet rat and ultrasound.

Speaker 4:

We've been working towards numerous educational projects and speaking together in numerous locations, and this year this kind of culminated in our joint position statement, which, you can imagine, took quite some time to get to the point where we were able to come to a consensus on what should be said about AI and what the colleges wanted to say, and so the way this kind of came together is through numerous hours of discussion with the committee itself. We put together a small working group led by myself, who was also chair of the committee at the time, to, I guess, draft the position based on those discussions, and then that position went back out to membership of the committee to review and discuss the items therein and the position itself and the recommendations. So that was kind of the process of putting it together and how it came together and we're really happy with what we've come up with and the information that is now available to veterinarians everywhere and we hope as well the public, to better understand AI.

Speaker 2:

Sounds like a very thorough process. So what are some of the important take home messages from this article?

Speaker 4:

I think what we really wanted folks to come away with on the article is our point of view of where things are at with artificial intelligence, so what kinds of things veterinarians should be aware of today when it comes to diagnostic imaging AI.

Speaker 4:

So this includes aspects of what veterinarians should expect from AI companies when it comes to information that's available on AI products, so what they should be looking for when they're thinking about deploying that in their practice and, at the same time, making some recommendations to strengthen what is available to them. Because one of the things that we've noticed in the field of imaging AI so far is that there's a relative lack of information for end users, which are the veterinarians, to make appropriate decisions about what artificial intelligence is being used in their practice and how they can employ that or deploy that effectively, ethically and safely for their patients. So the real crux of the paper is talking about those key core issues about how we can best approach AI and some fundamental principles about thinking of it, and we drew on a lot of the information that was put forward by the FDA primarily, along with their, I guess, their corollaries in other countries Canada, Health Canada and the UK, the UK MHRA.

Speaker 4:

They put together some great guiding documents that we've been leaning on as a committee and as a group to start to think about what should be available in imaging AI for veterinarians and what folks should be able to look for in a product such that they can use it effectively. Should be able to look for in a product such that they can use it effectively.

Speaker 3:

Yeah, all really great information. You've been so generous with your time, ryan, and it's clearly a passion of yours to keep up with AI and educate the rest of us. You've been generous with your time for our journals, as a reviewer and as an author, and obviously out of the speaking circuit as well. What sparked this interest in AI?

Speaker 4:

It's a good question.

Speaker 4:

I mean, I think I sort of I became boarded for the ACVR in 2019, which was a really good kind of time and sort of a lucky time to start to work on some projects in AI which I thought, you know, maybe a one-off or a two-off kind of thing, and has really kind of just turned into my main focus and my main passion.

Speaker 4:

And as I started to understand more about the field of imaging AI, what is available on the human health side, and especially some of the underlying regulatory principles even if there aren't true regulations but actually just regulatory principles and why we need safety and transparency with respect to these, I became really passionate about thinking about how we can better approach this in our field, because, ultimately, I think everybody that's involved in this really wants the same thing, which is, you know, better health outcomes for our patients and, in some ways, maybe a better life for veterinarians as well, and there's huge potential in AI for that, which kind of drives my passion forward. But I think that, unfortunately, in my opinion, the way in which we've gone about it so far has been a little challenging and a little fraught, and there are there are much better things that we can do, and that's kind of, I guess, what has sparked this passion.

Speaker 3:

You might know, I've been an equine orthopedic surgeon for more than 30 years, so I thought what you were going to say is you got bored being a radiologist.

Speaker 4:

No, it's.

Speaker 1:

It's typically folks that are equine surgeons that come over and join us on the radiology side.

Speaker 4:

So we'll be looking for you to start applying to programs soon. Ouch, I'm just kidding.

Speaker 3:

Yeah, that's fascinating. Like you know, it's a handful of years and you've already clearly established yourself as a leader and a key opinion leader. So well done, Thank you.

Speaker 4:

That's very kind of you.

Speaker 3:

And you talked Ryan about. Like you know, you're an author, you're here on the podcast, you're a reviewer, you are on the speaking circuit, part of this consensus statement. How else do we continue to educate our veterinary professionals on artificial intelligence?

Speaker 4:

I think that, honestly, it's such a challenging situation and a great question. Right now. A lot falls on a veterinarian, unfortunately, to seek out that information themselves and to learn it themselves, and that can be really challenging because there's so much that veterinarians need to know on a day-to-day basis. So I think that that's something that veterinarians need to be aware of is that right now, a lot of this is something that they need to educate themselves on, especially as they're already out in practice.

