Artificial intelligence is rapidly changing how wound care clinicians access information, document care, and support clinical decisions—but reliability and human oversight remain essential. In this episode of Wound Conversations, Catherine Milne, APRN, MSN, ANP/ACNS-BC, CWOCN-AP, and Elaine Song, MD, PhD, MBA, explore the evolution of AI in wound care and how clinicians can use these emerging tools responsibly in practice.
Key Takeaways
- AI is only as reliable as the information behind it. Training data, labeling errors, limited datasets, and gaps in representation can affect AI-generated results, making clinical evaluation and evidence-based verification essential.
- AI can support wound care beyond clinical assessment. Emerging applications include point-of-care information retrieval, documentation support, patient education, interactive clinical algorithms, and audits that help assess documentation against current coverage requirements.
- AI should augment—not replace—clinical expertise. Clinicians bring judgment, context, observation, and human interaction that AI cannot fully replicate. Used appropriately, AI can serve as a productivity and learning tool while the clinician remains responsible for interpreting information and making patient-care decisions.
Catherine Milne:
Hello and welcome to WoundConversations, the podcast where we share actionable insights from leaders in wound healing. I'm Cathy Milne, and today's episode is especially geared towards clinician looking to strengthen their clinical skills along with artificial intelligence, AI, because truly the art and science of wound healing require a strong foundation. We're taking approaches to artificial intelligence because before we choose treatment applied dressings or consider advanced therapies, everything starts with how we assess the wound and how we can apply artificial intelligence to our practice. We have to think about these new developments because we want to develop a consistent thoughtful approach early on because it can make all the difference in clinical outcomes and confidence. I'm really pleased to be joined by Elaine Song. So Dr. Song is a co-founder and chief executive officer of Wound Reference. Dr. Song has a medical and science and business background.
She's been a medical director for a regenerative medicine-focused biotech company and for a Joint Commission International Accredited Hospital Network. She served as a management consultant for Kaiser Permente, has practiced as a plastic surgeon in private practice in academia, and has conducted bench and clinical research in wound healing, microsurgery, and transplant immunology. Dr. Song is a professor in the Division of Plastic Surgery at the Federal University of Sao Paulo and a volunteer member of the Communications Committee for the Association for the Advancement of Wound Care. She has authored more than 250 scientific publications, book chapters, software registrations, and patents. Welcome, Elaine.
Dr. Elaine Song:
Thank you so much, Cathy. This is such a great pleasure to be here with you today. Thank you so much. I'm so honored.
Catherine Milne:
So I'm so glad you're here. So one of the things that has always intrigued me about you is that you had vision way back when that artificial intelligence could play a role in wound healing. So how did you come up with this? Or what vision did you have? What was that aha moment?
Dr. Elaine Song:
Thank you so much for this question. It's kind of an interesting question that makes me kind of think about it. When I saw that there had been a lot of investments coming into the AI space, in the software space, specifically speaking, then we thought, okay, so pretty much that seems like maybe it's going to be like an internet boom. But then I remember clearly it was November of maybe two, three years ago or something that ChatGPT came to light. And then it was a big surprise because I think at that time I actually tried it myself. I said, okay, give me a care plan for a wound X, Y, and Z. And then it came up with all this absurdities and then things that you will never do on a human being.
Catherine Milne:
So give me an example of that.
Dr. Elaine Song:
Something like, okay, apply a bandaid or make sure that you kind of wash it. It's very simple, but also things that are wrong. And then I though, okay, not ready, not ready. I brought it up to our clinical advisor and said, Hey, take a look at this output. What do you think?" And they were like, "Oh no, no, no, we cannot trust it." So basically I think at that time we were very wary because we thought, okay, this is great. That's wonderful potential, but it's dangerous. It's dangerous because it could be telling lay people things that they should be doing. They think it's right because the tone of ChatGPT, so it comes with authority. And then you're like, okay, I need to trust this guy, but then it's really kind of doing harm. And then that was back then. So we took a very kind of wonderful.
It was a wonderful vision, but we took a very precautious approach because we thought we can't do harm. I mean, it's beautiful, it has potential. That was back then. But then after that, being in touch with the investors and all that, they all ask, how about AI? How about AI? And then also our business advisors are asking, how about AI? And then we're like, okay, we have to take a deeper look into that and see how this can really be applicable in practice. So AI to be able to be usable, reliable, it has to be reliable. So it has to be vetted and then it has to have guardrails. So in that sense, back then we though about really incorporating AI into our platforms, but in a very responsible way. And it was great because back then we did a lot of research in terms of what are the best algorithms, what are the best machines that are out there, are the types of AIs that we can use?
And we were able to do a lot of experiments to make sure that whatever we had was actually safe in terms of providing answers that were actionable and evidence-based. I can discuss more a little bit about that if need be.
Catherine Milne:
So how have you seen AI evolve over time for wound healing?
