top of page

#208 - 🤖 Harnessing the Power of AI for Better Neonatal Outcomes (ft NeoMIND-AI founders Drs. Barry and McAdams)

Hello friends 👋

In this exciting episode of The Incubator podcast, hosts Ben and Daphna Barbeau dive into the promising world of artificial intelligence (AI) in neonatology. They are joined by special guests Dr. James Barry, medical director of the University of Colorado Hospital NICU, and Dr. Ryan McAdams, professor and chief of the Division of Neonatology and Newborn Nursery at the University of Wisconsin. Both guests are co-chairs of the Steering Committee for NeoMind AI, a collaborative focused on advancing neonatal machine learning innovations, development, and artificial intelligence.

The conversation explores the potential of AI to revolutionize neonatal care by leveraging the vast amounts of data generated in NICUs to improve outcomes, optimize care delivery, and enhance medical education. Dr. Barry and Dr. McAdams provide insights into the current state of AI in neonatology, the importance of collaboration in AI research, and the ethical considerations surrounding this technology.

NeoMind AI is introduced as an inclusive learning collaborative that aims to break down silos between organizations, educate neonatologists about AI, and foster collaboration on common data platforms. The guests emphasize the importance of clinician involvement in AI development and application to ensure safe, effective, and unbiased implementation.

This episode offers a fascinating look at the future of neonatology and the role AI will play in transforming the field. It serves as an invitation for listeners to join the NeoMind AI community and contribute to this exciting new frontier in neonatal care.


Learn more about the NeoMIND-AI team at:

Follow NeoMIND-AI on X -

McAdams RM, Kaur R, Sun Y, Bindra H, Cho SJ, Singh H.J Perinatol. 2022 Dec;42(12):1561-1575. doi: 10.1038/s41372-022-01392-8. Epub 2022 May 13.PMID: 35562414 Review.


Beam K, Sharma P, Levy P, Beam AL.J Perinatol. 2024 Jan;44(1):131-135. doi: 10.1038/s41372-023-01719-z. Epub 2023 Jul 13.PMID: 37443271 Review.


Sullivan BA, Beam K, Vesoulis ZA, Aziz KB, Husain AN, Knake LA, Moreira AG, Hooven TA, Weiss EM, Carr NR, El-Ferzli GT, Patel RM, Simek KA, Hernandez AJ, Barry JS, McAdams RM.J Perinatol. 2024 Jan;44(1):1-11. doi: 10.1038/s41372-023-01848-5. Epub 2023 Dec 15.PMID: 38097685 Review.



Short Bio: Dr. Ryan McAdams is chief of the Division of Neonatology and Newborn Nursery at the University of Wisconsin-Madison. He is also a member of the Division of Global Health, dedicating his expertise to improving neonatal care worldwide. As a co-founder of NeoMIND-AI, a collaborative of neonatologists, Dr. McAdams is dedicated to leveraging the power of artificial intelligence (AI) to enhance the quality and precision of clinical care for neonates. NeoMind-AI's goal is to create a future where neonatal care is more personalized, efficient, and effective, and where every child has the best possible start in life. The collaborative is guided by the values of innovation, empathy, collaboration, accountability, continuous learning, and excellence. Dr. McAdams' educational background includes a BS from the University of St. Thomas, an MD from the Medical College of Wisconsin, and a fellowship in Neonatology at Wilford Hall Medical Center. Throughout his career, Dr. McAdams has served as a neonatologist with the United States Air Force in Okinawa, Japan, and has participated in humanitarian global health work in various countries, including Peru, Mongolia, Cambodia, Zambia, Malawi, and Uganda. His experiences have shaped his strong conviction that every baby everywhere deserves the best care to survive and thrive, which has become the overarching focus of his research group. Dr. McAdams is committed to promoting equity, health advocacy for children, and physician wellness through his involvement in the Inclusion, Diversity, Equity, Accessibility and Anti-Racism Committee in the Department of Pediatrics. His clinical interests involve caring for babies in the UnityPoint Health-Meriter and American Family Children's Hospital NICUs, where he partners with families to promote positive experiences and outcomes. His research interests focus on improving neurodevelopmental outcomes in children, utilizing artificial intelligence and machine learning strategies, developing virtual reality simulation models, and promoting perinatal health equity.

Short Bio: Dr. Jim Barry is an Associate Professor of Pediatrics-Neonatology at the University of Colorado School of Medicine. As a co-founder of NeoMIND-AI, a collaborative of neonatologists, Dr. Barry aims to leverage artificial intelligence (AI) to enhance the quality and precision of clinical care for neonates. NeoMind-AI's goal is to create a future where neonatal care is more personalized, efficient, and effective, and where every child has the best possible start in life. Dr. Barry's educational background includes an MD from Creighton University School of Medicine, an MBA from the University of Colorado Denver, and a BS from Montana State University. He completed his internship and residency in Pediatrics at the University of Michigan Program and his fellowship in Neonatal-Perinatal Medicine at the University of Colorado (University Hospital) Program. As the Medical Director of the University of Colorado Hospital NICU, Dr. Barry provides care for critically ill newborns and their families. His commitment to patient care and education has been recognized through awards such as the Daniel M Hall Teaching Award and Physician of the Year by the Colorado Society of Respiratory Care. Dr. Barry's research focuses on improving outcomes for high-risk infants and advancing neonatal care through innovative approaches.


The transcript of today's episode can be found below 👇


Hello everybody, welcome back to the Incubator podcast. It is Sunday. We are doing an interview today with two very interesting guests. Daphne, how are you this morning?


Daphna Barbeau (00:10.856)

I'm doing well. I feel like we haven't been in the studio in a while, so I was extra pumped to get back in. And I know both of us, but especially you, have been looking forward to this interview.



Yeah, this is an interview that we've been meaning to schedule for a while. We've been talking to Jim and Ryan about this for some time. And so the date is finally here. We have the pleasure of having on with us Dr. James Berry, who is the medical director of the University of Colorado Hospital NICU and who is the co-chair of the Steering Committee for Neo Mind AI, which we will talk about more throughout the interview.


