Rethink Imaging

EP 39 • March 5, 2026

Total Risk in CT Imaging: Why Clinical Risk Outweighs Radiation Risk

Featured Guest

Dr. Francesco Ria, PhD

Medical Physicist and Imaging Scientist • Duke University

Listen Now40 min

Also on:

Let's Talk

Beyond dose

Ask us what your imaging data can actually show you.

Start a Conversation

Share This Episode

Subscribe to Rethink Imaging

CT safety conversations usually start and end with radiation dose. Dr. Francesco Ria, a medical physicist and imaging scientist at Duke University, joins the show to widen that frame with a concept he calls total risk. His point is that radiation is only one of the risks a patient carries through an imaging decision. A missed diagnosis, a delayed treatment, an unnecessary repeat scan, and an exam that never happens because someone feared the dose all carry risk too, and in many cases that clinical risk is the larger one.

Ria works through what total risk looks like in practice and why “how low can we go” is the wrong question to organize a protocol around. He shares research showing that across patient groups, clinical risk outweighed radiation risk by a wide margin, and he sets the numbers next to each other: surgical mortality sits just above 1 percent, while even pessimistic models put radiation-induced cancer risk near 0.04, roughly thirty times smaller. The takeaway is not to push dose higher. It is to optimize for diagnostic confidence, treat patients as patients rather than phantoms, and give physicists, radiologists, and technologists a shared way to weigh one risk against another.

Host

Chris St. John

Host, Rethink Imaging • Imalogix

Featured Guest

Dr. Francesco Ria, PhD

Medical Physicist and Imaging Scientist • Duke University

Watch the Episode

Key Takeaways

Total risk reframes CT safety as a balance rather than a single number. Radiation is one input alongside missed diagnoses, delayed treatment, repeat scans, and exams avoided out of fear.
Across patient groups in Dr. Ria’s research, clinical risk outweighed radiation risk by a wide margin, in some cases by around 400 percent.
Public attention often inverts the actual numbers. Surgical mortality runs just above 1 percent while radiation-induced cancer risk sits near 0.04 in pessimistic models, yet far more research focuses on the smaller risk.
Phantom measurements like CTDI describe how a scanner performs, not how a specific patient is affected. Organ sensitivity, disease prevalence, age, and the diagnostic task all change the real picture.
Dose optimization should protect diagnostic confidence. The goal is the right dose for the right patient and the right clinical question, not the lowest possible number.

Full Transcript

[00:00:00] Chris St John: Across all demographics, clinical risk outweighed radiation risk by at least four hundred percent in some groups and some much more. When I first saw that number, I was taken back. I was pretty shocked. It seemed high. Welcome to Frame by Frame Rethink Imaging, a podcast by Imologix. Here, we explore the intricate world of medical imaging, aiming to dissect the field and inspire both professionals and curious minds alike. I’m your host, Chris St. John. Hi, everyone, and welcome back to Rethink Imaging. Today, we are joined by Francesco Ria, PhD, and he’s a medical physicist and imaging scientist at Duke University where he works with the Carl e Ravin Advanced Imaging Labs and the Center for Virtual Imaging Trials. His research focuses on CT optimization, radiation risk, and diagnostic performance with particular emphasis on how imaging decisions impact real patient outcomes. Doctor Rhea is known for challenging the assumption that lowering radiation dose automatically means safer care and for developing quantitative frameworks that balance radiation risk against the clinical risk of missed or delayed diagnoses. He’s the lead author of a recent communications medicine paper proposing a total risk approach to CT optimization, reframing imaging safety around patient outcomes rather than just dose alone. Welcome, doctor Ria. Thank you for being here.

[00:01:25] Franceso Ria: Hi. Thank you for the invitation. It’s really a pleasure to be here.

[00:01:29] Chris St John: I’m super excited to talk to you today. Without without sharing too much

[00:01:33] Franceso Ria: Uh-huh. You

[00:01:33] Chris St John: know, someone close to me recently had a health scare, and we went through a bunch of imaging recently, you know, multiple head CTs, MRCT. And I found myself thinking about you because, you know, as the person there in the ER supporting them getting a head CT, all I could think was, turn up the dose. I want a higher dose.