Dr. Elaine Song:
Yes. For wound healing specifically speaking, it was very interesting to see because I think that basically we had an opportunity to do a lot of research in that realm as well. So really structured this and classified this in terms of AI. Well, it was a paper, so that was published and it was a great framework for us to be able to classify, understand the types of technologies that are out there. Pretty much the AI, the first era of AI, then the 1.0 that it has a 2.0 and 3.0. So the 1.0 actually dates back from many decades ago. And then it's pretty much what we call expert system. So AI is not really new. So it's basically, it was a term that was coined by this person called John McCarthy in 1955. The person's still alive and he's a professor at Stanford University or he was a computer science professor at Stanford University.
But back then, basically AI was being used for what we call expert systems. So this is called symbolic AI probabilistic models. And then it's more like if this, then that. So we can see this with older maybe decision support systems. So you put in, you choose some selections and then based off on the selections that you choose, the output is going to be something. So this is called AI 1.0. And then we also saw it evolving to AI 2.0, which is the deep learning, machine learning. And actually most of the technology that was there until last year, it was predominantly 2.0. So pretty much it's kind of like machine learning. So AI is trained in a set database. So it's limited to the type of database that is used in order to teach this AI. And then anything beyond that, it won't be able to know.
So for instance, classification of tissue types or diabetic ulcer classification or pressure injury classification, all of this has been trained using database of pictures. And then for instance, and I think that there's a lot of shortcomings on that because as I mentioned, so these are a set data. So pretty much anything beyond that, so if you have people of darker skin or different variations of that, it might actually make mistakes.
Catherine Milne:
I have this burning question. So if AI is learning by pictures and people label those pictures, what if the labels are wrong? Correct. Because we know in terms of the pressure injury literature that staff nurses, 65% of the staff nurses cannot stage correctly. So are we seeing a lot of wrong or misinformation?
Dr. Elaine Song:
Yeah. And I think that pretty much it's garbage in, garbage out. So the data set that has been used, typically speaking for the articles that I saw published on using AI 2.0, it's kind of like a set of data that's already out there. So pretty much it's used. The sole purpose is to train AI. I mean, there's other purposes as well, but it's kind of like a standardized set of pictures. And then of course people can create new types of databases, but it's using this type of database that's already has been labeled. But could you train AI using real world pictures? Yes, you could, but then you have this kind of issue because you don't know whoever's labeling that is the person right. It's so hard to agree on a stage two.
Catherine Milne:
I mean, there's a big discrepancy and mislabeling between incontinence-associated dermatitis and stage two pressure injuries. And then on top of that, we're starting to see definitions change. Oh my God. The NPIAP is redefining their stage two. And then also, I guess if they take older data from before we had a thing called deep tissue injury, and I don't know what we called it before that, a purple thing is probably what we called it. And then they changed the terminology and then we have ulcer and then it has evolved into injury. So how do we know that the information we're getting is correct?
Dr. Elaine Song:
Right. Right.
Catherine Milne:
So tell me about AI 3.0. Is that going to be different?
Dr. Elaine Song:
This is kind of like a big milestone, I think. This is where you were like, wow. Pretty much 3.0 is generative AI. So it's just like that ChatGPT that I told you a few years ago, but this is being used applied to in the clinical space. And then there are multiple ways that I see it evolving now, mostly in the practice. So I see this a lot in the practice, but also ourselves. So in our platforms, we incorporated AI to make sure that our information is retrieved in a very efficient way and provided in a very concise way as well. So it's a better way for the user to be able to find the answers that they need at the point of care. That's very customized to what the questions are with the references and everything else. But that's kind of one way of using it.
Other ways, of course, that we see they're also being used out there is, so everybody has already seen or everybody has, not everybody, but most people have already heard of some of the EMR features that are being launched out there that are able to capture the conversation.
Catherine Milne:
Okay. So you're at your provider's visit and it's listening in and transcribing your visit. Right,
Dr. Elaine Song:
Right. That also is enhanced with AI to make sure that the information is captured and documented in a logical way. Also, by utilizing AI, you could also make sure that the information that's documented is also parsed out or even the smart phrases or the small paragraph, chunks of paragraphs that are generated, you could have them ready that are also retrieved from AI as well. So real time. So I know that today, for instance, a lot of EMRs will actually allow you to save little paragraphs for efficiency. But then with AI, you could actually have those templates come in real time specifically to what you are documenting. And I think that one of the other powerful ways to utilize AI is in audits. So pretty much making sure that the documentation is aligned with the most up-to-date coverage policy determinations and whatnot.
Catherine Milne:
So that's interesting. So because there's been this big shift in CMS's view of their CTPs and there's certain documentation pieces that have to be there. And that list is pretty extensive. So how can AI help keep track of that?
Dr. Elaine Song:
Exactly. So pretty much AI can help because the one thing it's interesting to say is that, well, CMS itself is actually using contractors that are based off on AI right now in order to do pre-authorizations. So you have to counter with AI too. That's my opinion. Right.
Catherine Milne:
So there's so many other things that AI can do for us as wound healers. I know that you've worked with the Portuguese Society. Correct. Can you talk a little bit about that?
Dr. Elaine Song:
Yeah. So we've been having a really great partnership with a Portuguese society. And then with them, we co-authored several posters that were actually here at the SAWC presented and we have one today too, which is pretty exciting ...They're always making education that's very specific to Portugal. And they have those pocket guides. So the pocket guides that people use to laminate or just carry in their pockets. So
Catherine Milne:
These are home health nurses or nurses that practice wound care in Portugal.