And we are also joined by Dr. Ryan McAdams, who is a professor and chief of the Division of Neonatology and Newborn Nursery at the University of Wisconsin and who is the co-chair of the Steering Committee for Neonmind AI as well. Jim, Ryan, thank you so much for making time to be on with us today.


Ryan McAdams (01:06.119)

Thank you. It's great to be here.


James Barry (01:06.754)

But yeah, thanks for having us.



I guess my first question is going to be for you, Ryan. We always want to find, before we get into why your interest was peaked as a neonatologist with the concepts of artificial intelligence, machine learning and so on, can you tell us a little bit how you found your path to neonatology and how did you end up falling in love with that specialty?


Ryan McAdams (01:36.507)

It's a great question. It was, you know, when I was in medical school, I was enamored by a lot of different aspects of medicine. So it wasn't, you know, and then when I did my pediatric residency, I was fortunate enough to, I was in the military and we rotated at different hospitals. So on my rotation at UC Davis,


It was a really busy rotation, no work hour restrictions or anything, so lots of very long shifts. But I found neonatology very exciting from the pathophysiology. The attendings I had were engaging and pushed me a lot. And so it was probably one of the harder rotations that I did. And then I got offered to moonlight for during my chief year. So.


That's the year I think that I became much more interested in it. And I think encouragement is probably the word that is important, I think, for a lot of folks in their career. So if you have the right person pouring into you and encouraging you, that can really make the difference. So one of the attendings actually called my dad, and he encouraged him that I should go into neonatology. And I think that it incentivized me. And then moonlighting that year, those attendings,


I think believing in me and pushing me in that direction, that differentiated it between other specialties that I was entertaining. And so didn't really look back after I, I did fellowship and loved it. And so I just fell in love with the job and still I'm in love with it now. I think taking care of babies is such a privilege. So that's kind of where it came from and glad I made that choice.



Yeah, it's, it's pulling all the stops, calling your parents. I mean, that's something that's what I had not heard before. What about you, Jim?


Ryan McAdams (03:22.493)



James Barry (03:23.095)

Thanks for watching!


Mine's a little bit different. I've actually had an opportunity to think about this for quite some time. Probably when I was about five or six, I witnessed the birth of a cow. So I grew up in a town of 100 in Southwest Montana, a small ranching valley where livestock was more prevalent than humans and happened to see that next door to us, there was a lot of commotion going on in the ranch next to us.


So I hopped the fence and ran over and witnessed the birth of a calf that didn't come out so well. They had to use a tractor and chains to remove the calf from the mother. And the calf survived and the mother didn't. I remember the adults were there kind of trying to cover my eyes. And they thought, oh, we shouldn't let a six-year-old see this. And I was looking the whole time and really what I was thinking is, man, there's got to be a better way to do this.


And so I think when I reflect back, that probably some of my earlier age experiences with deliveries of animals probably put me on the path. When I was in medical school, I became very interested in pediatric cardiology and I was going to be a cardiologist. Once I did my residency, I really liked...


neonatology because it didn't focus just on one organ, but you're taking care of the whole patient. And then the physiology and research potentials that existed there were just amazing to me. And so I think my path was formed probably early on. And I still have that mantra that there's got to be a better way to do this. And that's probably why I'm doing what I'm doing today, being a medical director and being involved in other things is trying to improve the.


James Barry (05:11.874)

the care that we're providing for our patients and our families. And maybe it does all go back to that day when I witnessed the Malone's delivering a calf with a tractor and chains.



We've asked the question about wine neonatology to so many, and this is by far the two most interesting answers. We've...


Daphna Barbeau (05:28.162)

I agree, both of them.


Ryan McAdams (05:30.908)

On Jim, I just want to make sure that with that cow story, this is a dad joke. I hope you do cord milking Right. So cord milking has to be in your repertoire as a unit


Daphna Barbeau (05:42.918)

Cord milking.


James Barry (05:44.514)

I did milk those cows actually. The Malone's allowed me to milk the cows, so I knew how to do that. But it was not the umbilical cord. It was attached to the udder.


Daphna Barbeau (05:52.656)






What's interesting about your path and the first question I would like to ask you related to artificial intelligence and more of automation and machine learning is we always say that neonatology is a very technically advanced field when we compare this to other specialties around the hospital. And this is by far, this is by no means demeaning to the other specialties, but we have the ventilators, we have lots of gadgets in and around the NICU. Everything is super small. And so


I'm just curious how all this was not satisfying enough for you that you felt we need to explore this next frontier. There's something even bigger coming on the horizon. And when did that happen specifically related to new technologies? And we'll ask about AI in a little bit, but when did that happen? Ryan, do you want to take this?


Ryan McAdams (06:49.339)

Yeah, so probably four or five years ago. And I think despite all that technology, we still have babies die, right? We take care of really sick kids. So we have babies die and then we also have babies that have neurodevelopmental impairment, their vision can be affected. So I think having done neonatology and seen a lot of babies not make it die, and then also babies that had some outcomes that you'd...


love to have prevented and helped them with. And then having done a lot of global health where you see different magnitudes of death in areas like sub-Saharan Africa. So despite the technology, obviously there's a disparity with that globally. So we can get pretty US centric where the 4 million deliveries that occur each year, we're the third largest country, we've got the richest country in the world, and yet those outcomes still occur here. But...


the majority of the world, 140 million births, it's in low and middle income countries and there's a lot of death and dying. So I think for me, it was saying, well, there's this, despite our technology, there's this newer technology that is becoming more common and how can we maybe incorporate that, learn about it, apply it, so that we can leverage some of the...


current strengths we have with technology in a very heavy data centric field where we're getting gigabytes of data per day on babies and maybe use that in a way to enhance care delivery in all settings. And so that was, for me, just being frustrated with why can't our outcomes be even better? Maybe this is a way forward for that. So that's, I think, what was an incentive for me early on.



What about you Jim?



Looks like you might have, yeah.


Daphna Barbeau (08:49.177)

We froze.



in a very stoic position.


Daphna Barbeau (08:54.86)

Very start. He's really, really thinking about it. It's okay, we're not live. We'll be fine. We'll see if we can get him back online.


Ryan McAdams (08:56.833)

He's thinking about this question deeply.



