← Becker’s Healthcare Podcast
Loading summary
Transcript31 lines
- [00:01]
A
Philips is a health tech leader focused on innovation that improves the health and well being of people. Our healthcare technology and informatics solutions help care teams diagnose, treat and manage more patients with greater precision, speed and confidence. Across the care journey with Philips, clinicians are empowered with streamlined insights in the moments that matter for every patient. Better care for more people. Phillips.
- [00:29]
B
Hi everyone, this is Lucas Voss with Becker's Healthcare. Thanks so much for tuning in to the Beckers Healthcare podcast series. It's great to have you. Today we're talking about the role that AI plays in radiology informatics innovation and joining me for today's discussion, very excited to have them, Dr. Paul Yee, Associate professor of Radiology and Chief of the Intelligent Imaging informatics section at St. Jude Children's Research Hospital. And Dr. Sanjay Gandhi, chief medical officer for Philips Enterprise Informatics. Dr. Yi and Dr. Gandhi, thank you. Thanks so much for being here today. It's great to have you.
- [01:00]
C
Thank you, Lucas. It's great to be here.
- [01:02]
D
Thanks for having us, Lukas.
- [01:04]
B
Absolutely. I do want to start off with introductions for those that might not know you yet. At least if you could just share a little bit about yourself and your work in healthcare. Dr. Yi, we'll start off with you.
- [01:14]
D
Sure. Hey everyone. My name is Paul Yee. I'm a physician, scientist and radiologist and I lead the Intelligent Imaging informatics section at St. Jude Children's Research Hospital. My work sits at the intersection of clinical care, imaging, informatics and AI research. Our mission is to build an AI driven radiology ecosystem that improves diagnostic accuracy, reduces clinical burden and accelerates how quickly we can get answers for children with cancer. This is in alignment with St. Jude's mission of finding cures and saving children. In that vein, I focus on developing and safely translating AI tools that solve real workflow problems and deliver measurable value in everyday care, in addition to helping accelerate discovery about pediatric cancer and other catastrophic diseases.
- [01:59]
B
Very excited to hear about more. Some of this that you've mentioned is very exciting. Dr. Gandhi, over to you.
- [02:06]
C
Thank you. I'm an interventional cardiologist. I currently serve two roles at Philips. I am the Chief Medical Officer for our Philips Enterprise Informatics business, but also serve as the global lead for medical affairs and clinical partnership teams at Philips. Prior to joining Philips, I've been in academic practice for about 20 plus years. I still practice one day a week and my practice focused on medical education, clinical research and hospital strategy. My passion is integrating technology and data to improve care delivery so we can provide better care for more people.
- [02:45]
B
Dr. Yi, radiology is facing a number of challenges right now. Staffing shortages, declining reimbursements, and so much more, certainly at St. Jude Children's Research Hospital. What are some of the issues that are most pressing for you right now and how is AI helping your team to address them?
- [03:04]
D
Yeah, so, Lucas, you're absolutely right that Radiology has a number of challenges and they often seem to be conflicting. We've got staffing shortages, declining reimbursements, but at the same time, the demand is just exploding nationwide. But I will say that at St. Jude, we have a little bit of a unique environment. We're a clinical trials hospital, it's built into our name. We're not St. Jude Hospital, we're not St. Jude Children's Hospital, but we're St. Jude Children's Research Hospital. So what this means is that by design, we're really capped at a number of patients. We don't have an emergency department. You can't walk in off the street. By and large, you have to be accepted for a clinical trial to come to St. Jude to get care. But what that means is that we're able to provide care at an extremely specialized level. We also never send a bill to our patients, that our families, so everything's covered free of charge. So I think that our challenges might look a little bit different, at least the first look, from a typical health system. But I will say we still feel the same pressures. Staffing shortages. I think every practice in the country right now is struggling to find enough radiologists to cover the shifts to make sure that there's adequate coverage for all of the different requirements for practice. Increasing image demand. Imaging is the mainstay of so many disciplines, including pediatric cancer. Then finally, the need to maintain exceptional quality. This applies whether you're dealing with kids, adults, cancer or non cancer. What I'll say is, although our environment is pretty unique, there's a lot of lessons and commonalities that we do have with other practices where AI has helped us. Most, I would say in the immediate term is improving efficiency and safety. For example, what we're using right now is AI driven image reconstruction. That's allowed us to cut certain scan times dramatically. For example, PET CT scans, these typically take us around 15 to 20 minutes. But using these AI driven image reconstruction tools, we can actually slash that down to just about five. And what this means for young children, this means less sedation, better image quality. Believe it or not, even though we're doing quicker images, we actually get better image quality related to the AI and then overall smoother Experience for families. On the flip side, what this means for radiologists is, hey, we get our images done more efficiently, they're easier to read, and ultimately we can bring that value back to the patients to make more accurate diagnoses and help their doctors make the best decisions for them. One final note on the research side, since we are a research hospital, we're also using AI to improve tumor measurement and response assessments. And these have huge implications, especially as we move into this era of precision medicine. What radiologists do currently we draw a measurement, we say the tumor is bigger because of this one measurement. In reality, these are pretty coarse and crude. So you can imagine that AI can help us do more precise measurements, maybe on a volume level. And this is not something that's fully deployed clinically yet, but, but it's shaping a future where radiologists can spend more time integrating information and less time on these repetitive tasks.