Dr. Elaine Song:
Nurses that practice wound care in Portugal. Yeah. So all types. Acute care setting, outpatient. And
Catherine Milne:
They would carry these laminated cards.
Dr. Elaine Song:
Yeah.
Catherine Milne:
Okay.
Dr. Elaine Song:
They call it pocket guides.
Catherine Milne:
Pocket guides. Okay.
Dr. Elaine Song:
Yeah. So I'm not sure if they laminated or not, but definitely those are printable. And then -
Catherine Milne:
And so what was on these cards? How did -
Dr. Elaine Song:
Oh, yes. Yes. Those are algorithms. Okay, they're algorithms. Quick algorithms. So for instance, how to manage a venous leg ulcer or how to approach a diabetic foot ulcer. It's pretty awesome. They have, I think, over maybe 20 now of those pocket guides, and there's a really, I think, great. And then it's also adapted to the Portuguese reality, which is pretty nice. So anyway, so we kind of partner with them in the sense that they wanted to digitalize those pocket guides and then make it more available at the point of care. So as opposed to having to print it out or having to carry it around or whenever there's an update.
It takes a long time for the pocket guide to be updated and distributed again. So how can you shorten all this process? So we help them digitalize those pocket guides and then make them into interactive multi-layered algorithms. And then those algorithms are digital. So pretty much whatever you have it on paper, now you have it on digital and then they're interactive. And then the second iteration was actually out of feedback of the society. So the members are like, "Yeah, this is great, but I really need answers fast." So I am used to just typing my question and then getting my answer right away. So the second evolution here is actually AI. So we're using AI to retrieve all the information that comes from those Elko's digital pocket guides and just use this as the knowledge base so that when the user types in whatever question they have, it will retrieve the answers based off on their pocket guides as references.
And then if they click on the reference, it will open up the algorithm itself and they can look at all the step by steps and whatnot.
Catherine Milne:
Wow. So there are so many things that AI can also do. They can do patient education, specific patient education. Love it.
Dr. Elaine Song:
Yes.
Catherine Milne:
But here's the issue. Will we be replaced in 15, 20 years?
Dr. Elaine Song:
That's a great question. Well, AI always tells us that it's about complimenting, but complementing the clinical decision making whatnot. And then at the end of the day, it's not replacing. It's not replacing the professional. And then I think that's kind of truth because pretty much you still, maybe what we're going to see is people who are the patients who come to you and said, "Oh, that's what I saw on the internet. That's what ChatGPT told me." And then you have to be able to counter argument and be prepared with the evidence. Because out there, as I mentioned, so if you go to the internet, if you go out there, it's like wild AI. I always take it with a grain of salt. You always have to look at this answer and not believe it at first sight because you don't know where it's coming from. Where's the reference? So where is all the information? So basically if patients are also exposed to that kind of information and then they come to you and then they say, "Okay, that Dr. Google told me that." So we have to be ready to counter with evidence and help discern what's right, what's wrong. But yeah, I think that professionals are still not replaceable even with very advanced AI. But I think that professionals should be able to leverage AI to be more efficient.
Catherine Milne:
Wonderful. So Elaine, I want to thank you so much for joining us and sharing your expertise. This was a very valuable conversation. Thank you so much. If you have one thing you want to share with our audience about AI, what would it be?
Dr. Elaine Song:
I would say it's a wonderful resource to be used as a productivity tool, but always be taken with a grain of salt. So basically that's my approach to AI today as well. But I was just kind of thinking, what if everybody's just start thinking about AI and then dictating their lives based off of what AI says? What's going to become of us? So pretty much, i think we should be the people guiding it. So we should be the people who are making the decisions. AI can help you bring information and whatnot. But at the end of the day, we are still humans. So we need to be able to drive this and direct everything.
Catherine Milne:
Wonderful. Wonderful.
Dr. Elaine Song:
Thank you. Thank you so much. Can I ask you a question?
Catherine Milne:
Oh, okay.
Dr. Elaine Song:
How about you? What do you think about if you were to say something about AI, what would you say?
Catherine Milne:
At this point, at least in my practice, it does not replace clinical decision-making. There's so many human variables that the AI cannot pick up. It can't pick up the facial feature when the patient says, "It may hear, I don't have many pain," but they have a grimace in their forehead. And you know as one human looking at another human, that patient's in pain and I need to do something about that. AI can't get that. But maybe eventually they will. But right now there still needs to be a lot of clinical intervention from a clinician. I find AI when it is reliable that it can help a clinician learn, even an experienced clinician, because we still come up with things that we've never seen before in front of us. And for the new person coming into wound care, it's so valuable if you have a reliable source of information.
And there's lots out there. So use AI to help you become a better clinician. So that's my thought.
Dr. Elaine Song:
Wonderful. As always, very inspirational.
Catherine Milne:
Wound Conversations, as always, is brought to you by WoundSource, the trusted resource for wound care professionals and WoundCon, your connection to global virtual education in wound management. Listen today and apply tomorrow.