Yeah, yeah, we'll be fine. Mm hmm. Usually, actually, let me see if I have his cell, because usually he could just log back in and that'll be fine. The whole reason we're using this platform is that whatever he said until now is still recorded. Oh, he's back, okay.


Ryan McAdams (09:08.867)

I have a cell like that. I'm going to text.


Daphna Barbeau (09:11.999)

Yes, sir.


Daphna Barbeau (09:16.68)

Oh, and you're back. That's okay.


James Barry (09:19.53)

I don't know what happened.



Hold on. No worries. We lost you for a second. That's OK. And so I guess maybe you can give us your take on what piqued your interest as a neonatologist with these new technologies, especially considering how already technically advanced our field is.


James Barry (09:39.142)

Yeah, and so for me, it fits into my mantra, which is we have to use what we have better. And so part of my career, I was a basic science researcher in fellowship and then early in my junior faculty years, and then switched to become medical director because I saw opportunities to improve the care that we're providing. And at first I was trained in kind of adult learning theory.


dipped my toe in the water of simulation, was asked to be a simulation director for a center, and I declined that, and really what I determined over the past decade is that my whole approach is to develop tools that can improve the care that we provide. So simulation is one where you can improve teamwork, communication, technical abilities. And a few years ago, I was talking with our department head, and he brought up


the idea that AI was going to improve a lot of healthcare. And when I pressed him about that, he gave me an answer, but it didn't really satisfy me completely. And what I realized is I knew very little about artificial intelligence. And I decided that if one of my leaders was talking about AI improving our care, then I really should start trying to learn what AI is and how it can be leveraged as a tool to improve our neonatal care.


And basically, the way I look at it is that we're surrounded by data. It's data that's not being used very well and effectively. People think that about 3% to 5% of all the data in EHRs is actually used by us when we're providing care. So that means 95% of it is sitting there and wasted in unstructured types of data for them. Now with the ability of AI with increased computational power,


data storage allows us to look at that data differently and detect patterns that will help us care for our patients. If you look at physiological monitoring data, there's a lot of data that's hidden in there just within the fetal heart rate or newborn heart rate variability, oxygen saturations. So there's a lot that we can use and leverage and now is really the time because AI is now evolving so rapidly.


James Barry (12:05.282)

that there's things that we can do every day to improve our care.


So that's how I got involved. Sorry.


Daphna Barbeau (12:10.376)

Um, I really, no, that's perfect. That gets me into my first set of questions. And I really appreciate how you're helping us describe what this actually looks like. So I'm not as tech savvy as my, my partner here, Ben and, um, in the two of you, obviously. Um, so for those people in the community who are still just, just learning the lingo, getting their feet wet, maybe you guys can really, um, help us with some of the definitions, you know, what, what.


What is AI as it relates to medicine? What does machine learning mean? I think you've described it kind of in broad strokes, but so people can really start to connect with the community, especially the community you guys have built at NeoMind AI. Help the people who really don't quite understand yet what you guys are doing.


James Barry (13:05.038)

So you winna go first Ryan or you want me to just take a stab?


Ryan McAdams (13:08.684)

Take a stab, bro.


James Barry (13:10.338)

So I'll say that I think artificial intelligence shouldn't be called that. Because I think now it's kind of a buzzword that people throw around AI a lot. I think AI should really stand for augmented intelligence. That we should use the capabilities of machine learning and deep learning to support what we're doing, especially in healthcare, to improve.


Daphna Barbeau (13:17.514)



James Barry (13:35.89)

aspects of our care that we just can't do human, humanly possible. Sorry, there's somebody trying to get into the room. Go ahead, Ryan, why don't you take it from


Ryan McAdams (13:49.875)

When I think of artificial intelligence, it's having a machine.


James Barry (13:55.184)

Did you make a reservation? Yeah. OK, my apologies.


Ryan McAdams (14:02.331)

It's having a machine do a task that we instruct it to do and be able to achieve it. And so it's, if you think of whether that's explaining something, recommending something, if I type something into my phone, I winna get directions somewhere, it's giving an instruction and then having it carry out that task at a simple level. And there's different forms of artificial intelligence. So there's terms that are


used in the artificial intelligence community that maybe folks are less familiar with or maybe the word isn't the best word like Jim had kind of mentioned with the word artificial being Substituted thought with augmented so there's weak AI and weak AI could be like you in 1997 where we did deep mind I think taught and how to play chess and it beat the then champion Gary Kasparov at chess


But the AI could do 100 to 300 million calculations per second. Now, I wouldn't call that weak. I'm like, that's amazing. You can do that many moves per second and win a chess. But then it couldn't. That's all it could do was chess. So it's very narrow scoped. It couldn't play Monopoly. It couldn't recommend a movie. Whereas like Siri or Alexa or someone could recommend a movie that I should watch, but it can't play chess. So there's these.


Daphna Barbeau (15:07.444)



Ryan McAdams (15:23.811)

Weak AI means it doesn't mean it's not intelligent or can't achieve a high level task, but it's limited on what that can do. You'll hear words like artificial general intelligence, which that's like human level intelligence that that's what a lot of the open AI and a lot of these companies are working towards.


And that's an AI that can do multiple things that we can do. So it has human-like capability in many realms. And it's debatable when that will occur. The time frame from that keeps getting sooner. So a few years ago, it was like, oh, 2040. Maybe that'll happen. And now it's like, well, maybe it'll happen in five years. Maybe it'll happen in a year. I think we're just not clear on when that will happen. When we talk about machine learning,


That's a subset of AI where using algorithms like linear regression or other things, you can, it will take data and it doesn't necessarily need to be programmed for every little aspect of it, but it can look at that data and then give an output.


to that data. So using models to digest data, and those models will learn from that data and then give an output. And then there's deep learning, which is a way to represent data in a more complex way. So deep learning takes, it needs more data. So a lot of, a huge body of data, but then it uses what's kind of modeled after the brain, like neural networks. So there's an input and there's all these layers, which can be millions if not billions of layers where that data is going to get


That input is going to get assessed and then there's going to be an output So there's an encoder and a decoder and that's where like large language models where it can using things like called transformers where it can look at a huge body of data in parallel and And not forget what it's looking at and produce this huge output whether that's like write me a short book


Ryan McAdams (17:23.091)

or create an image, so text to image data. So that's a very sophisticated way that's really where the field is taken off with the deep learning because it's allowing us like these large language models and also this, now they have Sora, which is text to like short movies. So really high sophisticated outputs. And so when you hear artificial intelligence, that's like the umbrella term, machine learning's under that and then deep learning is even under that.


