Imaging aided prevention | Radiology Rewired | Season 2 Ep. 2

AI is accelerating MRI, moving imaging upstream into prevention, and catching disease years before symptoms. In this episode of Radiology Rewired, Dr. Dan Sodickson, a pioneer of parallel imaging and modern MRI, explains his take on imaging's evolving role in patient care.

After nearly two decades leading imaging research at NYU Langone, Dr. Sodickson is now Chief Medical Scientist at Function Health, tackling a different question: what happens when radiology's most powerful tools are used before disease rears its head, not after?

We get into the ideas reshaping the future of imaging: how AI is compressing 20-minute MRI scans into a single minute, why "scanners with memory" could make follow-up imaging faster and cheaper, how combining MRI with blood work and prior scans can detect cancer years before symptoms, and why Dr. Sodickson believes faster, more accessible imaging is inevitable.

Radiology Rewired · Season 2, Episode 2

🎙️ SPEAKERS

👨‍⚕️ Guest: Dr. Dan Sodickson, MD, PhD 
Chief Medical Scientist, Function Health

🎙️ Host: Dr. Vivek Singh, MD
Neuroradiologist, MUSC
Host, Radiology Rewired Podcast

🎙️ Listen on Apple Podcasts and Spotify
🔗 Learn more: https://www.rapidai.com/podcasts  

Transcript

SPEAKER_00 0:10

Today, we're joined by Dr. Dan Sodickson, a radiologist, physicist, and biomedical engineer whose work in parallel imaging fundamentally changed MRI by accelerating scan times and unlocking the data-rich capabilities that define modern imaging. After nearly two decades leading innovation at NYU Langone and running one of the country's top imaging research centers, he's now chief medical scientist at Function Health, focused on a very different question. What happens when imaging moves upstream into prevention, continuous insight, and consumer-driven care? Dr. Sodicson, so great to have you with us.

SPEAKER_02 0:47

It's great to be here. Thanks so much for having me.

SPEAKER_00 0:50

So I wanted to start with your story. You were telling me about it before this interview started and kind of how you got started in medicine, became like this MD PhD interested in both radiology and image acquisition, as well as now, you know, AI and all this stuff.

SPEAKER_02 1:04

So from when I was young, my dad always used to tell me, make room for serendipity. And that's what it was. It was a complete accident. Um, actually a series of accidents. So I started out training in physics, but in a program that introduced physicist physicists and engineers to the day-to-day practice of medicine, gave us sort of half of the medical school coursework. And some of us decided after that, well, this is pretty cool. I think I want to go on. And so I hopped from the PhD into an MD program without any prior planning. And towards the end of the MD training, I was in a rotation with a cardiac imager, a guy named Warren Manning, uh, who told me to look around and find something of interest. And I found myself coming back to the fact that we just we couldn't image fast enough to image the beating heart the way we wanted to. And I puzzled about it for a little while and actually came up with a way that I thought we could image faster by taking lots of data in parallel rather than one line at a time, and came to him at the end of that rotation and said, So I kind of want to work on this. I think maybe this is something. And he said yes, to my amazement. And so I went right back into physics. I started developing, and it's what became parallel imaging, um, a way of basically accelerating MRI by getting lots of data uh at once. So that was my complete and accidental entry into imaging. I fell in love immediately as a physicist in medicine because MRI is physics and medicine working in tandem. Um, and it's been off to the races ever since.

SPEAKER_00 2:59

Yeah. So tell me a bit a little bit about parallel imaging and what exactly that means. Can we get like a simplified explanation for the audience just to see how that impacts what we're going to see moving forward? Sure, sure, sure.

SPEAKER_02 3:10

Um well, MRI machines are often called scanners. And that's actually quite apt because if you think about it, they're like a digital scanner, or if you're old like me, a fax machine, where you basically feed the paper in at one line at a time. MRI machines also traditionally used to gather one line of data at a time and build up the image bit by bit. Parallel imaging was essentially driven by the recognition that if we have multiple different detectors around the body, each one seeing something different, we can gather multiple parts of that image in parallel. And so if you have 10 detectors, you can in principle go up to 10 times faster and practice a little less. So what it meant was that MRI with this technology on board could now operate faster or in the same time could get better image quality. And that turned out to be of great value across the field, because as we know in radiology, time is money. And also for patients, nobody really that I know wants to spend more time in our tubes. I heard it's not a good experience. Yeah, yeah, yeah.

SPEAKER_00 4:26

So it's almost like uh kind of like an analogy would just be like if you're drawing a picture or something, instead of one person drawing the picture, you have like eight people kind of contributing to different portions of the picture, and then you get your full picture because you have contributions from eight different sensors, right?

SPEAKER_02 4:40

That's right. And another way to think about it is through the analogy of our eyes. Our eyes are massively parallel imaging devices. We have all of the rods and cone cells in our retina, which are essentially capturing the entire scene at once. Right. Would that we could do that with medical imaging, but at least we can get closer with parallel imaging.

SPEAKER_00 5:00

Yeah, makes sense. So tell me about just how you've seen imaging change, especially with the advent of all these new AI algorithms to kind of speed things up. Like when we're talking about parallel imaging and how that's going to change MRI scan times, what are we talking about? And what's the progression been over the past few years? And what can we expect moving forward?

SPEAKER_02 5:17

Well, it's interesting. There's kind of a direct lineage between older acceleration techniques like parallel imaging, or about 10 years after that, there was something called compressed sensing that entered the scene. But another 10 years later, it was all AI. And in fact, sort of an interesting little tidbit is that though most people think of AI as occurring after you gather an image, right? You you feed a traditional image into an AI system and it emulates humans, it makes a call, it does an interpretation. Actually, one of the first things that happened with AI and radiology was that people figured out how to get less data and reconstruct a pristine image from that smaller amount of data, which could be gathered in a shorter time. So the accelerations we could get with parallel imaging were maybe two or three times faster than if you gathered everything one line at a time. Now we're up to four or five times faster with AI, and that's just the barest beginning. And actually, I think the way AI is used for image acceleration is interestingly different from a lot of the things that get talked about in medical AI. I think of a lot of AI and medicine nowadays as downstream. Yeah. You gather data in the usual way with your blood tests or your scanners or your sensors, and then you feed all of that into a machine learning model to do whatever you need it to do. But actually, there's an upstream use of AI as well, which is to change the data that we gather in the first place. And AI, it turns out, in this era, allows us to operate with less data, with lower quality data, allows us to gather images in settings that we might not be able to in the past. Low-field MRI scanners. In the course of my career, I watched this fascinating sort of reversal. Scanners got higher and higher and higher in field strength, stronger and stronger and stronger magnets. And then did an about face, and everybody's doing a race to the bottom now in order to get accessibility, in order to get scanners where they haven't been before. AI, an upstream AI that helps you get good images off of this more limited data is going to be completely instrumental in that. So I think AI has both downstream and upstream applications in radiology.