[00:01:56] Franceso Ria: Okay. Alright. That sounds like the battle I have with my daughter because I like loud music, and she doesn’t. So

[00:02:04] Chris St John: You know, it was just funny. So for decades, CT optimization has been, you know, very much about lowering radiation dose. And in this recent paper, you write that optimization has been one-sided and strongly biased towards radiation risk only. And I’m curious, what was the moment or the realization that made you question whether that focus was actually protecting patients?

[00:02:29] Franceso Ria: Yeah. Thank you for the question. That’s really interesting and also made made me think a little bit of the last few years. I don’t think there is a there was a moment when we realized this. I always use the plural way because it is a team effort. I think that before we start thinking outside the box, we need to realize that there is a box. And so there was some sort of discomfort that I always felt in together with my coworkers because there is always this apologetic interaction with other medical professor, with patients. I am sorry. We are hurting too much. We’re giving you those. But then we are part of medicine. And in medicine, I mean, we are not allowed to do anything if we are not doing more good than harm. So how does radiology fit into the medicine framework if we are only talking about about risk? And we know that what we are doing is helpful, is useful. I mean, all the other medical specialties ask the diagnostic imaging to provide essential information for the patient, for the diagnosis. So there is something good that we are doing, but we are never able to put a number on it, and so we just decide to give it a try.

[00:03:40] Chris St John: Yeah. And so, I mean, like, radiation risk dominates how everyone is thinking about CT safety, and I believe you call that dominance contrary to common assumption. Why do you think that radiation fear became so central in the imaging culture compared to diagnostic accuracy or patient outcomes?

[00:03:57] Franceso Ria: Okay. Allow me to make, like, a little digression. I come from a small town, and it’s very easy, uh, to be called with some names. You know? Like, the people give you nicknames or everything. I don’t want to pass like the guy that advocate for a high dose. I want to be the guy that advocate for the right dose. Right. That’s what we are trying to do. And sometimes it’s higher, sometimes it’s it’s lower. So I think that the fear of radiation comes from a real fear. I mean, it’s something that it we cannot see, we cannot smell, we cannot touch, but it’s still harmful. And, I mean, can be harnessed to build bombs and can be harnessed to cure tumors, you know, with radiation therapy. And then there, of course, there is a whole Hollywood and comics, and they always use this this concept of radiation to give superpowers, of to give mutations, whatever. So, I mean, it’s easy to be to be afraid of something that we don’t really know. But also I think that in our field, because we use them every day, there is some sort of legacy. In the past, traditional CT or radiography or mammography, usually the the relationship between radiation dose, radiation exposure, and the quality of the image was pretty straightforward. Because I increased the dose and I got better images. Because the system, like, it was manual. I have to select what kind of current or tension I want to use, and the outcome was, I mean, the side by the operator. Right now, there are a lot of automatic system implemented in the machine. So we don’t know we lost this traditional relationship. So it doesn’t make sense anymore to talk only about dose because that’s just part of the equation. But for the same dose, for different patient, we can get two different image qualities. And so say this dose is higher or lower doesn’t really tell us, uh, uh, much about about the overall procedure.

[00:05:45] Chris St John: And so you point out that, you know, we built all these models for radiation risk, and there’s been, quote, a paucity of studies assessing clinical risk. Practically speaking, you know, what risks were we not seeing before you started, you know, this quantification work?

[00:06:03] Franceso Ria: Yeah. Every word that we’ve brought in that paper, and it was carefully weighted. So first of all, we decide to follow this clinical risk approach because there was a need to compare apple with apple. So let’s make an example, you know, like, let’s talk about surgery. We know that the surgery procedure is justified if the risk of dying without getting the surgery procedure, it’s higher than the risk of dying during the procedure. So you even risk to benefit needs to be characterized in terms of risk to risk. That’s the way that we can compare output to output. So if the we decide to do the same in radiology, and we say, okay. What is the benefit of radiology? The affected diagnosis. Therefore, what is the risk associated with the misdiagnosis? And that’s what we are trying to compare with the radiation risk. There are this is just only one risk that we are adding into the traditional narrative that was based only on radiation risk, but there are other risks that we can incorporate in the future and thinking about the risk of contrast. Contrast media is there half of our exam to tier performing contrast media. And it would be nice to put some numbers there also to see how many adverse events we have, what is the kind of dose that can be tolerated, can we optimize the dose. And then there are other risk that, I mean, every other medical procedure every other medical specialties also also include, and this risk can be associated with the history of the patient. You know? Even investigating the same pathology, if the history of the story of the patient is different, can have different different kind of risk. So definitely, once that we got outside the box that I I mentioned at the beginning, and we say, okay, radiation risk is only one part of what we are trying to quantify, then, of course, I mean, there are a lot of other risks that can try to incorporate in this model.