So I'll pause there, but hopefully that helped bring some clarity to the terms.


Daphna Barbeau (17:56.372)

Absolutely. I think that was very valuable. Thank you. And Jim, I'll let you tell us a little bit about how we're already using AI in neonatology. Where are we in terms of the field?


James Barry (18:10.882)

Well, I would say that we've been using AI for quite some time. So when I give talks about this, I ask the audience if they have used artificial intelligence in clinical care. Most of the audience shakes their head and says no. But indeed, we've been using it in different forms for quite some time. So probably the longest standing that I can think of is the EKG. So we get 12 lead EKGs on our patients, pediatric patients,


and there's black squiggly lines that run across a red lined grid system, and that there's an automatic interpretation that comes with those EKGs. And that's actually a machine looking at those black squiggly lines and providing us some information that a human should be able to recognize by looking at an EKG. So we've all been using it for quite some time in different forms, the autocomplete in our texting and our emails.


is another form of it. I would say that I've really started using the large language models and I'll use ChatGPT as what I've become comfortable with since its inception and release to the public. And I use it every single day. I help it, it helps me brainstorm ideas, it helps me refine my writing, it helps me respond to emails. But I'd have to say that it's really evolved in the last year. So when I look at...


Even its evolution, I think of the cell phone that we had in 2000 that was this big clunky thing. And now this computer in our pocket that we have today, that's how quickly I think even Chachi BT has evolved with its capabilities in the past year in a few months. So I use it for many things. You can take a picture of a ventilator screen and say, analyze this image, and it actually tells you what all the values are on the ventilator.


and even give some insight on compliance and exhale title volume and what you should be thinking about. So it has the ability to do things that are pretty remarkable. And if you understand that it was just trained on the internet and the web, should even amaze you even further. And probably one of my aha moments, I was listening to, I think his name's either Henry or Peter Lee from Microsoft.


James Barry (20:36.926)

And he was giving a talk on how Microsoft started partnering with OpenAI. And he gave the example of Charlotte's Web. He asked ChatGPT, can you tell me a little bit about Charlotte's Web? And it did. And it talked about Wilbur and Charlotte and how they were friends. And then, so he said, how does Charlotte Web teach us about humanity? And it listed off all these things, friendship, loyalty, cross species.


friendships, death. And then he asked it, well, how does Charlotte's Web teach us, or how is it similar to the Department of Pediatrics at Stanford Medicine? And it provided an example there. And there was a connection between Stanford Pediatrics and Charlotte's Web. So its ability to form connections by something, I don't think the internet probably has any words connected with Charlotte's Web and Stanford Pediatric Medicine.


somehow it has the ability to provide inferences and insight that is pretty remarkable. And so that's just examples of how I use it today. And I think in the next year, it's going to become more multimodal. So it's going to be able to use text, images, video as input and create that same output. It's already been creating that output for the past year. And I think that that's


will be one of the remarkable things. And I think that medical education will be transformed by its use. And there's some institutes around the country that are already using it pretty significantly. There's one in particular where medical students admit patients overnight. It reads through their H&P and provides feedback on what they're missing from their history and physical. It looks at their diagnosis and say, hey, you know, you missed on your training exam.


three questions based on this diagnosis. Here are three references that you may want to read. And that's in an email that's sent to that medical student by 8 a.m. the next morning. So it's pretty remarkable things that are already occurring today.


Daphna Barbeau (22:47.595)



Daphna Barbeau (22:52.)

Yeah, and it sounds like the turnaround time is pretty useful for sure. That was helpful. Before we get into what you guys are working on and the future of AI and neonatology, I'm actually going to direct this question to Ryan. I think a lot of people who maybe are less familiar with the technology worry that if we push forward into this world of AI, might we lose some of the...


quote unquote, humanism in medicine. And I know this is particularly important to you. You're coming to us literally from an art studio. How do we balance the two?


Ryan McAdams (23:28.467)

That's a essential question and a urgent question because I think that's going to apply to AI outside of medicine, but certainly within medicine where we take care of humans, we value humans, there's the human touch, there's the things that you know, like if you don't, as an adult, if you don't touch the patient, that changes the visit. So there's this human-centric need, I think, for AI that we cannot lose sight of. So...


you know, there can be a lot of worries about, you know, being replaced and things like that. And that's a topic I can talk about. But I think at the heart of it, we need to always keep humans at the center. Like, why are we, why are we using this technology in the first place? It's to enhance care, to improve outcomes.


But it's really to make the human part of us better. So like, will it free up some more time so I can spend time speaking with the family at the bedside? And can it help improve the care outcomes of that child and how is it going to do that? How's it going to do it in a way that I still have essential skills if needed to use them? So if everything's so automated, but then for some reason it's not working,


It's like the car, like it's an automated car, but if it stops working, can you drive it? Can you override it? So these are like key questions for ethics and just in general, like a lot of the people, what some folks might be worried about, afraid about, we have to like consider that. So there's this tension where if it can save lives and enhance care, let's use it, right? Let's adopt it. Let's not be too latent with that and resistant where we're like, wow, let's just drag our feet.


And five years, 10 years later, we're starting to use it, but that was hundreds of thousands of lives that could have been saved or changed. So we don't want to miss that. And in medicine, we tend to be really slow. We work at almost a glacial pace sometimes where I think with this and the way it's moving in society, we need to embrace it a little bit more, but have the right people involved.


Ryan McAdams (25:39.075)

so that it's done in a safe, effective way, but at the heart of it. And that's why I think you need people that are just beyond like a data scientist or someone who's tech-savvy, not that they couldn't put the humans first, but you need people who that's their main drivers to put the humans first so that this is a, the group that's involved with making these decisions has those values and those priorities. And that's why I think it's really important with AI that


people who aren't tech savvy, can still be part of this and include it. Because if this isn't something we want to leave them behind, we want them absolutely weighing in and making sure the things that they prioritize, the child, the family, their colleagues, all the humans that are involved, that we really value them and prioritize them. So I think that's, for me, one of the most important things in this process is that we never lose sight of that. Because if we have, I feel like we're going to lose our way.