Speaker 4:

As part of the position, we came forward with the recommendation that veterinary colleges start to think about ways in which to integrate artificial intelligence into their curriculum. This should become a core competency of how new graduates are leaving school, and that, too, will become a huge body of work that we need to think about how we're educating our new graduates, not only on using the existing tools that are out there, but also thinking critically about where those tools best apply or where they don't apply. And then we really need some support from the companies themselves to assist with transparency, because the education can only go so far. We really need to understand the tools that exist and then we need to, you know, think about what is out there and what those companies can provide to veterinarians so that they better understand the tools that are available to them.

Speaker 2:

So what are the next steps for research in AI?

Speaker 4:

I think the world is our oyster when it comes to that. Steps for research in AI. I think the world is our oyster when it comes to that. You know, there's almost nothing that AI can't touch within our profession in one way or another.

Speaker 4:

For me, it is incredibly key that we not only think about, you know, also proving that what they can do improves health outcomes, and that's kind of the key behind evidence-based medicine, right?

Speaker 4:

Is that we're not just because we have these fancy tools or these AI systems or whatever they are, that we're using them, but rather that when we deploy them, we think about how they're impacting patient health. So to me, it's not enough if we say you know, our AI systems can detect these findings with X percentage accuracy and these positive and negative predictive values, et cetera. We really need to point out and show that that leads to a better health outcome. Otherwise, we have no business charging our clients for deploying that piece of technology. We have no business trying to integrate it into our practices. We really need to have evidence behind what we're doing, and I do think that that is possible and we will get there. That is possible and we will get there. The challenge just becomes what are the incentives and in many ways the economic incentives for us to actually do that, rather than kind of putting the cart in front of the horse, so to speak?

Speaker 2:

Very well said. Are there any commercially available AI products for diagnostic imaging that meet the required standards for transparency, validation or safety?

Speaker 4:

No. So, as part of our statement, we looked at what was available from a perspective of transparency and came to the conclusion that none of the available products meet the criteria established by the FDA, health Canada and UKHMHRA for transparency, for machine learning enabled medical devices, and so that's. You know a document that is available through learning practices that those groups came forth with a number of years ago, which is that states that the end users of products need to have enough information to be able to make informed decisions about those products, and they need to understand enough about how those systems are made and especially the underlying data sets that went into those systems, and that is all lacking for the products that exist. We really need to understand way more about that before we're able to feel or, in my opinion, before veterinarians should feel confident using these products.

Speaker 2:

Yeah, thank you. And for those of you just joining us, we're discussing the ACVR and this ECVDI position statement on AI with our guest Ryan.

Speaker 3:

Ryan, you've authored a lot of manuscripts, but that's different than a position statement. How did all your previous training culminate into and you said you were the lead on the position statement as well. How did you get to that point?

Speaker 4:

Yeah, you know, it is a very different kind of approach. You know, certainly this isn't primary research by any means, and even putting it through, you know the review process and thinking about the editorials review that came back, it's all a very different process. It's a really good question and I'm a bit stumped for the answer, I guess In some ways I think that I drew on the same sort of principles. I wanted to. You know, we as a group wanted to approach this from as much of a scientific basis as we could.

Speaker 4:

We look through the literature, we look to what should be expected of AI, and we've had many, many hours of discussion on where AI should sit and specifically imaging AI should sit within the profession and what position we as the colleges wanted to take on that AI. And so you know a lot of it, instead of, I guess, coming from it from a research project, it's the culmination of, you know, all of those discussions. So, rather than putting together data points to come to conclusions, we had discussion points to then come to conclusions and positions. So I tried as best as we could to have it in a very similar fashion. You know, still collaborate with the experts that need to be involved, from machine learning experts to radiologists, to folks in the industry of imaging, ai and trying to put together the best possible position and recommendations for the profession as a whole, but I guess just trying to do it as scientifically as we could, if that makes sense.

Speaker 3:

Yeah, during that process, in the end did you find one kind of trick, for lack of a better or effective tool? So you said like, oh, you had a discussion point, but we all know when you get a room full of people it starts to deteriorate. Maybe might be the right word, and then it's everybody's opinion. You're like hang on a minute, let's get back to the discussion point. And we have this in every aspect of our life. What one thing really worked for you in the end?