[00:07:48] Chris St John: Yeah. I mean, it’s it’s so interesting to me that the risk to risk comparison is such a subversion of even, like, of the language that we use societally to talk about risk in general. Right? Like Yeah. We are constantly saying risk benefit, risk benefit, risk benefit. You know, I I don’t think at any point until, you know, at RSNA, when I heard your talk, was I actually hearing anybody talk about risk to risk?

[00:08:13] Franceso Ria: You can apply this to every kind of human activities. You know? Like, I drive every day from my home to work, uh, and there is a risk of getting in a car accident. So I can reduce the speed, and the risk reduces. But probably the risk of being late to my meeting and getting, like, my post upset increases. You know? Like, there is always a risk to risk comparison. So do I want to take the risk or spend five hours in the car to make, like, two miles, or do I want to take the risk of speeding a little bit more? You know, it’s it’s it’s always a risk to risk in in everything we do, even even the lottery. I mean, they sell the lottery to us like, oh, you you invest a dollar, you can win a billion. But, really, they should tell us that you they we should compare the risk of investing $1 with the risk of not winning, which is much higher than the percentage of win. So everything can be framed in this way, and probably it’s it’s more understandable because, again, we are comparing apple with apple, the same units.

[00:09:12] Chris St John: Yeah. For sure. But, yeah, we just, like, we we that’s not the way, I think, that we have just, you know, grown and learned to think about it. So I I just I love this this reframing of benefit as risk, specifically defining the clinical risk as harm from misdiagnosis or nondiagnosis. So was it important to put radiation risk and clinical risk into the same units instead of comparing dose to image quality or or other detection metrics?

[00:09:39] Franceso Ria: This depends of what we are trying to do. So if we are trying to see the performance to assess to measure the performance of the scanner, then we can use traditional measure. We can get a phantom measure the CTDI board. You tell me if I’m using words that the audience may not like. But we can take a phantom measure measure an exposure.

[00:09:59] Chris St John: I think I think we can all get behind phantom and CTD eyeball on this podcast. Okay.

[00:10:03] Franceso Ria: Fair enough. Because we physicists are quite of nerves, so sometimes we like like no man pressure very much too much. So we can get a phantom, and we can measure the exposure in the phantom. We can measure the image quality and perfect. I can assess how the scanner it it’s working, but patients are not phantom. So the CTDI for the patient is not enough because the same CTDI to different organs is related to different risk. So the same image quality can bring to different diagnostic readings. And we saw it, for instance, now in photon counting, which is the topic of the hour. So the main concern with photon counting is that the noise magnitude of the images is very high. But there are some task like, uh, in thorax exam, you know, in the chest, for instance, where the resolution is so high and that even this extra noise doesn’t really bother the radiology. The radiology are very happy, while the same noise can be a problem in other anatomical regions. So measure the image quality without talking about the task that we are investigating is is not enough. It’s just it it is good to assess the performance of the scanner, but, again, patients are not phantoms, and pathologists are not inserting a phantom. Like, everybody’s different, and we should take

[00:11:14] Chris St John: this into consideration. Forgive my naivete here, but so you’re what you’re saying is, like, different scans, like a head CT versus an abdomen, pelvis or something, different noise thresholds are going to be different in order for those images to be to be readable.