SPEAKER_00 7:51

Yeah, and tying into that, I think it it kind of ties into one of the common themes that we bring up on the show is that I think a lot of the discussion about AI and radiology has been based on detection and improving workflow and improving efficiency for radiologists and seeing things earlier. But I think the bigger and you know kind of under talked about side is the image acquisition side and acceleration size. And when we brought it up in a couple episodes, but you know, screening populations are going to explode in the next 10 years, because you, like you're mentioning, low-field MRI, increased access to these tools, you know, radiation dose reduction in CT, um, reconstruction of other sequences from one sequence in NMRI, like getting, you know, getting post-contrast images without giving the patient actually any dye, you're able to reconstruct it with these AI algorithms. So, in my thought is if you're able to scan a patient, you know, we're running scanners 24 hours a day right now. If you're able to cut down, like you're saying, like by half, by 10 times and and get those scans done way quicker, there's no reason that more people aren't going to get more imaging because it's the superior test. And if it's it's more accessible and it's quicker and it's more comfortable for patients, that's just gonna lead to more imaging in the future.

SPEAKER_02 9:04

Absolutely it is. And and we'll talk, I know, about the implications of that increase in imaging and the and the change in imaging's role in our lives. But you're absolutely right about the role of AI and and modern sort of technologies in changing not just the current applications of imaging, but the way we do it in the first place. Uh and I'll tell you one kind of cool thing, uh, a little bit of research from my NYU lab that we're also interested in carrying on at function. It turns out that if you have already have images of a patient. So let's say this isn't the first time they're being seen, they're they're being followed up for risk of prostate cancer, for example, yearly, um, or you even just have one prior image, you can feed that into an AI system and get an acceleration not of two times or 10 times, but of 20 times or 30 times. So you can take a 20-minute exam and cut it down to a minute if you have some previous information about that person. Because it turns out if all you need to do is detect change from a previous time, you don't need a lot of data. So down the line, you could imagine that not only will imaging machines get more accessible, but interval scans will become much, much quicker, easier, and cheaper. And it'll become that much easier to follow people with imaging over time.

SPEAKER_00 10:40

Yeah. So for you know, for our audience, like a large percentage of what we're reading in the reading room is follow-up scans. We're following these post-treatment changes for this cancer patient, or you know, they have this tumor that's been treated and we're looking for recurrence. These patients come back every, you know, depending on how aggressive their cancer is and things like that, like three to six months at times. You know, so if you're able to do this much quicker and even at shorter intervals, if you want to, like again, adds adds to the credence that the imaging volumes are going to continue to skyrocket in the future, you know.

SPEAKER_02 11:12

Absolutely. Absolutely.

SPEAKER_00 11:13

And I wanted to ask you, so we have talked a lot about AI's role in like the current medical sphere where we're too looking at detection um and diagnosis. You're working a lot in the preventative space and kind of looking at healthier patients and finding disease earlier. So interested in hearing about that and kind of what your take is on all this and and and how you got into this space and what you're doing now. Yeah.

SPEAKER_02 11:38

Well, the story of how I got into it is really a story of a gradually growing frustration that these devices that I've devoted my career to building, to modifying, to using your MRI machines, your CT, pet, ultrasound, all of these remarkable devices that can slice through the body any which way without making a single cut, we basically only use them once we already know you're sick. Yeah. And to me, as I got on in years, that started seeming like more and more of a crying shame. Because I know how much we can see on an MRI, for example. I know how well we can characterize disease. So why are we using our best tools last rather than first? Now, there are many reasons that we can get into why that's been the case, although I think that's changing. But for me, that perspective is what ultimately led me into this work. I started shifting my group's research at NYU into proactive imaging to see if there's any way we could drive down false positive rates, any way we could use prior information to help us. I started advising a company called ESRA that does proactive imaging. It was one of the early ones in this space that got a fair amount of attention. And then Ezra last year was acquired by Function, which is a proactive health company that now does large panel blood tests and regular imaging in otherwise asymptomatic members. And when Ezra was acquired, that was the time I thought, okay, time to put my money where my mouth is. I think it's time to figure out how we use imaging in concert with all of these other tests proactively as a kind of early warning system for disease.

SPEAKER_00 13:35

So tell me more about how function health is working. Like what is the general workup? Like when a patient walks in, what kind of workup are they going to get? And what can they expect from some of these biomarkers and these and these MRIs? Like are you using them in tandem or what's going on?

unknown 13:48

Yeah.

SPEAKER_02 13:49

So the basic premise is that in order to be proactive about your health, you need to get your biology online. In this modern era, bits and bytes can be interpreted by all of the algorithms that are being developed, and they can be interpreted in context of one another. So the first step when you join as a function member is to get an entry panel of blood tests, 160 blood tests, including a lot of the familiar ones you get at annual checkups and a whole lot that aren't done, sort of digging into details of your lipid composition and other things that are predictive of cardiovascular health or brain health or um or other aspects of your health. And then also on entry, the idea is you get a baseline MRI, which is a multi-station MRI, head, abdomen, pelvis, which is not really whole body MRI. I think whole body MRI has gotten a bit of a bad rap out there. Um what this is, is actually sort of state-of-the-art active surveillance protocols for key cancers that you might be worried about and various other conditions in those three areas of the body strung together. And then you can also get any number of other tests if you want, but the idea is with those baselines, we have a starting understanding of what your health state is. And then from then on, we want to guide you to figure out what you need to do next. So I think of it almost like developing a GPS for your health. It knows where you've been, it understands the kind of health landscape, and it helps guide you to avoid the outcomes you don't want and to achieve the outcomes you do, like living a hundred healthy years.