- [06:08]
C
So, you know, Paul, maybe I can. I just have a follow up question since you mentioned, you know, you're a research hospital, you see a lot of patients with cancer, especially pediatric patients. How do you utilize AI in forums like Tumor Board, where you have this data that is spread out across the hospital and you're trying to bring it in one place so you can make those decisions for your patients?
- [06:33]
D
Yeah. Well, I'll say this with a caveat, that I have a lot of hopes and aspirations for AI and Timberboard, but it's not a clinical reality yet. I do think that we have the individual pieces, but putting it together I think will take some time. But the way I see AI shaping things like tumor boards, which for the audience is where we have different doctors and different specialties come together, we discuss tough cases that might not be so straightforward and, and we try to come to a consensus about what's the best treatment option, what's the next best diagnostic step. And if you think about it, it's sort of like the Jedi Council. If you're a Star wars fan, you've got all these experts, you got people, maybe you got Yoda, you've got Obi Wan, you got people with different experiences, different kind of strengths, different weaknesses as well, where we can learn to kind of prop each other up. What I envision is rather than a Council of Jedis, we've got a council of so called AI agents where each agent specializes in a particular field. So you've got an AI agent that's an oncologist, an AI agent that's a radiologist, a pathologist, a surgeon. And rather than doing these tumor boards where we come together once a week, where we're prone to scheduling conflicts and just the reality of limited time and resources. These AI agents can talk to each other. They can scour the Internet, scour the medical literature for all of the latest and greatest, and simulate that interface and that interaction really in the background. You can imagine that you could do this not just for the patients who have the luxury of having a tumor board accessible to them or have the good fortune of being accepted for the tumor board, but every patient could potentially get this interdisciplinary council consensus. That's the vision. There's a lot of pieces that go into it. It's a lot easier said than done, but that's what I hope to see maybe in the next 10, 20 years, maybe even sooner.
- [08:27]
B
Dr. Gandhi, very poignant follow up question. Thank you so much for that. I was wondering, right as we're seeing health systems expand their use of AI in clinical areas like imaging, right. Informatics. What risks should leaders be aware of and mindful of? And how are you working? How is Philips working to mitigate some of that risk?
- [08:46]
C
So clearly when you talk of AI, there's a lot of risks. People talk of bias, fairness, data privacy, security. But I think if I was to focus on two things, I would say one thing that the leaders need to be mindful of is transparency and explainability of AI. For people to trust AI, they need to understand it better, whether it is what goes in the algorithm, how the algorithm was trained, what the output is relating to and does it apply to my patients. So I think having that transparency and explainability is very important as we think of AI, especially in healthcare, where patients trust the system, they want to trust their doctors, and for the doctors to really trust the output from any technology. I think the second key piece in my mind in terms of some of the risk is integrating AI into the workflows. We're used to a certain workflow and when you disrupt a clinician's workflow, it can really lead to bad output at the other end, regardless of how good a tool is. So we want solutions that are integrated into the workflow of clinicians so they can support them, they can augment them, but not replace them. And I think those are some of the key balance, key risks that I would kind of keep in mind as you look at implementing AI at scale in healthcare system. Philips has a responsible AI approach that has eight pillars, but I think the first and the key pillar is around human oversight. We want AI that supports, not replaces, clinical judgment. Clearly, as we design these products as we design these solutions, we want to put patients and clinicians at the center of that journey so that when these solutions come in the market, they're not just a tool, they're something that is deeply integrated with the healthcare ecosystem.