[00:11:29] Franceso Ria: Yes. I mean, yeah, different task. Yeah. Depends on what you’re trying to see. I mean, again, we can ask the radiologists about this, but trying to investigate kidney stones, it does not require this kind of, you know, fancy images because, I mean, it’s a piece of rock into the human body. So we should be able to see even if the images, you know, are not are not perfect. But if we are trying to detect a subtle lesion in the liver, then we really need, like, a lower noise or other kind of image quality feature. It depends on what we are trying to what we are trying to investigate. And this is also something that we cannot stress enough. So many times I read paper and say, oh, yeah. We increase we reduce the we reduce the CTDI. Therefore, we’ve patients have increased. That’s I I don’t know. Maybe, yes, but I don’t know because I don’t know where do you give the CTDI, to the head, to the chest, to the abdomen. I don’t know what kind of organs were exposed. I don’t know the age of the patient. Like, if we don’t contextualize this kind of analysis, again, we cannot draw conclusions and make sense.

[00:12:29] Chris St John: Yeah. Absolutely. I mean, like, you know, not to tie back to my personal story, but, like, after the second head CT and they continued to say everything that looked fine, I’m like, it’s ahead. Crank it up. Like, I’m not I’m not on here trying to trying to preach to the world that we need to crank up our doses. But just, you know, it’s very top of mind right now. Right?

[00:12:50] Franceso Ria: Yeah. Yeah. It brings me since you are bringing me an example, this conversation is bringing me a story from a few years ago. I think it was, like, 02/1516. We did a study within Duke University and my previous Italian institution, CTI from Milan. And we asked the patient what they know about radiation dose. It was the topic of the hour that I’m and then I have no idea why, but we put this question into this survey. We asked the patient, do you prefer having the radiation dose of your exam reduced or the image quality improved? It was, like, very straightforward. It’s like, no context whatsoever. Just tell us there were, like, seven other patient. It was a a significant amount of of patient. They split almost fifty fifty. They were, like, 45, 55, I think. So patients understand this issue that we are facing every day. And if they are coming to the hospital because they want hands worse, the same way that if they are sick and they are taking an antibiotics, they want the right dose on antibiotics. They don’t want too much. They don’t want too low. They want a dose that is safe and is effective. That’s what we should try. So we should give a a radiation dose. We should use a radiation dose that is safe for the patient, but at the same time, it allows us to answer to the to the question the clinical question that we’re investigating.

[00:14:06] Chris St John: Yeah. Okay. Well, I might I might be getting myself a little bit into the weeds here, but I wanna keep moving. So in your model, total risk, I believe, is literally just radiation risk plus clinical risk, which sounds surface level quite simple, but the inputs going into that are really not. Right? You have disease prevalence and false positives. The AUC, the area under curve, life expectancy loss. I’m I’m curious about, you know, how those different variables surprised you and how you saw them, you know, affecting total risk. Well, the thing that we did we never considered before, it was the ethnicity,

[00:14:46] Franceso Ria: the ethnic group, the race. And people from Europe will forgive me. Race is not a it’s a sort of a bad word in Europe by United States. I mean, it’s medically appropriate. But, uh, yeah, that was something that we never consider. But because the side effect, the adverse event associated with radiation is usually induced cancers. And different ethnicity, different races have different incidence of cancer, but also different survival for cancer. So if even when we’re just trying to talk about radiation risk only, we need to incorporate this this demographic and this ethnic and trace information into our model. So this was the thing that we never considered before. There was to us, it was like, I don’t know, interesting, uh, because we knew that AUC is basically how good the radiologists can read an image, and we knew that life expectancies were. But, I mean, incorporating not just the age and the sex, but also the race of the patient was something that surprised us a little bit. Yeah.

[00:15:47] Chris St John: Yeah. And this follow-up may be, like, a little bit delayed here, but just still still talking about, you know, like, the reframing to benefits back to risk. I’m I’m just kinda curious. Is this there’s a piece of it that feels very mathematical. Obviously, like, this is this is what your paper is about, this this quantification. But, you know, there’s also, like, a mindset shift that needs to be happening as well. Correct? Yeah. I I guess, where do how do you balance those two, like, math versus mindset? That’s that’s a great quest. It’s a great question because

[00:16:17] Franceso Ria: we are physicists that we are men of science, at least with our science background, and we work in medicine that is mostly an art. You know? Mhmm. So we need to try but, yeah, we knew already that something has to be done in terms of switching our focus. I think the mathematic provided the language to communicate this new model. Because I can go to people and say, listen, you need to change the dose or to increase the dose, reduce the dose, but they want to know why. And I can tell them, oh, listen. You’re going to see the image better or the image is worse. That’s fine, but it’s still qualitative. You know? It’s based on preferences, based on experience. But but if we have some numbers that we can show, I think that’s really what we need to support this shift in the mindset that you that you talked about. And let me tell you something. When we showed the results of our primary studies to simulation, whether to their radiologists, they are they are quite happy because they say, yes. We know that we are doing good things for the patient. Thank you for showing us how good how much good we are doing. You know? Instead and so it’s I don’t know. It’s optimistic message that we are trying to broadcast as well.