SPEAKER_00 15:46

Yeah, makes sense. So then are you is it most like a consultation type thing? You're constantly checking in and getting recommendations from the company about what you should do next or what supplements you need to take, or how's that work?

SPEAKER_02 15:56

So it's a work in progress. This is a you know a startup company that uh is only, I think, uh four or five years old at present, but growing like crazy. At the moment, you get a detailed clinician report on your blood panel, what each value means, what it might you know mean for you, um, and also on your imaging.

unknown 16:18

Okay.

SPEAKER_02 16:18

And these are radiologist reports done by human radiologists, digested by AI to make them a little more understandable for the member. Down the line, though, the medical intelligence lab that I joined function to lead is going to be devising more and more advanced forms of guidance, prediction models that tell you your risk of developing a certain disease in a few years' time, other things that we can use to really guide your health. So this is those initial reports are just the beginning. We'll also be able to tell you when we want to see you next. Maybe when does it make sense to have a coronary artery calcium scan? Yeah. You know, uh uh all of those things.

SPEAKER_00 17:04

As an neuroadiologist, I've watched imaging become more powerful than ever, but also more fragmented. With more data, more tools, and more systems. It's a challenge to bring everything together in a way that supports fast, confident decisions. Rapid AI helps bring imaging insights, patient context, and clinical workflows into one connected enterprise platform designed to reduce friction and support better decisions across the patient journey. At the center is Navigator Pro, a radiology workspace that prioritizes what matters most, surfaces deep clinical context automatically, and streamlines reporting and collaborations within the workflows radiologists already use. Learn more about it at rabbitai.com. With the whole body screening MRI, like you said, you know, I think there's been you know some good news, some bad news regarding like the whole body MRIs. Like there was a recently, you know, that a PRNUVO case where they're like the all like ICA, you know, high-grade narrowing was just not called. And I think me and you know that a lot of these imaging protocols aren't set to evaluate that really well. You know, that's not like what the whole body MRI is not great for looking at the vessels a lot of the times. And there's sometimes artifact, and you might you might not see anything there, or it's not even in your real cue. So I guess my question is how are people being advised, or like how are patients being told that that what what are the limitations of this whole body MRI? Are we looking for tumors only or anything like that? And how do you account for all these incidental findings?

SPEAKER_02 18:33

Yeah, yeah. I think the first order business is, as you said, to set expectations that this is not a diagnostic exam. And to the general public, those words mean nothing. Yeah. But we understand that what that means is we're not looking to uh find absolutely anything that might uh uh be associated with symptoms you're having, you're not necessarily having symptoms, you're coming in healthy. The idea is really to get a baseline and to cover some of the most worrisome things that we might find. I think one of the challenges that uh is related to the Pranuvo suit is that if you promise to find absolutely everything, yeah, there will be things you don't find because imaging is imperfect and radiologists, human radiologists and AI are imperfect. So I think that initial framing is key. This is a baseline scan to uh rule out some of the most worrisome things. But I think the real magic comes when the scans are repeated. And I think a lot of the debate around proactive imaging comes from this conception that it's a screen, it's a one-shot thing. But actually, we all know from uh radiology that if you have prior images, you can rule out many things that might be of concern if they haven't changed from last time. So I think that's the key. It's then to set the expectations for the first scan, but then to go and get the next scan and interpret that scan in context. Got it.

SPEAKER_00 20:18

So the real value lies in being um someone who's gonna be diligent and follow up with you know continued surveillance scans, because at that point, then if there is some minor change, it's gonna pop up very, very quickly. That's right.

SPEAKER_02 20:30

Basically, the better we know you, the better we can predict your health, the better we can understand the imaging. And I think there's one other piece that to me has been really compelling, which is the com the combination of imaging with all of the other information. Right, right. So in some ways, for me, when the imaging was combined with the blood testing, that was the magic formula that pulled me from academia to industry for the first time. And in fact, some of the research my group was doing a little bit before then was very intriguing and really suggestive. We trained an AI model to take your current biparametric prostate images and predict your risk of developing prostate cancer in five years, for example. And the model on its own, using just today's snapshot, did fine, yeah, you know, but a pretty high false positive rate, like 64% false positive rate. Not that much different from humans because it's hard to predict five years out. Yeah, exactly. Yeah. But then, interestingly, when we fed that same model prior scans, and when we fed it even a small bit of blood test information, PSA, plus some age, maybe prostate volume, we found we could drive the false positive rate lower and lower and lower until it was below 10%. Wow. Okay. So from 64% to 9% just by incorporating Other test information and prior imaging information. And I think that's the key. In medicine, context is everything. Right. Right. And AI, for the first time, allows us to put all of this diverse information in context.

SPEAKER_00 22:13

Yeah, because there's so many things that you we don't have like the research to back this correlates with this thing in imaging. You know, it's it's those two things aren't don't have a lot of crosstalk. Like obvious, of course, you're gonna have elevated PSA when you have prostate cancer, but that's when it's already full blown and you know, things like that. But these little micro changes in whether it's imaging or whether it's the lab values, and then kind of combining that, um, those micro changes result in some significant conclusions by this black box algorithm, you know? That's crazy. That's right. And um, I think it kind of speaks to some of the technology elsewhere, like you know, I think we were talking about the before the interview. There was just an article a couple days ago um about a study, you know, where they're using this new algorithm to try to look at pancreatic cancer earlier um than when it's detected. As you know, pancreatic cancer has a horrible prognosis, often diagnosed way late into disease when it's already spread to other organs, and patients have a very high mortality rate um related to it. So catching disease earlier is is super important, but it's so hard to do that. Um, you know, as someone who looks at pancreases all the time, they all look like little sponges on CT, okay? If you if there's something going on, it's not good. If you're able to see it on CT, that means that the patient's probably pretty advanced. Like we are often taught like, you know, if you see a little thing in the pancreas and you find any single little lymph node that's a little bit round, it's probably metastasized already. And you're just seeing the results of that. So hearing these kind of like, you know, combining medical imaging with the health data to predict disease earlier is really exciting. And then just on its own, this imaging algorithm from this study was able to start detecting early signs of pancreatic cancer just on a non-contrast CT, um, you know, two to three years earlier before it developed, which is just crazy to me. So when you're talking about, hey, what if you had all this, you know, blood work as well, and you throw that into the equation, then you start thinking about how early can we start predicting these things?