Speaker 4:

to be like, hey, we only have 20 minutes left of our time, let's get back to the discussion point, or what worked for you in this group. I have to admit that I was not always successful in that and I think there were many instances where you know we would go over our scheduled time. There were many meetings where the conversation perhaps went sideways away from what we might intend and try to get to, but in the end we would just kind of come back together and I think that dividing things up into smaller working groups was probably the best. Part of it is that you know, once the larger group had a discussion saying you know who has the bandwidth and the space to actually take all of the things that we've been talking about. Of those authors that are listed on the manuscript, even though it is a position you know from the colleges as a whole those that smaller group of people are the ones who kind of wrote the, the initial drafts before sending it back to the committee for review.

Speaker 2:

What a process Sounds like a great learning opportunity.

Speaker 4:

Yeah, it definitely was. I definitely have appreciated being involved in that and have appreciated being able to kind of take that work and also provide recommendations to other groups. You know the AAVSB just put out their white paper on AI. I've been involved with my local regulatory body here in Ontario talking about emerging technologies and AI and I think that that you know that work has been instrumental and members of our committee are also now on the AVMA task force on emerging technologies and so you know there's so much that we can draw on that experience and bring it forth to the profession as a whole. It's been really great.

Speaker 2:

Very cool. Now this next set of questions is going to be very important for our listeners, and the first one is going to be revolving around the veterinarian's perspective. So what is one piece of information the veterinarian should know about the use of AI for veterinary diagnostic imaging?

Speaker 4:

I think that you know to me. Veterinarians should know that AI is not always accurate. So veterinarians need to separate the romance claims that companies make, which are similar to romance claims that pet food companies make about how great their products are, from the science behind them. And veterinarians should treat artificial intelligence like any other diagnostic test. So where they would look for any sort of validatory measures and information on how well a snap test works or how well any other diagnostic test works, they should use the same criteria and scrutiny for artificial intelligence.

Speaker 2:

And on the other side of the relationship, what's one thing clients should know about the use of AI for veterinary diagnostic imaging?

Speaker 4:

Clients should know that AI is starting to be used within veterinary practices and they should be aware of that, and they should feel comfortable asking their veterinarian whether or not they're using it and, if they are using it, be prepared to ask some questions about the data privacy associated with that, the efficacy of the ai, how the veterinarian comes to feel confident about using that tool and why. And, in turn, the veterinarian should start to be more prepared to answer questions like that as our clientele becomes more educated and more understanding of some of these things.

Speaker 3:

Yeah, really great points. Thanks again, ryan, for everything you do for our profession in this crazy time of AI. As we wind down, we'd like to ask a little more personal question, and this is one that my daughter got in an interview recently. If you could idea what spice I would be, but I don't think he passed the interview.

Speaker 4:

I have no idea. I will say that I love to cook, um, and I particularly like the feel of a nice spice grinder. I got them for Christmas this past year from my in-laws, um, but I have no idea what spice I would be. That's such a good question. I don't pass the interview. I got nothing. I'm drawing a complete blank.

Speaker 3:

She said salt because it's highly versatile.

Speaker 4:

I like that. That's good.

Speaker 3:

I did too.

Speaker 4:

I couldn't come up with one. I hope she got the job.

Speaker 3:

She did not. I'm not.

Speaker 2:

Next time, when I was interviewing for rotating internships, one of the institutions asked me if I was a refrigerator appliance, what I would be and why, and that was. That was also like a hard question. I was like I don't even know how to answer that one. I think I ended up saying like a refrigerator or something, because you're just continuously no-transcript, cool and calm. I started translating to working in ER. It was hard.

Speaker 4:

We used to ask everyone what they would bring to a desert island. Three things they would bring to a desert island. That was one of our questions at NC State and I remember when I interviewed I said that I would bring my podcast so it's great to be here, a camping stove and my cat and I was then asked if the cat was for company or for eating, because I brought the camping stove.

Speaker 3:

That's awesome. Well, it could be both.

Speaker 4:

eventually, that's what I said, and I knew it was going to be a great fit.

Speaker 2:

Oh, very nice. Well, thank you so much, Ryan, for being here again on our podcast and for sharing to the position statement with our journals as well.

Speaker 4:

It's been my pleasure, and thank you so much for your journal's interest and for publishing on AI. I think it's so important for everyone to start to understand more about this, so thank you for taking that on as well.

Speaker 2:

And to our listeners. You can read the ACVR and the ECVDI's position statement on AI and Javma. I'm Sarah Wright with Lisa Fortier. Be on the lookout for next week's episode and don't forget to leave us a rating and review on Apple Podcasts or whatever platform you listen to.

People on this episode