[00:17:29] Chris St John: Yeah. Yeah. And so you simulated a million digital patients undergoing abdominal CT for liver cancer. Yeah. I’m curious why y’all went with liver cancer and why, you know, such a large simulated population was well, I mean, you know, the the bigger the sample size, the better. But, you know, why liver cancer and why this specific population size?

[00:17:52] Franceso Ria: I have to say that the main reason is why because at the beginning, we had a small dataset of liver cancer patients that were acquired with several doses. So it was easy to start. No researcher will ever say that that’s the reason why we do some studies, but sometimes we have to use whatever we have in our home.

[00:18:07] Chris St John: Hey. Dig out. You take the win. Right?

[00:18:10] Franceso Ria: Exact exactly. So, yeah, we had this small data set of patients with different doses, and so I start looking at these things here. And then we realized that, I mean, liver cancer is is the third cause of cancer in in United States and in the world, so it’s very it’s very timely. It is also hard to diagnose because liver is soft tissues and the lesion usually are very subtle. So it’s it’s a good task to measure this, to try to to test this this model. And then there are a lot of data in the in the literature, in the the in the National Cancer Institute, they have a lot of data about mortality, about incidents, about survival. So it was a safe, uh, first step. But, of course, I mean, the the mathematical equation is the same. We can apply the same model to every kind of task that we want. We just need to change the the parameters. But, yes, liver cancer was, like, a safest and also a very impactful pathology to to to investigate for several reason. Why a million is a round number?

[00:19:15] Chris St John: Uh, Yeah. Yeah.

[00:19:16] Franceso Ria: But, also, if you start thinking, I mean, we need to simulate different radiation exposure condition. So we went for, like, zero to fifty milligray of CTDI just to include everything. We needed all the ages from zero to 100. We need male and female. We need four different ethnicities. I mean, if you start putting all this combination of permutation together, you get to get a big number. So, yeah, a million was was a safe number to show some robust results.

[00:19:43] Chris St John: Yeah. And how I mean, how do you even like, what is it even like, I you know, I understand that you simulated digital patients, but what does that actually look like in practice?

[00:19:52] Franceso Ria: Uh, a spreadsheet. Yeah. Okay. No. No. No. I’m kidding. Yeah. Several spreadsheets. No. Yeah. How does it look in practice? It looks like a statistician and physicist working together to try to put all these variables in, uh, into some kind of software and making this software to work, and then spit out some some data. So there are like, every computer things, we have some inputs that are the age of patient, uh, the sex, the ethnicity, the the the presence of cancer or not or whatever. And then we have some output at the end by, yes, it’s mostly computer based at this level. Informed by clinical practice, of course, because, I mean, we put them number that makes sense. We work with the radiologists as well.

[00:20:38] Chris St John: Let’s dig into, you know, your your results a little bit. Right? So across all demographics, clinical risk outweigh radiation risk by am I correct that it is at least four hundred percent in some groups and and some much more?

[00:20:50] Franceso Ria: It yeah. That’s yeah. Some groups. Yes. Mhmm.

[00:20:53] Chris St John: I mean, were you when I first saw that number, I was I was taken back. I was pretty shocked. It seemed high. But, you know, did was that something you expected, or was it surprising to you as well?

[00:21:04] Franceso Ria: No. It wasn’t surprising. Uh, first of all, then, again, then these are simulation. So, I mean, these are not absolute numbers. We need to go to test and securing a page. But, I mean, the math is robust. Again, I don’t know why we treat radiology differently than any other medical specialty. FDA approves drugs and medical devices based on safety and effectiveness. You know? If we are doing something in radiology, it means that we are doing more good than our. So we were expecting already that the clinical risk outweigh the the radiation risk. For sure. Otherwise, I mean, I don’t know what we’re doing here. You know? Like, it’s not as, uh, so this this was expected. Yes. You’re right. Seeing it in a curve, in a plot, it was something fun to see Yeah. For the first time. I I I don’t want to lie. It it it was fun, but, also, that’s what we were hoping for. And, again, we think that that’s what we have to do in practice.