So then I would love for a... Oh, yes, Jim, go ahead.


James Barry (26:45.59)

I just wanted to add that we all have electronic medical records like EPIC, CERN, or other things. How many of us were involved in the development of those electronic medical records? Virtually none of us. I would say that most of us in neonatology in the ICU, that medical record and information system doesn't help us very well.


So I think it's our prerogative as neonatologists and clinicians to be involved in the application and development and study of AI in our healthcare spaces. Because if you look at the publications right now, there's a lot of them that are coming out. They're fantastic, interesting, and exciting. If you look at the authors on them, there will be papers that will have 30 authors, and there will be one MD.


that's on that authorship list. So right now, this field is really heavy in the data science side, but the clinical application side is really where people like you all and us are needed because we have to make sure that it's applied, implemented safely, effectively, and as unbiased as possible and safe for our patients. And so we have to be involved as neonatologists. It's hard. I'm not a data scientist, computer science,


computer scientist or mathematician, but I've spent the last three years trying to learn about it. And if I can, I think anybody can.



I mean, I was going to ask a question, but then your comment is leading us in a different one, which I want to explore. I think my follow-up question would be, what do you guys feel is the state of our NICU when it comes to the adoption of all the new technologies that we've talked about in terms of AI machine learning? Is this a case of we're going to have to retrofit that new technology into our current system, or will our NICUs have to...


change dramatically to become a new form of a new entity that will then adopt AI and machine learning. I'm just, you may not have, obviously, you're going to share your opinions and we don't know what the future holds, but I'm curious to hear your thoughts. And Brian, if you want to go first.


Ryan McAdams (29:08.219)

Yeah, so it's like, do you need to renovate the house? Is it renovatable? Like, can you change it? Do you need a new footprint? I think the reality of, I think of like how cities change. So when I lived in Seattle, homes would get torn down all the time and there's this new beautiful home and some would just get, like we remodeled ours which was a lot of work. Maybe we should have bulldozed it and rebuilt it. But, and then the neighbor's house is falling apart. So you had all these like different homes, right? That were


there and I think if you think of NICUs around the world and around the country, like what does this look like? Because, you know, new one, there is remodeling, there's updates, but like how do you incorporate this? What does that look like? Is it more structural or is it heavily like is it the machines that we're going to be bringing in? And I think they'll probably be both. I think there'll be some NICUs of the future that will have, the design will be a little different. It'll have monitoring systems


that are going to do a much better job of some of the environmental factors that aren't being captured. Some of that can be condensed though down to a camera. So you could have a device that is multimodal. It's a camera that could have a really good field of view. It could narrow in on the baby. It could never miss anything. It's going to have.


eyes at all times, it'd be a sentinel looking over that baby. It could incorporate sound monitoring. So we can monitor decibel levels and harmonics and understand like which sounds maybe influence outcomes or, you know, apneic events or long-term outcomes, which sounds are beneficial. Like if the lack of that sound actually might have a negative impact if you don't incorporate it. So.


And then you could think of light levels, so luminometers and things that are measuring like certain that are influencing circadian rhythms and other things that influence the baby. In a lot of the world air quality is not ideal. And so it can measure particulate matter in the air. So you can have a lot of Tracking of the environment. What does that environment look like different people moving through different interactions and then understand which


Ryan McAdams (31:14.163)

For precision medicine, for certain cohorts of babies or different babies, what optimizes their care? How do you enhance their care? And then you'd have real-time feedback all the time. How are you doing with that? It could be a very dynamic process. So.


Some of these things to incorporate them wouldn't necessarily require a lot. You know, you need WIFI, you need maybe it's, is it cloud-based? Is it a local server? I would envision as things continue to evolve with the internet of things, all the devices have their own like machine learning embedded, but they all talk to each other. So all the ventilators in the whole unit are talking to each other and they can talk to each other for that patient with the pumps and all the different devices and Nears monitoring, EG.


you know, any kind of fixed interminute continuous data can all be, um, there's a communication that's happening, but then you could have that whole network within the unit. And then if you think of like federated learning where you have multiple units all over the world, they're all talking to each other and learning. So the scale of this, where then you're taking all that data and understanding it, um, could really inform in the best way for that baby that's in front of you, how to take care of them. I think one of the challenges is


privacy and how do you share data and are people willing to share data and how do you do that? So that's, I think a lot of where we're, I don't know if the word stuck is the right word right now, but a lot of the reason things aren't necessarily moving forward a little faster is sharing data is a real challenge and how to do that. I think the devices will continue to kind of evolve to allow this, how that will change, like what exactly the unit looks like.





Ryan McAdams (32:53.231)

I think remains to be seen. I think though with anything, if you think of a home, back to that home example, homes kind of modernize. And so that I think the same thing will happen with NICUs and some will look really modern and some will look, well, it looks, you're moving in that direction. You fixed the kitchen. You still need to do the bathroom in the basement.






Mm-hmm. Jim, any thoughts on that?


James Barry (33:13.686)

I'll say that I would take a little more practical approach. Ryan is fantastic in the way he thinks about this. I always enjoy listening to him talk and describe what he's thinking. How many of you have actually spoken with your hospital CEOs and CFOs? And what's their take on artificial intelligence?



Well, from experience, the cost is really what always comes back about some of the things that Ryan just mentioned. Like, you want to upgrade the monitors that can capture that data and store it on a server. How much is that going to cost us? I mean, that's usually the point of contention.


James Barry (33:56.57)

It certainly is. And so right now, children's hospitals and most hospitals around the country have thinned and negative operating margins and profit margins. Hospitals are risk adverse because of that, especially children's hospitals. So first of all, you have to identify, is your hospital or health care system AI ready? Are they willing to say AI is something that we have to move towards in the future? About six months ago, there was a survey


talked to healthcare executives and said, do you think AI is going to be important for your healthcare system in the future? 96% of them said yes. Then the follow-up question was, well, how many of you have an AI strategy? Just simply an AI strategy, 9%. So what I would first say is that you probably need to talk with your hospital leadership on the administrative side and see what they're thinking. Meet with the chief strategy officer.


and see what kind of insight they have on AI, then you could dig a little deeper and see how many data scientists your healthcare system actually has and what the infrastructure looks like. And what's going to happen is the healthcare systems are going to have to ask themselves kind of one question. Are we going to do this in-house or are we going to buy it off the shelf? And so most healthcare systems won't have the money or the infrastructure to develop it themselves.