SPEAKER_02 24:09

Absolutely. You throw in genomics, proteomics, you establish somebody's baseline risk from their underlying genetic profile, and now you update that with their essentially biology at the time, you're really starting to cook with gas.

SPEAKER_00 24:28

And can you tell me like a little bit about so there's this seems to be a divide these days uh between like what public healthcare looks like or academic healthcare looks like and what private healthcare looks like and the access and the information that you get from new companies like function health and and um you know whole body MRI? Like, but you have to pay to play, right? Like you you gotta you gotta have you gotta have the interest and the funds to be able to get these tests done and to continuously follow up with these scans, you know. Um, so can you talk about a little bit about how you've seen that divide occur and what you see in the future? Are we moving toward one day where maybe we're getting things all combined like this for a regular patient in the ER? Like when's that gonna happen?

SPEAKER_02 25:07

Yeah, yeah, yeah. Really good question. And really, if it were about concierge care for the wealthy, I wouldn't have joined function health. Um, in fact, a lot of my interests and the interests of a lot of the imaging field have been moving more and more towards accessibility. Yes. Low-field MRI, as we were discussing before, getting out to places in the world that's never had imaging. I think one of the interesting things about this moment in history is that costs are going to be driven down. As we were talking about, if you have prior information, if you have context, you can image in a much more accessible and expensive way. So I think ironically, even though right now private health, if you want to call it that way, or direct to consumer health is paid directly by the consumer, I think down the line it's gonna end up being cheaper than traditional health. First of all, I think once we demonstrate the outcomes that in fact we are saving lives and saving downstream costs associated with complex care of advanced disease, I think insurance companies are gonna take note and start covering this because everybody knows that an ounce of prevention is worth a pound of cure, right? Yeah, and insurance companies are very good at weighing the ounces at a pound. Um but I also think our ability to deliver care is going to get more and more effective and cheaper and cheaper the more context we build. Right. So essentially you start out, you know, almost naked like a baby without a lot of data on your health. But as you go through life, I feel like it's this timeline that follows you and accumulates. And all of that can be used to deliver insights more and more cheaply. So I actually see it as a democratization of health, okay, giving people autonomy and the ability to catch things early, quickly, and cheaply.

SPEAKER_00 27:14

That's such a great point. I like didn't even think about it that way. Cause yeah, you know, you're preventing by finding it earlier, you're preventing the development of disease or at least getting it treated early, which, yeah, like you're saying, gonna lead to less admissions, less surgeries, less complications, all that kind of stuff that occurs because we're catching disease too late. Um and that's gonna, yeah, trickle, trickle down to that, to the cost for the consumer. You mentioned um insurance, which I think is a great point too in all of this and how it ties into the situation. Obviously, premiums, everybody's insurance is, you know, health insurance is going, you know, skyrocketing for everybody. Do you see a future or are we working toward a future where you could get some kind of preventative workup like this? And you let's say that you get, you know, a stamp of approval from the company based on your blood work, based on your screening MRI. How does that change your insurance premiums? Is that going to be something that maybe then you don't you're you're not paying as much because they know you have a clean bill of health prior to getting insurance? Or how you you think that there might be something in the future that alludes to that?

SPEAKER_02 28:12

It's interesting. I I think that's a double-edged sword. I mean, I I'm remembering the movie Gattaca, uh, where essentially everything is built on knowledge of your genome. Right. Um but but I do think that we're heading towards a future where really the focus is on preservation of health, not treatment of disease. I mean, people talk about this dichotomy that health care is really sick care. As somebody who's grown up in academic medical systems, you know, I want sick care. Sick care is important. People will always get sick. Right. But I do think there's there will be a shift more and more towards preserving health, and I think that is gonna change the insurance structure. Yeah. Um I think if we have less and less highly expensive advanced care to cover, I do think that premiums could change for the larger body of healthy people. Yeah. And change in a good way, I hope.

SPEAKER_00 29:22

Yeah. I mean, I'm just thinking, I'm like the only thing I can think of is like an analogy to like um, you know, like safe driver discounts, right? Like, you know, you um you have this safety thing in your car that prevents you from doing things. Okay, you're gonna get a discount on your insurance. It's almost like that. Oh, you did a pre-screen and it was clean bill of health. Yeah, it's less likely, or you know, this person's following up, or they're in the subscription-based system. Okay, they're probably good, they don't have to pay as much or something like that. Just extrapolating here, but I'm just curious, like, you know.

SPEAKER_02 29:50

No, I I I think it's definitely a direction we could be going in. I also think, just as you were speaking, I was thinking, it also reflects a kind of change of the role of imaging in our lives. You and I are imagers, right? We we live in and breathe it professionally, but for most people, it's a sporadic thing. Yeah. They want it to be a sporadic thing. Yeah, yeah. They don't want any of to need imaging because imaging is what tells you you're sick. Yeah. Is it any wonder that radiology departments don't get as many philanthropic contributions as surgery departments? We're the people who tell you you're sick, and then we hand you the people who cure you. Yeah. But I think more and more imaging can become the thing that tells you you're still okay. Can be that reassurance, and in fact, can be a more intimate and continuous part of your life. And actually, this is something I talk about a bunch in my recent book, The Future of Seeing, um, when I'm talking about where imaging is heading. Yeah, I think it's not necessarily going to be this episodic thing anymore. I think it's gonna be part of our daily lives. And I think that's gonna be enabled by AI, by all of this context we have. There's something I call the everywhere scanner vision that I've had for a number of years now, which is okay, you go in, you get your initial baseline scans in a high-field three Tesla scanner, you get all your fancy blood tests at a phlebotomy center. But then for the next time you get scanned, maybe it can be in an MRI machine that I've built into a seat.

SPEAKER_01 31:29

Yeah.

SPEAKER_02 31:30

Or a bed at home or in a CVS, because all it needs to do is measure change. And maybe all these wearables we love to wear, you know, my rings, my watches, and so on, maybe they can be on the lookout for change, referenced against this baseline of imaging. Yeah. So that effectively monitoring for change becomes continuous in our lives. We have a safety net that doesn't let us slip through. And anytime, I mean, it doesn't bother us unless something's going wrong. If we wake up and our sensors tell us, you know what, you're not yourself today. They raise a flag, you go in for more imaging, and you catch whatever's happening early.