[00:22:01] Chris St John: Yeah. I mean, and and, you know, and it’s always exciting.

[00:22:04] Franceso Ria: Excited to add. I don’t have anything exciting to add,

[00:22:06] Chris St John: but I mean, that’s exciting enough for me. Nothing nothing excites me like like spreadsheets and percentages. Let me tell you. Alright. Careful. So, obviously, clinicians are not a monolith. Right? Everybody has different opinions. Like, there’s Yes. Lots of opinions about dose and cumulative dose and this and that, and, you know, everybody’s constantly having these internal conversations. But, like, do you would you expect many clinicians out there to to really, like, have an understanding of what that ratio could possibly be. I mean, now they’re gonna have access to this paper and stuff, but I’m I’m curious what you think, you know, your your average clinician might expect this to look like, uh, when you know, you don’t you don’t have to answer for them because you’re not them. But I figured No. Yeah.

[00:22:50] Franceso Ria: I’m not talking. I don’t want to ask for for for for MDs. But, again, I’m pretty sure that there is a a common consensus about it. You know? I spoke with gastroenterologist, for instance, about the same project, and we thought that this model was so cool. And we spoke with the gastroenterologist, say, oh, this is quite simplistic because when I press when I ask for a radiology exam, I’m considering the patient history, the patient background. I go back to generations. You know? Like, they put a lot of consideration into it, but still they know that when they are asking an exam or a procedure in general is something that is going to be is going to be beneficial for the patient. So I’m pretty sure that there is common not pretty sure. There is for sure a common understanding that, uh, the clinical benefit outweighs the risk. It’s also in radiology, not just another medical medical specialty. But, again, if I can piggyback a little bit on on your question, I I I think that there is a little bit of overemphasis probably in radiation risk in in in our field. So recently, we presented our estimate this this study. So for fun, we went to look what is mortality risk from surgery. And it is a little bit above one percent, which is reasonable. Then we work we we look at the most pessimistic study in radiology, and we saw that the risk of having an induced cancer, not not dying, but having a cancer Right. In the other even with the most pessimistic model, it’s zero point zero four. So 30 times smaller. Then we went into the literature, and we say, okay. How many paper last year talked about radiation induced cancer? 35. How many papers talk about surgery mortality? One. So let me say it again. Surgery mortality is 30 times higher than radiation induced cancer, and they talk about surgery mortality 30 times less than in radiology. I think that we are being overapologetic, probably. And I don’t I mean, it’s always good. It’s better saying the sorry, of course. But, I mean, let’s just don’t get depressed.

[00:24:58] Chris St John: Well yeah. I mean, you know, and this

[00:24:59] Franceso Ria: this don’t get said. You know? That

[00:25:02] Chris St John: is the nature of the press. Right? It it is it is much you know, it’s much more financially viable to publish stories about things that scare people than to sit you know, than to publish a story that says, like, you’re probably fine if you get a CT. Like, who’s gonna like, how do you turn that into clickbait? Right? How do you drive ad money from saying, like, you know, getting a CT is probably okay?

[00:25:27] Franceso Ria: Yeah. And, I mean, when when we say at the beginning that it is a legacy to talk a lot about radiation dose, it is really a legacy because the, like, regulations always push towards radiation dose monitoring. How many radiation dose monitoring softwares we have in the market? Dozens. How many image quality monitoring software we have? Two or three, I think. So there is really, you know, a mismatch, a disconnect between these, uh, these two things. But I don’t see any other medical specialties having session and sessions about reducing the risk of dying. You know? I don’t go to a surgery conference, and they have sessions about, yeah, they they reduce the risk. I mean, that’s that’s the goal of everybody, but they are not so exclusively focused on this.