- [10:45]
B
Dr. Yee, from the application process then, and then the application piece, what guardrails did you put in place to ensure AI is used safely within your teams and in use and responsibly throughout an everyday practice?
- [11:01]
D
Yeah, I think that the first thing I'll say is that every new AI model or use case really requires a fresh look and frankly, a fresh sense of humility. We don't assume that what worked last time is going to teach us everything we need to know about the next one. Just the two examples I mentioned. When we talk about things like image reconstruction, that's a very different thing than talking about tumor measurements on images. One of them has everything to do with the behind the scenes. When we actually get the images, when we reconstruct them, when we send them to the radiologist, then the second part has everything to do with what the radiologist actually does. So that being said, when I think about how do we ensure safe and responsible use of AI in everyday practice, I really think of three pillars. The first one is awareness. I think the adage goes, you can't solve the problem that you're not aware of. So the first thing we do is we deeply study and are aware of what are the known risks, whether it's algorithmic bias and fairness, brittleness to everyday variations in how images are acquired, or maybe in human computer interaction looking at things like automation, bias and overconfidence. I think second, after we're aware of what we're looking out for, we can then figure out how do we best measure them. Some of the things that we do is build validation pipelines that look, not just that, how do we measure these things from a statistical standpoint, but also ask will these things bear out clinically? One example that I like to use is if you have a nationwide sample and you're looking at something like a quality metric, you might be able to say that, hey, a 0.001% difference is statistically significant, but I think it's questionable if that actually means anything clinically. I think it's important that whatever we measure, it's placed in the clinical context. When we think about how does this impact radiologists interpreting the study, how does this impact the patient's outcome? Maybe how does it impact the patient's experience? Then I think finally, once you have an idea of what you're measuring, how to measure it. Then it's figuring out a governance system, meaning how do we have a process and a framework where we can actually operationalize these metrics? And so every tool that we use at St. Jude goes through interdisciplinary review. And this includes radiologists, physicists, informaticists, IT and clinical operations. And the idea is, again, this council kind of analogy is none of us know everything. And in fact, it really does take a village because I might be a radiologist interpreting studies, but there's a technologist who's interacting with the patient, actually bringing them onto the scanner. They're actually acquiring the images. There's going to be a physician at the bedside communicating these results to the patient themselves. I think it's important that we keep in mind everyone's unique vantage point, their unique perspectives. That combination, I think is enough to keep us humble enough to recognize new risks, but also confident enough to deploy AI safely where it adds real value and feel like we've got enough checks and balances to make sure we're pushing the boundaries, but doing so in a way that's going to be safe and trustworthy.
- [14:02]
B
Yeah, we come back to the council approach. Right. There is a lot of folks involved, governance council, so to speak. A lot of organizations are employing them right now to be able to do exactly what you just described, which is really, really key. I want to close us off with a little bit of a forward looking perspective here on things. And I'm really excited about this question because it always yields great responses. I feel like, how do you both envision AI reshaping radiology over the next three years? And most importantly, what excites you the most about that future? Dr. Gandhi, we'll kick it off with you.
- [14:39]
C
So I don't know, Lukas, if you're a glass full or a glass empty kind of a person.
- [14:44]
B
Glass full.