SPEAKER_00 32:11

Yeah. So we talked about a little bit about, you know, the future of what it looks like, you know, the CVS scanners and things like that. So I'm just curious your take on how things are going to change on the healthcare side of things. Um, we talked a little bit about, you know, may maybe patients are getting these freaking follow-ups. How does the job of a radiologist change in 10 years?

SPEAKER_02 32:30

Yeah. Well, it's interesting. People talk a lot about disruptive technologies that, you know, come and eat the technologies that birth them and so on. I don't see it happening that way in imaging. As we start moving towards proactive imaging, catching things early. Okay, maybe we don't want radiologists to be reading all of those normal scans, those millions of normal scans that are coming through. Maybe AI can do that really well if its job is only to raise a flag. But then there are going to be all of these people who have the flag raised, who then need the follow-up. And in some ways, that's like the perfect incoming patient for a radiologist. High suspicion of disease, you know, a lot of prior context and work up, and your job is to figure out what's really going on. Yeah. So I sort of feel like the job of radiologists could get more intellectually satisfying, not working through 99 normal exams to find the one abnormal, but really using your expertise to figure out what needs to happen. And I think in terms of volumes, it could actually drive volumes for radiologists up. Even if the first scans are being done sort of out in the field, they're going to be all of these otherwise healthy people who are coming in for triage. Already, you know, Jeff Hinton back in 2016 said, you know, we should stop training radiologists because AI is going to do their job for them. Well, now there's a shortage of human radiologists, interestingly enough. I think that might continue. I think radiologists are still going to be needed to digest a lot of this information, aided by AI, presumably. And by the time AI has figured out how to take over radiologists' jobs, it will have already taken over so many other jobs. We'll all have to figure out what else to do with ourselves. Yeah.

SPEAKER_00 34:30

I think that we've we've kind of harped on that throughout the uh you know first couple seasons. I think I almost led the season one with that uh Jeffrey Hinton quote because it it set up a season to kind of talk about how the changes have occurred in the past 10 years where, you know, we're now it's literally the best job market that we've ever had in the history of the field. You can kind of like pick your poison of what you want to do. You can you can say no to this. I don't want to do that, I want to work from home and I want this much vacation. You know, it's it's it's dealer's choice right now for for radiology. Um, and I think I talk to a lot of medical students, and we talk about this a lot on the show, where there's just always this uncertainty or, you know, I don't know if it's even gonna be around in a few years. And that, you know, I read this article and they said this, and that this new thing can read the CT by itself in 90s, you know, 90 seconds, and you know, all these things, like these headlines are coming out, and we talk about the bottleneck in training. We're not increasing residency positions, you know. Like if if we start losing interest from the medical student side, then we're in a real predicament here because right now, as as you know, imaging volumes are like this, radiology trainees just like this, stagnant. You know, you have to change, you know, things on the Congress level, things change medical positions, all this stuff has to happen to produce more radiologists. I get worried about like, okay, you lose interest on one side, and then you have continued explosion of imaging, and then we're starting to not have as many experts in the field, and we start delegating tasks that we shouldn't delegate to AI. And that's where I'm I'm concerned about that stuff. But I think that, you know, that's where we really need to work to find the right tools that help us bridge the gap and make sure that we talk about these kinds of things and so and make sure that medical students hear this stuff because I I just tell everybody the job is not going to be the same. Your job is gonna be completely different. It was completely different. You know, I talked to one radiologist, you know, a couple days ago where they were like, you know, hanging plane films when he first started, and you know, all this stuff is a rabid acceleration. It's only going to accelerate even more rapidly as we move toward the future. So, like you're talking about with the blood markers and the biomarkers and kind of connecting the imaging data. I mean, I see a future where in 10 years you're sitting back there and I've got a genomic output on this screen, I've got biomarkers flagged on this screen, I've got, like you're saying, two accelerated scans and another 10 priors because this patient's been screened annually. And now I'm taking all of this data together, maybe using an AI algorithm platform to kind of synthesize all of this data information and making a conclusion that is very, very definitive about what's going on with the patient.

SPEAKER_02 37:07

Absolutely. And I think you're doing it as the expert being fed the interesting cases. Yeah, exactly. Not just the run of the mill. So arguably, you know, at least for a significant time, the job of radiologists should get more and more interesting and rich. Yeah. Um, and I think one important thing, you know, one important message for medical students is if you look back at the history of imaging, it's never stayed still. It's not like we're gonna be doing just the current exams that we do today. Every time a new probe of the universe has been discovered, someone's figured out how to turn it into an image. Yes. So our imaging devices, I think, are gonna get more interesting. We're gonna be able to detect new types of contrasts and so on. I mean, one should never bet against the advancement of imaging technologies. So I think a radiologist's job in the future could be as different from what we do today as, you know, I don't know, uh an old glassmaker before the invention of the telescope. Right. And a modern imager today.

SPEAKER_00 38:23

Yeah. I mean, I I think we talk about the current state of radiology um where people are pretty isolated. Um they're cut off from clinicians these days, a lot of them are working remotely. Um your whole day is just grinding through ER stat, and then once you get time, you go and work on the outpatient list, and that's your whole day. Um, you know, communication is still not great. It's getting improved with new tools. Um, and so I'm already seeing more interaction with clinicians than I did in training, which is crazy to me because I work remotely. You know, I'm like remote most of the time. But all these new tools that are coming out with like, you know, better imaging or better messaging between clinicians and the radiologists, where I can sit there in a text box and chat with the ER doctor live as I'm looking at the scan. And, you know, then we're talking about different features coming up where, you know, we can do a console feature where we can show the ER doctor the images live and be like on the phone with them, be like scrolling through and it's on their screen. It's a screen share thing. So then that produces like a different role of the radiologist. Again, we're moving back to, hey, we're at the at the center of healthcare here, and we're helping guide management rather than the clinician kind of reading this impression from someone in California who didn't do a thorough chart review, and it's not very specific for the patient. Um, so I I see it as an evolution in care, and I think there's gonna be a time where we become very, very central to the role of diagnosis and even more central and even more clinician-facing, because, like you're saying, the normals have been screened out. We're yeah, we're getting more patient volumes and stuff like that, but the stuff that we're seeing is actually abnormal, it's actually complicated, and you really have to use your brain in the way that we love to do in radiology. Like that's what we got into it for, not to just grind, grind, grind and feel like you're you know providing minimal value and stuff like that. You know, this is, I think, going to improve burnout for people so much, um, and is going to expand the role. Like that intern year you do before radiology residency, it's gonna matter again, I think. You know, where you're starting to look at labs and you're starting, you're like, you know, starting to put things together again, whereas before it's just like, I don't know, man, it's black and white here, correlate clinically, you know.