[00:26:06] Chris St John: Yeah. Alright. Well, so, I mean, like, let’s build on that. Right? So Okay. You know, in your results, radiation risk averaged about point zero zero nine deaths per 100 patients, right, while clinical risk averaged point zero four five. So how should we interpret that difference in real word world terms, and what does that actually mean for, you know, a patient sitting in a scanner? Yeah.

[00:26:29] Franceso Ria: Do you want politically correct answers? Do you want the real answer?

[00:26:34] Chris St John: Obviously, I want the real answer, but, you know, if we if we got you know, maybe start correct and then get into the truth. Yeah.

[00:26:41] Franceso Ria: I mean, if I’m writing a grant to say, oh, this paved the road toward the personalization of care and, therefore, toward the minimization of risk, whatever. That that’s what we always said. The real answer is we don’t know what it means for the patient. No. It’s fair. Because all the theory, all the medicine is based on experience, on population. I have no idea what’s gonna happen to the single patient. You know? And a patient I mean, a a a patient need a a knee prosthesis, and the doctor say, okay. Fine. You do this surgery, and you’re gonna be fine. And then after two months, the patient go back and say, listen. I cannot work. And the surgery say, yes. There are complications. Complications is a fancy word that we use in medicine, but then unknown. We don’t know. You know? Like, even a drug is approved for marketing, but the clinical trial never ends. There is phase four of the clinical trial, which is post marketing surveillance. What what does it mean? It means that every patient that is taking that trunk can react differently even if we have years of data of million of patients that that, um, that use these drugs. Human bodies are different, so we have no idea what’s gonna happen to it, but we can run from experience. I know that ibuprofen is effective ninety nine percent of the time, and I want to get a drug that is effective ninety nine percent of the time. Instead, that drug is effective fifty percent of the time. You know? So what we try to show with this 1,000,000 simulation things is that for the general population, again, it’s better to focus on improving the quality of the image. It’s better to focus on improving the diagnosis, and it’s better to focus on giving the radiologists better images. Right. For the single patient, I have no idea. But overall and let me add a a number that always blows my mind. Yeah. Life expectancy in The United States but across the world in the last seventy years. So the last time I checked, grew by other other I think, like, thirteen years, something like that. So, basically, every year, we are giving two more months of life expectancy. And that’s to me, it’s impressive. Every year, we live two months longer. So what we are doing makes sense. Medicine makes sense. We are giving good results. We are advancing the care. We are advancing the life expectancy. So we are working in the right direction. Then, of course, human body are not machines. Even when I go to the car dealer and and take my car, sometimes they make mistakes, you know, and these are machines. So human body are a little bit more complicated, but we are definitely going in the right direction.

[00:29:06] Chris St John: Yeah. So one of the most counterintuitive findings is that in over ninety percent of cases, the CT dose that minimized total risk was actually higher than the dose patients are typically receiving. Let’s talk about how, you know, in in some cases, increasing doses is often reducing overall harm.

[00:29:23] Franceso Ria: Yeah. And, again, I mean, I cannot stress this enough. This is liver cancer. This is a simulation for a broken wrist. This scenario will be different, You know? Of course. Because, I mean, the images will look different. We don’t need all these dose. So this is just a case of liver cancer, and these are simulation. Yes. We found out that for ninety percent of the cases, we can increase the dose and get overall risk that is lower for the patient. Now in most of the case, the difference was mathematically perceivable. Like, it was ten percent, five percent, six percent. We can argue until tomorrow if it makes sense to increase the the the dose to reduce the the the the risk by five percent. This is a topic for regulators. So it’s not a topic for luckily, it’s not a topic that that I need to answer. But, yeah, the message that we are trying to to broadcast is that there is room for optimization. Luckily, we are not alone in this because, for instance, ICRP in publication one thirty five few years ago clearly stated that heat radiation dose is low enough. Image quality should be the main focus. And they brought in the ICRP that optimization can result in an increasing of dose. Can. Again, I don’t want to be the guy that advocates for higher dose. I want to advocate for the right the right dose. But, like, playing with radiation dose should not be a taboo. It’s a tool that we have. I mean, uh, we have this tool of looking inside a person without cutting the person open. Let’s use it in a way that that is affected and that makes sense.

[00:30:48] Chris St John: Yeah. So, I mean, I I I’m glad you you brought up the ICRP guidance.