Now there are some healthcare systems around the country that are already doing that and they're leading the way, but most won't be that way. Then they have to have an AI strategy in terms of evaluating a lot of these off the shelf products and to see if they would be of value to their healthcare system and align with their strategic goals. So that's my approach is trying to understand what healthcare system leaders are thinking.


and trying to talk with them about some of the potential for AI in the future, exactly what Ryan described.



That's very interesting. My next question would then focus on the individuals and people like ourselves who are trying to work with new technologies specifically in the field of research. You both are very well published on this subject. The question, my question for you is, do you think that in these early phases of artificial intelligence, we're maybe trying to answer too big of questions?


based on where we are at, where we see a lot of papers coming out with predicting BPD and management of PDA and predicting mortality and long-term outcomes. And these are the big, big questions of neonatology. And I was wondering the other day about, like, maybe we're not grabbing the lower, the low hanging fruits and we're really trying to shoot for the moon very early on as the technology is just being adopted around the world. What...


What are your thoughts on that? You can go first.


Ryan McAdams (37:00.219)

So I think that's a good question to contemplate. And I think things can be done in parallel. So people can be doing the research on BPD and NAC and there's some really exciting work with ROP. But what are we not thinking about? And that gets maybe back to the healthcare economics and the money piece of it. So I think one thing in medicine that I felt like hasn't been done well, period, in our field, which requires transparency is to understand


cost better. I mean, if you were to really ask a lot of neonatologists to truly understand like at a high level, like how does the money work in the hospital? I think people would shrug. They're like, well, I know I bill for things and I often don't know it's like a menu. You don't know the prices. I mean, now they're starting to show some things like, oh, this study will cost this or this will cost this. But I do this all the time on rounds. I'm like, do you know what that costs that we're going to order?


Daphna Barbeau (37:45.828)



Ryan McAdams (37:55.143)

And sometimes just finding that out, what it's going to cost took me months for someone to find me. Well, this is what that test costs. So I'm like, oh, it's good that we know that because we were going to, we didn't really need to get it and we can save a lot of money right now. So I think understanding like beyond just like billing, but like, what does this look like for cost and then. How could we save money? How could AI save us money?


Daphna Barbeau (37:58.5)

You want to see?


Ryan McAdams (38:20.319)

Because I think those are those important questions to understand this technology to say, well, if it works, this is why we should do it. Like it's going to help, but that translates into dollars. And so I think you have to have, my driver isn't financial, but I cannot ignore financial because someone has to pay for this stuff. And the people that have to pay for it,





Ryan McAdams (38:44.539)

they have priorities that they have to consider. So it doesn't mean they don't care about the patients, but they have to really focus on the budget that they're working with and profit margins and things like that. So I think we have to be doing work in that realm. We have to be doing work in the ethics piece of it, the safety piece of it. Some stuff that may not be as fun as like, I want to solve BPD, but how do you do it in a safe and ethical way? So that's where...


getting more folks involved is key. It gets back to like, who's working on this and who are you working with? So to Jim's point, that paper with only one doctor on it, like who else was on it? Was an ethicist on it? Was someone who was an expert in economics on it? How about economics related to healthcare? How about economics related to healthcare with neonatology? So some of these opportunities, I think, we need to be focusing on that as well. So I think...


you do need to step back and say, right, if we looked at the current landscape, so there's all these papers on BPD and these predictive analytics, but we're not looking at any of these other areas, that's potentially a missed opportunity that will prevent this from moving forward. Cause you may figure the BPD thing out, but you didn't figure out the healthcare economic piece and nobody's going to pay for that tool that you want to get in the NICU.



Mm-hmm. Jim, do you have any thoughts on this? Because otherwise, I have another question for you. Go ahead.


James Barry (40:05.11)

Just mine's short in that I think that we have to be cautious because in healthcare quite frequently we have technology that we're trying to look for the problem to put the technology to focus on. And really what we need to do is start with a problem first and then work back and say, is AI something that can help us solve that problem or add insight? And I do think that AI will become more widespread in...


Non-clinical arenas first, helping healthcare executives and administrators improve their revenue cycle management. It'll improve supply chain management. I do think that there's an opportunity for chat bots to act as medical assistants for patients and their families. And that's where I see kind of the next step. But we have to make sure we're solving their problem first and not just applying a technology because it exists. We've gotten ourselves in trouble before in neonatology.


doing just that.


Ryan McAdams (41:06.375)

But I'd like to add a little bit on that because I think it's a little bit more complicated than just there's a problem and this is the tool I need. I need a screwdriver. I need a hammer. I need a shovel. Because AI is magical in a way. Like if you think of what it can do and it can do it in a non-human way, it can look at a problem in a way that we don't. How did you get to that conclusion? I would have never thought by looking at.





Ryan McAdams (41:31.783)

you know, and a segmented black and white image of the eye from a retcam image that you could tell the race of the baby or the sex of the baby or that the baby has lung disease, like, aren't we looking just for our OP. And so there's these abilities that it can have that we're not that aren't logical to us as the way we think as humans. So I think the other piece I agree with what Jim said, but then the like asterisks for that is like, because this


tool where we don't totally understand its capabilities and it's getting more and more powerful, applying it to areas that you may not necessarily think it might work, there may be something that comes out of that that's a value. And so that's also like, well, how do you make sure you're not on some like tangential.


you know, chase, it's going to really not get to you to where you want to be. I think there's some folks need to be doing that work as well, because there may be something that comes out of it. That's unexpected, which is like medicine in general. There's so many stories where like how we discovered things. Um, so I think that.






It might bring back serendipity to medical research and discovery. That's kind of cool. Yes, Jim.


James Barry (42:44.598)

And that's why we need clinicians involved because you all have the perspective that data scientists don't. And have the intuition and insight that will help when some pattern is recognized by deep learning that us humans would have never recognized. We have to be there to say, yeah, we can verify that. That is true and it's accurate, it's reproducible and it's reliable. But that's why we need clinicians involved.