SPEAKER_02 40:38

Here, here, here, here. No, I think it's interesting. I think as radiologists, we've gotten as a field, we've gotten used to operating a little bit in isolation in our closed reading rooms in the dark, yeah, looking just at the information that's fed to us because so much information is flowing in. Yeah. But actually, I think there's a real opportunity for radiology to reconnect with the rest of the clinical world because our scans don't occur in isolation. Yeah. Right? We need increasingly to know what the broad context is. I I like to say also that radiology is great with space, but we still have a lot to learn about time. Yes. Integrating information over time is something that is very fulfilling and very valuable for patients, as we've been talking about. But old-fashioned radiology kind of got used to looking at what was in front of us. And now I think we get to step back and look more broadly at the timeline for an individual, at this broad range of information that we get to integrate to help in the diagnostic process.

SPEAKER_00 41:56

So we talked a lot about all the pros of this preventative imaging and kind of getting all this data. Um, I think one of the things that comes up, especially with whole body MRI, is incidental findings or you know, getting this lab value that may indicate you're going to develop this cancer and then the anxiety that comes with it for patients. And and and people see that as a big downside for this stuff. So can you talk a little bit more about that and kind of how you guys are balancing that?

SPEAKER_02 42:21

Absolutely. No, and that's a real practical question. I think in some ways, imaging has kind of an image problem because it's associated with anxiety because imaging is associated with having a problem. Yes. The minute you even go into an imaging device, you know, people start sweating. Um I think there's the practical question of how to handle incidental findings, and then there's the broader question of anxiety. I think it changes if this is your first time or if you've been here before, if you've if if this is a follow-up imaging exam, right? So if it's the first time, I think you know obviously we need to call those findings. And what Ezra, the imaging arm of function, does is assign a score from one to five that's modeled after the rads scoring system. So pi rads, lie rads, bi rads. There's an e-score, which is from one to five, which indicates sort of the level of concern. Got it. So if there's a finding, but it's not really a concerning finding, you'll get an E-score of one or two. If there's something that is glaring and really needs immediate action, it's five. And the in-between area, which is the muddy, uncertain one, just as it is in all of our practice, is three. Three, okay. So right off the bat, there's kind of an indication of how concerned you should be. And I think then the other important thing is once you're coming back now for your follow-ups, now we can say, you know what, it looks the same as it did last time, no interval change. And so I expect the incidental findings are going to drop more and more. As people got it. Right. So there's caution up front and then progressive improvement the more you come in. So I think this is also, people aren't used to it. The notion that you should get imaged regularly, I think that's key because the more you are imaged, the fewer incidental findings, the fewer false positives, the fewer items of concern. The one last thing I'll say about anxiety, you know, and people often approach me with this. Well, you know, what if there's uh a small cerebral aneurysm that discussed that's discovered, and now I have this time bomb in my head, and what am I gonna do? Well, it's an interesting question down the line as to whether that should be called at all before we have some context. But I think down the line, imaging can be the thing that tells you you're still okay, not just the thing that tells you you're sick. So in some ways there's some anxiety about finding something initially, but then if you get regular imaging and you see, you know what, it's the same. Yeah. I think of it almost like our sensory system, right? You step out onto a street and you look over and you see a car coming at you, you get this jolt of adrenaline, which is very uncomfortable. But then you step back, the car goes by, and you're still alive, and the adrenaline drops down again. I don't mind a little bit of adrenaline as long as it's not a constant pump. Yeah. Down the line, I think with imaging, when there's a finding, it'll be that little burst of adrenaline that allows you to step back from the car, but then it'll kind of just dissipate into the distance.

SPEAKER_00 46:05

Got it. No, that makes sense. So I think I think what I'm understanding is like I think the most utility lies in, like we were talking about someone who's going to have serial exams, you know, the one-off full-body MRI is probably not an amazing idea just because it can raise these questionable three findings. And then if you're not getting regular image, then maybe you do lead, you do, you know, result in unnecessary work up, unnecessary biopsies, and some complications from procedures and stuff like that. I think that's what people get worried about.

SPEAKER_02 46:34

That's right. And I think we tend to think of imaging screening, even this idea of whole body MRI as a one-shot thing. Oh, I'm okay. Yeah, yeah. Well, no, I mean, health is dynamic.

SPEAKER_01 46:48

Yeah.

SPEAKER_02 46:49

You want to follow it over time. And so I do think that the whole proposition changes if it becomes a regular part of your life.

SPEAKER_00 46:58

Okay. Yeah. 3D reconstruction has become essential to modern imaging, but the process is still slow, manual, and fragmented. Techs spend valuable time recreating reconstructions, outsourcing information to labs, and they introduce delays and inconsistencies, and radiologists are left waiting for the information they need. There's a much better way to do things. Lumina 3D from Rapid AI fully automates high-quality reconstruction of the head and neck, eliminating the manual burden on your team and the unpredictability of outside labs. Consistent outputs are proven to reduce reconstruction time by more than 24 minutes and enhance radiologists' diagnostic accuracy. Lumina 3D gives the imaging teams better results delivered to their workstations without the weight or the variability. Learn more at rapidai.com. And so you mentioned like the anxiety of, you know, let's say someone does have a small aneurysm in their brain and they got the full body MRI and it came up. You mentioned how, you know, one of the cons is, you know, the anxiety that comes with it, the patient's worried about this stuff. I mean, I get that side of things, but I know if it was me in that scene, I think a lot of people would agree. Like, wouldn't you rather want to know about this finding and just have it followed up? And especially in the day and age where, you know, we have algorithms like you know, like rapid aneurysm and all these different detectors that can actually look at the morphology aneurysm and really accurately detect like little morphological changes that may suggest that is prone to rupture. You know, when we looked at these before, it was just pixels on a screen. I'm like, eh, it's a millimeter bigger, like in this plane when I'm measuring it this way. It was much harder to track things. And I think that's a key point in this is that, yeah, you can get some anxiety from the findings, but when our tracking and our AI technology is so much better in terms of quantifying the interval changes, then like you're saying, that reassurance becomes so much more valuable in a year.