- [14:45]
C
So it's a solaceful. No, I'm actually there's a lot of hype around AI, but truly there's a lot of hyper as well. I believe that the next few years would be truly transformative. AI will move from isolated tools that we have now, where people are looking at isolated algorithms for implementation, to deeply integrated systems that can automate workflows, enhance decision making and personalized care. So if you imagine end to end automation, from image acquisition to reporting, plus the kind of multi agentic work that Dr. Yi was describing, with multiple agents working together to provide advanced insights at point of care, what this would mean for the health care system would be Less administrative burden, faster diagnosis, and more time for clinicians to truly focus on their patients. To answer your question about what excites me most, I think I'm most excited about AI's ability to enable proactive personalized care while keeping clinicians and patients firmly in them.
- [15:51]
B
Dr. Yi, same question for you. How are you seeing this evolve? And then most importantly, what are you excited about?
- [15:57]
D
Yeah, well, first, I think Sanjay really put it super well. I think that it will be transformative. I think AI is going to reshape every step. Not just the radiology life cycle, if you will, but really the entire patient care ecosystem over the next few years. When I think about this as a radiologist, I don't think it's just going to impact image interpretation, but everything from planning the studies, figuring out when to schedule a patient, how to protocol, or figure out which type of study to obtain, down to image acquisition, reconstruction, reporting, follow up, and communication ultimately with patients and clinicians. What excites me most is not just automation, but I would say twofold. One is augmentation, the idea that AI can help clinicians do more than what we could do before. For example, there's a lot of things that I do in my daily practice that are pretty rote. I have to drag and drop images into a particular arrangement on my screen. Can imagine that's not exactly the most fun part of my day. If AI can do this, it just gives me more brain space and bandwidth, the focus on the things I enjoy and the things that honestly are going to make a difference in a patient. It's going to be things like actually interpreting the images, making sure that we're getting the right diagnosis once we found the findings. I think also in addition to augmentation, I'm excited about this possibility of orchestration. We talked about agentic workflows. I think it's incredibly exciting, as Sanjay put really well, that these AI tools have been very, shall we say, isolated. They do things extremely well, but they're kind of like one trick ponies. They might be able to find a pneumothorax on a chest X ray, or they might be able to tell you if a report contains an incidental finding to get follow up on. But with these agentic workflows, we're going to have something closer to a true ecosystem where it's not just, hey, here's a tool, it does this extremely well, but here's a tool, it does it extremely well. And now this other AI agent can make sense of it and decide on the next step. And that might sound like a little bit scary, but I think that again, if we use these safe and trustworthy kind of guidelines, we really have an exciting future ahead. And so if I had to summarize it, I think we're moving from point solutions to real AI ecosystems that are going to really make a big difference.
- [18:15]
B
Yeah, beautifully said. I want to close this off again with giving the floor over to you both. Dr. Ghani, we'll start with you. Anything else to add that we didn't touch on? Any final thoughts you'd like to share?
- [18:27]
C
I think one thing that we didn't touch on is that none of us can do it alone. I know we came tangentially where Paul said we all need to work together, whether it is academia, whether it's industry, whether it's regulators, whether it is private sector and technology companies. I think the future of AI to be successful needs to be collaborative. I think it needs to be data driven, and it needs to be patient centered. So if we do it well, we can truly democratize access, we can reduce disparities and truly empower physicians. And this is what we are truly committed to do.
- [19:03]
B
As Dr. Yi, anything else to add from your end here?
- [19:07]
D
Yeah, I'm incredibly optimistic about the future of AI and imaging, but I really think it takes a village. I believe that the organizations that will succeed are ultimately the ones that bring together clinicians, engineers, IT professionals, physicists, administrators, industry partners, all of the above. But really, collaboration is how we turn AI from exciting prototypes into tools that genuinely improve care. And so I think the technology is incredibly exciting. I think that at face value, one could say that it's ready. But now it's about people, process and collaboration and making it work in this amazing healthcare ecosystem that we're blessed to be a part of.
- [19:44]
B
Dr. Yoon, Dr. Ghani, thanks so much for being here today and taking some time for us. What a great conversation. Thanks for being here.
- [19:50]
C
Thank you. This was a wonderful conversation. And thank you, Paul.
- [19:54]
D
Yeah, thank you both.
- [19:55]
B
Absolutely. And we also want to thank our podcast sponsor, Philips. You can tune into more podcasts from Becker's Healthcare by visiting our podcast page at beckershospitalreview.
- [20:04]
C
Com.