Mm-hmm. So we need to talk about Neomind AI. And I think the questions that we've addressed until now are a great introduction to what you guys and the rest of the team at Neomind AI are thinking of on a day-to-day basis. Neomind AI is a collaborative. You can find out more about it on the web at And Neomind AI stands for Neonatal Machine Learning.


innovations, development, and artificial intelligence. Can one of you, either one of you can take this question, but what was the driving force behind the creation of this organism and what do you aim to achieve in both the short and the long term?


Ryan McAdams (44:03.571)

Do you want to just go ahead?


James Barry (44:06.154)

I was going to let Ryan take that. So it happened, I would say, somewhat serendipitously where several of us recognized that there was neonatologist involved in the study of AI in neonatal critical care. So we got together really as a learning collaborative to begin with. And then it kind of grew into Neomind AI, where we're


Our approach is to both be a learning and collaborative organization, trying to break down silos between kind of organizations and groups and learn from one another what we're doing and then also trying to get our institutions to work together on common data platforms so we can move the AI algorithms between institutions to test.


the models versus having to worry about data privacy and having data move between institutions. But I'd say overall, really, the idea was to come together and learn from one another. And I've been amazed at the individuals involved are so bright. They're so dedicated. They're just so great at what they do. And I always get excited to...


listen to them talk. We have open sessions that occur every Thursday at one o'clock or every fourth Thursday at one o'clock Mountain Standard Time where a speaker talks about their research and what they're doing and then really to collaborate with others around the country. We've had pediatric cardiologists on there. We've had data scientists on there. And so it's really been fun. I've learned so much and I really, my approach is making sure that we're breaking down silos.


and that we're trying to educate neonatologists around the country in every opportunity. So we're trying to get members of our group on medical education conferences, medical conferences, all the ones that you've heard about in pediatrics. We have workshops or talks that are being presented there. And the idea is to make sure that people in neonatology in pediatrics or AI literate, or at least AI conversant, so anybody can be involved.


Daphna Barbeau (46:32.208)

I love that. And we know that all research requires collaboration, but it actually sounds like, you know, in terms of things like machine learning, I mean, collaboration is actually critical to getting the bulk of data needed for the work. Can you tell us a little bit about what the current focus for research is for the group and how you guys, you know, set the priority for what to focus on next?


Ryan McAdams (46:58.991)

Yeah, so the group is what, like Jim described, it's folks from all over the country and just some really amazing individuals, very bright and kind. I think the one thing I love about the group is people want to work together and collaborate. So it's kind of the best of medicine where you, like you have partners that really want to help solve a problem so we can deliver better care. They recognize the value of this emerging technology.


They want to participate in it. We want to lead the way so we do it right. And we want to have enough people involved so that it's a lot of different, we want to be very inclusive for a lot of different inputs so that how we go about trying to deliver precision medicine and medicine that's efficient and effective is done in the right way. So one of the, you know, the initial thing we wanted to work together is just to write a white paper on


kind of the current state and describe some of the ways we're thinking about AI. And that we were fortunate to get that published in Journal of Perinatology.


A lot of the work now, so there's groups that are working together within that. So, you know, of our NeoMind AI group, some folks are doing projects, like maybe there's three or four people or five people doing projects together. As a larger group, a lot of what we've discussed is data sharing and how do we go about doing that and what are the hurdles for that? Cause there's lots of hurdles and some institutions are farther ahead than others. Some folks in the group have different skillsets that they can


leverage more, they're more familiar with like data science, or they're working with Epic on things, or they've worked at Epic. I think a lot of it is trying to get our data in a way that we can collaborate. So one of our folks in the group at Johns Hopkins, Kaiser Aziz, he's done a lot with data sets and converting into a format called OMOP, which allows this kind of standardized definitions. And if other centers have that,


Ryan McAdams (48:58.547)

Then you can start to pool that data and you can share, you don't have to share data, you don't have to send data anywhere, it can stay local, but you can share the model, the machine learning model. And then it's like you have all the data standardized by definitions and so it's looking apples to apples and that gives you larger cohorts of patients. Cause a lot of the problems we're trying to solve, whether it's NAC or BPD, sepsis.


you need a lot of data and you need good data and you can't answer it on, we see this all the time, with randomized controlled trials and they'll do a meta analysis and it's like, well, that trial when we got more data, that it pan out, that's actually not an effective treatment. And so we're trying to solve this in a way, like how do we work together and overcome some of those barriers and what does that look like? And


Daphna Barbeau (49:36.752)



Ryan McAdams (49:50.467)

And so proof of concept for that, but also hopefully coming up with a way that other institutions then can say like, oh, these five places did it. Here's kind of a template on how to do it. It looks like it's safe. This is a pathway and there'll be some nuance for each institution, but we could follow this path. We can.


talk to Kaiser and ask them how did you guys do this and get them on the call. And so I think these are some early ways to just share data. And then there'll be individual projects, but Jim might also want to talk about the group is, it's focused more on than just like a data project. We have areas like education and a lot of other areas that we need to focus on with this. And Jim, maybe you just want to add a little bit around some of the other.


things that we talk about in areas of need.


James Barry (50:44.238)

We certainly understand that there's a lot of ethical challenges potentially. And so we do have an interest group focused on the application of AI in neonatology and really coming from an ethical standpoint. And recently I listened to a medical legal expert neonatologist on the West Coast who spoke on an AAP conference.


we're inviting him to speak at our Neomind AI Open Session. So I think that there's areas of focus for research and scholarly collaboration like Ryan outlined, but I do think that some targets going forward are ethics. We certainly have to be involved in that, the medical, legal, and regulatory aspects. And one big area that I think is of high value and interest for us all is medical education.


Because right now, most of the medical educators in this country do not understand AI. They're the ones who are teaching our medical students, our residents, our fellows. And if they're not teaching them just simple tools on creating some AI conversationalism, we're really doing our trainees a disservice. So from my perspective, my role in the next one to two years is really trying to be involved in medical education and trying to help.


systems kind of find their footing on how to teach residents, students, and fellows how to use some of AI's capabilities. I look at it the same way of Microsoft Word Editor. We use that. It helps us spell check grammar, so forth. AI should just be looked as a tool, especially these large language models that we can use to support our work and make us more efficient.