SPEAKER_02 48:56

Absolutely. Really, what we have now is a combination of the best in artificial intelligence, machine intelligence, and human intelligence. Yeah. So now you have tracking algorithms, you have neurosurgeons who can be monitoring and can be ready to intervene if ever it gets to the stage where it's necessary. Right. So we have this whole safety net that hasn't always been there. And so if there's a safety net there, why wouldn't you want to know? Right.

SPEAKER_00 49:26

I think a lot of it is just, you know, like we're doing here is having these conversations and making sure people are informed about where we are in the technology and and kind of what this means for for you now and what this means for you in five years, you know. Um, I know just from our discussions and everything that I've learned recently, I'm probably more inclined to do preventative screening just because I know that, you know, if I do follow it annually and I know how accurate this stuff is, I can get a pretty clear picture of what's going on. At the very least, get some semblance of reassurance. And I think as we progress and and and the technology improves, and more importantly, the information gets out there that this stuff exists and this stuff works pretty well, um, I think it's gonna take over. I think there's gonna be a huge push for people, you know, lay people trying to just get this stuff, get into the pool, whether that's an annual follow-up or a follow-up every three years, just to provide some reassurance.

SPEAKER_02 50:19

Well, and I think there's a way in which, interestingly, technology is recapitulating biology. Because if you think about the evolution of our brains and our nervous systems over time rather than just dealing with the moment, right? Maybe a single-celled organism swimming around, sees that chemical gradient, wants to go find food, more and more they became prediction machines. And so when we walk around in the world, our brains are constantly predicting what's coming so we can avoid harm. Well, in medicine, arguably, we're still a little bit back in that responding in the moment phase. But down the line, wouldn't it be great if medicine can be proactive, predictive, and personalized? Yeah. So that we're really looking at what's going to happen, not just looking at what has already happened. And I think that's a fundamental change that is going to be good for doctors and for patients everywhere.

SPEAKER_00 51:24

Yeah, that's a that's a great take. Almost like a sixth sense for danger via, you know, blood screening and imaging.

SPEAKER_02 51:30

But if you think about it, imaging is just an extension of vision. Yeah. Right? It's the visual sense, but with the capacity to see inside our bodies. Right, right. In a lot more detail. So I think we already have a sixth sense. It's called medical imaging. And, you know, we need to generate the data that shows the downstream benefits and so on. And I'm working hard on that. But um, but yeah, it's it's fun, right? I mean, I it's amazed me that in all of the years I've been working on it, imaging has never flattened out. I mean, never, never asymptotes. So like every time I think, oh, this is gonna start getting boring. And then, you know, a few years ago, AI, geez, huh?

SPEAKER_00 52:13

I mean and I mean the other thing we didn't really talk about was just like uh even in residency, there were new imaging modalities, or not modalities, but imaging techniques coming out all the time. Yeah. Like, you know, these biomarkers that go straight to the cancer, and now you have to read the scan of like, you know, these like what are they? Like, I can't even remember the bio, like the the targeted agents and stuff, but like, you know, you give this injection, this nuclear medicine injection, and it goes straight to the cancer, and now there's a new scan, and it has its own nuances and interpretation. Like, oh no, you don't look at those values, and this tissue is going to take that stuff up, so you gotta ignore that stuff. But I'm like, I'm looking at the scan, I'm like reading it that day, and I'm like, what is this protocol? Yeah, yeah, we're doing this new thing with Lutitia 43 and you know, whatever name it is like for the molecule, and then it's like a new scan. It's like, yeah, so here's uh a paper about like how to look at these, and so you know, when people talk about you know just the future of imaging and stuff, they're they're very much looking at a silo of like what we currently do and not factoring in that with all this new technology, with all of this functional health data and and these biomarkers that are getting discovered, don't you think there's gonna be new neck nuclear medicine imaging that's like a tag scan and it's gonna go straight to the thing where you can get even more information or certain drug treatments that are nuclear medicine and imaging tagged, like all these things, like and then you have to develop an algorithm that does that to screen those normal populations, and it's going to continue to occur.

SPEAKER_02 53:40

Well, and listen, one thing we didn't even talk about was the discovery potential of the data from all of this proactive monitoring. Right, right? I mean, the UK Biobank has 500,000 people who've been followed, you know, with some regularity, with a set of tests and with some imaging. But if you start getting millions of people getting tested and scanned and monitored regularly, that is an utter gold mine. And you start discovering, oh, you know what? These markers, yeah, actually, they're very predictive. And so we should be doing them in everyone. You know, we discover the next Apolypo Bs, we discover the next diffusion MRI, you know, we figure out how to put all of it together. Um, I think just the capacity for discovery in this modern world of big data is remarkable. Yeah.

SPEAKER_00 54:30

And we talked a little bit about this, about you know, the acceleration algorithms and things like that. How, and you you mentioned that the costs would likely go down. I think that's something that's really exciting for the future of imaging because it's lead to more imaging, lead to better patient outcomes. Do you have any like idea of what that might look like or what kind of cost difference we're looking at? Or, you know, or like a proportion, like let's just say that an MRI now takes like five minutes, you know. I I can't I can't imagine it's gonna cost the same. Is it half? Is it a third, maybe? Like what are we what are we looking at?

SPEAKER_02 55:03

I mean, generally speaking, we know that imaging costs scale with the slot time, right? So if you're you if if you've blocked a 15-minute slot for imaging, that's half as expensive, at least on the technical fee. Yes, yeah, um, as a 30-minute slot. So you can imagine that the technical fees can actually drop down in proportion to the imaging time. Yeah. And if we can image 20 times faster, you do the math. Yeah. Um, so it's a pretty exciting possibility. There will still be interpretation time if there's a professional radiologist reading it. If it's an AI triage, maybe not so much. Yeah. So I really think the cost balance can shift a lot.

SPEAKER_00 55:52

Yeah. And and that's that's really exciting just because when we're talking about, you know, how patients need to be plugged in on a like a yearly basis or, you know, following it up consistently, I think the biggest barrier right now is access and cost, right? And and and the ability to do this stuff. So if we're moving toward a future where scans are faster, scans are way cheaper, and you have much more data and and and um you know, data to show that you have good outcomes if you do follow your health regularly, I mean, then it becomes like a standard screening thing. You know, everybody pays this much, your scans only this much per year, and you get instead of like a yearly annual physical, you get your yearly annual imaging, blood work, and that's how we monitor people and only flag what needs to be flagged.

SPEAKER_02 56:39

I I love that picture, the notion of the annual physical, including broad blood tests and imaging. Why not, right? I mean, our annual physical is basically dictated by tradition, but we live in the modern world where we have access to all of this information. Why should we not have that as part of our annual physical? Now, I do want to come back to one thing that I think you mentioned, which is also the kind of burgeoning just of the amount of imaging that's being done. And even if imaging individually gets cheaper, if you're imaging that much more frequently and that many more people, is that a burden to society? That's something that I get asked a lot as well. There's a fascinating study by a group that explores benefits programs for companies that basically looked at the costs, did a cost-benefit analysis for employee screening programs to you know, for early detection for, say, cancers. And they found a fascinating thing. They found that because of all the downstream costs that are saved from expensive care, ultimately it's dramatically more favorable financially to have these screening programs in place. But that benefit doesn't kick in until something like year three. Because initially you're triggering all of these other follow-up tests, all of the downstream, you know, costs from that those initial findings. Only later do you save all the costs because you have saved, you know, the cost of care. Yeah. So in fact, I think we have to think about it with a little patience as a society, too. There might be an increase in cost initially as we're finding all of these things. Yes. But as we're actually curing people or preventing advanced disease, then all of the cost savings are going to start flowing back.

SPEAKER_00 58:40

Aaron Ross Powell What are the next steps in getting preventative imaging and preventative health deployed on a larger scale in healthcare systems? Is it is it a cost issue? Is it a technology issue? Uh it's probably multifactorial, of course, but I just want to get your thoughts.

SPEAKER_02 58:58

Aaron Powell That's a really good and hard question. I think I know it's not a question of demand, because if you look at the growth in the number of users that have signed up for a function, yeah, clearly there's a hunger out there. And people want to get this sort of control over their health. I think partly it's an attitudinal change. I think people are used to, in our field, certainly, the bad old days of mall scanners and false positive findings. And everyone is very cautious. I I have a number of radiologists come to me in in complete good faith and say, you're doing more harm than good. And to which I reply, Well, wait, what harm am I doing? If I'm finding things early, okay, maybe I find a few too many things. But lives are also saved. So you know, I think I think people have to get used to the capabilities of our new technologies. And they have to get used to the idea that imaging is not the last bastion of diagnosis only. I think that will help. And then I think certainly the costs of scanners, the difficulty even of citing a modern high-field MRI machine is a big obstacle. Ezra, in order to build up its network, partnered with, I think, 70 different imaging networks with 200 different imaging facilities around the country.

SPEAKER_00 1:00:40

Wow.

SPEAKER_02 1:00:41

So talk about a heterogeneous complex imaging network. Yeah. Um down the line, if we could build cheaper scanners with less infrastructure surrounding them for interval scans, for example, I think that would be a real enabler as well.

SPEAKER_00 1:00:58

Yeah. When uh are we going to see portable low-field MRIs out in the field, do you think?

SPEAKER_02 1:01:04

Well, the amazing thing is they're already there, not in routine use by radiologists, but you know, hyperfine. As its 64 millitesla MRI machine. We've we've we we had uh have a couple at NYU. Um they're not going to produce images that a neuroradiologist is uh going to want to stare at for too long in all cases. But again, for interval scans, we just need the AI to connect them to the previous images. So I think the technology is already potentially available. We really what we really need to do is develop all the algorithms that take your old data and bring it into today's scans. And that's something we don't have, right? We have pack systems which store your old data, but it's not like they're accessible to today's MRI machines easily. So I think there's some infrastructure we need to build uh in order to give our scanners memory. I actually like to think of this whole thing as scanners with memory.

SPEAKER_00 1:02:09

Yeah, that's a good way to think about it. Yeah. And if it's auto-integrating all of this stuff into the new context when it's scanning you again, then again, you're gonna have better pictures, better reconstructions, all that kind of stuff and a quicker scan.

SPEAKER_02 1:02:20

Absolutely. So I think there's some interesting technological enablers that people are starting to work on. And then there's just everybody getting used to this new role that imaging has in the modern world.

SPEAKER_00 1:02:31

Yeah. Wow. So I'm just so excited about the future that. I mean, uh it sounds like from everything you've been saying, we are going to see skyrocketing imaging, more accessible imaging, probably cheaper imaging, and more effective imaging in the sense that we're able to tell you much more with any given scan.

SPEAKER_02 1:02:52

Absolutely. I think I think that feels to me inevitable. And I say that partly because I'm bullish about the future, but partly because for this book, I went back and studied the history of imaging from the early oceans when we first developed eyes on through telescopes and microscopes and, you know, uh X-ray machines, MRI machines. One of the things that became clear to me is every time we extend our vision, we invariably expand our minds and we become capable of things that we had never imagined. I mean, how would the inventors of the telescope back in, you know, the early 1600s have imagined that there would be telescopes orbiting Earth out in space, staring off at things that you know it's taken light millions of years to get to us to see. Yeah. Right? I think why should medical imaging be any different?

SPEAKER_00 1:03:55

Love that analogy.

unknown 1:03:58

Dr.

SPEAKER_00 1:03:59

Sodicson, this has been incredible. I really appreciate you coming here. This was such an insightful conversation, and I'm so looking forward to seeing what happens next on the preventative imaging side and preventative healthcare side.

SPEAKER_02 1:04:10

I love talking about this, and I have really loved talking about this with you. Thanks so much. Thank you so much.