So that's another area that I think is right for us to be involved in and kind of help march forward.


Daphna Barbeau (52:45.36)

I appreciate that. I guess you guys have sold me on how important this is for the future of neonatology. I love on your website, there are some resources for people who want to learn a bit more, the publications that are coming out from different team members in the group. People can follow you also on Twitter, x, at Neomindai.


But how else can people who really want to learn more, really want to kind of dive into the deep end, how can they get involved? Can anybody join your weekly meeting?


Ryan McAdams (53:23.291)

Yeah, it's open to anyone. So if they're interested, they're welcome to join. I think we, you know, we want to be very inclusive and not limit this at all. I think, you know, different folks with different backgrounds, different opinions is quite valuable. And I think, you know, a barrier to this, self perceived barrier could be like, well, I don't understand a lot of this, or I don't, I'm not really good at math, or I can't program.


And I'm like, yeah, that's me too. And I think this is just where you can, this group is a friendly group and you can get involved at any different level. And I think what we're hoping is just to get folks on board. So we don't want this to be exclusive in any way. It's really like welcoming, like, hey, let's participate. Let's have you participate. Cause I think of it as someone might come with a really creative idea and we could say, oh yeah, we could use AI.


that could solve that problem or we could tackle that together. And they didn't need to be an AI expert to look at a clinical problem and say, well, how would I solve this? And I've been thinking about it. And I think that's where these exciting kind of brainstorming conversations take place. And those are some of the most engaging, enlightening, invigorating conversations where you can discuss a problem that we're all dealing with at the bedside. And what are some of the tools we could use to solve that? And when you have a large group, folks might say, well, no,


this way or I would do it that way and then you can narrow it down to like well this actually might be the best approach let's see how we could work on that. So I think a fellow or an undergraduate student or a medical student really any level we would love folks to get involved.


Daphna Barbeau (55:05.408)

And how's the best way for people to contact you then to start joining?


James Barry (55:10.522)

so they can contact us on the website. So there's a contact portion where people can write in their interest and leave an email. I'm more than happy to have people email me and get you invited to our open session, which again is on Thursdays at one o'clock, the fourth Thursday of the month at one o'clock Mountain Standard Time. And really we're trying to develop an inclusive collaborative group that really wants insight from


Ryan McAdams (55:11.583)

See you on the phone.


James Barry (55:39.774)

everyone because everyone has a unique perspective on AI, neonatal critical care, and we'll learn from one another. So it's been really exciting and fun up until this point. And I can only imagine as the weeks and months go on that this group will get larger and we'll probably break down into smaller subgroups to talk about things. So the EHR group that Ryan mentioned.


They already have a small working group on how to improve epic functionality in particular, how to use the cosmos within Epic. So that's going to be our next open session. We'll be presented by Lindsay Kinnick out of University of Iowa and Kevin Dysart out of Nemours in Delaware. So they're going to provide some insight on how to leverage the cosmos within Epic. And so we just want people to.


Reach out to us in any way. They can reach out to me in LinkedIn as well, as well as Twitter, email, and then contact us through the website. So four different ways.


Ryan McAdams (56:48.891)

And one thing I want to emphasize, AI, it is here. And it's not going away. This isn't a fad. This is such a monumental game changer in our world. We think of things in our field like discovery of surfactant or steroids. Like...


this is going to blow all that away. Like, you know, and people right now would say, really? It's like, yes, this is like, you know, if you talk to some of the big thinkers in the field, there's this book, The Coming Wave by Mustafa Suleiman, he talks about inflection points, like discovery of fire, you know, penicillin, you know, electricity, the wheel, you know, all these things in history that have really been game changers, AI is that, and it's here. And so I think having folks get involved, participate,


because you're going to be using it all the time. And for me, it's like, well, let's be inclusive, have a friendly introduction, get folks on board, and then they're going to start to learn. It's just like anything else. It's whether it's a language or anything else, you're starting to use it, you're going to learn the language. And I think this AI is just like that. And then you can start to leverage it as a tool and participate more. So I think we want to...


make raise awareness for it but also say like hey come on board get involved we need you.



Yeah, I think I was listening to a Harvard Business Review podcast where they were mentioning how the introduction of industrial production raised productivity by like 20 something percent and that so far AI within like a year or so of existence has raised productivity in certain industries by 50%. Going back to your point, Ryan, about this is going to be a monumental shift.


in humanity in general, not just neonatology, not just medicine. So I think all that is very important. Daphna, any last question before we part ways?


Daphna Barbeau (58:54.96)

Now, I really wanted to thank you guys because I think it put into for… And listen, Ben is talking about the work the group is doing all the time, but this really helped me put into perspective that everybody can get involved, that every research question potentially, I think, can be applied through this lens and potentially give us maybe cleaner data. So I think it's really exciting and I hope people will take…



Yeah. And I want to emphasize one more thing that you guys mentioned, obviously, several times, is that you're not going to get on these calls and find out that it's a bunch of very tech-savvy people that are going to be very... No, it's people like all of us who are wanting to learn more about this. And it's kind of nice, like you said, to see how every one person has a slightly different level of expertise and maybe one little aspect of artificial intelligence and how together...


Daphna Barbeau (59:23.952)

take advantage of the opportunity.


Daphna Barbeau (59:33.452)




you were able to do some nice things. So, all right, so Thursday, one o'clock, they'll see you there. James Berry, Ryan McAdams, thank you so much for making the time to be on with us today. We will leave some of the papers that you mentioned in the show notes, your contact information, the website for NeoMind, which can be accessed both via and


Daphna Barbeau (59:53.284)

I already have a project in mind that I'm going to bring to the group.



And for the people attending Delphi this year, you will both be speaking at the Delphi Neonatal Innovation Conference and also be holding a workshop. So we're excited to have you there in person. Ryan, Jim, thank you so much for being on with us today.


Ryan McAdams (01:00:30.535)

Thank you so much. Appreciate it.


James Barry (01:00:31.786)

Yeah, thank you. It was fun.






bottom of page