[00:30:52] Franceso Ria: Is it

[00:30:53] Chris St John: You know, that that optimization sometimes includes increasing dose to preserve diagnostic information. Why do you think that nuance has been so hard to, you know, take into a daily radiology practice?

[00:31:06] Franceso Ria: Yeah. I think it’s a lack of data, to be honest. Mhmm. When I came at Duke, like, over ten years ago, I found out that they were already measuring image call, at least noise magnitude in all the patients that underwent CT. And to me, it was mind blowing because Duke was the only center in the world. And we are still leader in the world about image quality, uh, estimation. It’s not easy, but luckily now with the new computer system, we can do this quite no. I don’t wanna say easily, but in a way, that mean without meaning supercomputer. So, yeah, there was this lack of data. And so we play with the data that we had. And radiation dose based this where this data because regulations require to monitor radiation dose, and so we try we try to use it. Also, remember what we said at the beginning that in the past, radiation dose could be used to predict the image quality. So it it it I mean, there was some kind some kind of legacy that we that we carry on with us. But, again, once that we start seeing that the Georges’ world is not only dimensional, but there are more dimensions, then, of course, we can, uh, we we we can try to to how can I say to personalize everything better? Uh, if you want to talk to personalize, at least we can talk about more task specific optimization. Why do I need to use the same abdomen protocol for everything that is in the abdomen? You know? Like, the RT. Again, we don’t do that, but there is room to make more nuances or to create more task specific protocol. Again, if I have a patient with cirrhosis of a specific ethnic group with some specific risk factors, I know that the risk of having liver cancer is higher. I want to be sure that I can see it. So perhaps I can be a little bit more aggressive. A child without any history, maybe I want to be more conservative. Again, we need to and now we have the tools to fix. Like, in the past, it was impossible. Now we have the informatic. We have the computer power. We have the information system that can really support this kind of decision making.

[00:33:02] Chris St John: Yeah. Okay. Continuing on the regulation train, you explicitly reference recent regulatory pressure like CMS scrutiny of CT dose and suggest that an, quote, over exaggerated and obsessive, end quote, focus on radiation could negatively impact patient safety. Mhmm. I’m curious, like, do you expect this message to actually land well with policymakers? Maybe in

[00:33:26] Franceso Ria: a few years. And the way the the the the reason why I’m saying this is because regulation is always late. You know? So the moment regulation is something that we have, that we need to abide by, that we need to respect, that we need to do our our test to that. But the moment the regulation is approved, it’s already late. So if we respect the regulation, we are already ten years late from there. No? I can think about dozen of examples of even the mammography rules or whatever. They were created when radiologists was a different beast. Like, there wasn’t artificial intelligence. There was automatic expert. Yeah. We respect the things because we have to respect that, but is it enough? Probably, it’s not enough. So regulation is one step towards, you know, providing the best the best of care. The regulators, yes. I think that I mean, we are doing this for every medical field. Again, FDA clearly says that and all the regulatory agency that a procedure should be approved or a drug should be approved based on safety and effectiveness. I would like to see this applied also to radiology. So I hope that the next set of routes will ask us to, in some way, to estimate the effectiveness of a radiological procedure.

[00:34:36] Chris St John: Yeah. And I I really like your point about, like, the the delay of regulations. Right? Like, even if we just talk about the new CMS measure, right, which, like, you know, we we we’ve seen in practice some people changing their approach based on this measure. Right? There there have been some good results of it. But at the same time, right, we this whole measure is built around data that I believe is from 2014, is my memory. You know, that, like, pre photon counting. Right? Like Yeah. You know, so we’re Pretty much as

[00:35:04] Franceso Ria: the tech you know, we

[00:35:06] Chris St John: need to be able to keep up.

[00:35:07] Franceso Ria: Yes. I think that also we should try to learn from our story. I don’t wanna say mistake because nobody makes a mistake in purpose. You know? Like but you analyze the story, and you realize that you can do something different. So in yeah. Of course. Let’s talk about radiation dose and then translate this to image quality and CMS. A couple of years ago, we did maybe four years ago, we published a study, and we look how many radiation dose metrics we use in CT. We counted 11 radiation dose metrics that we use in CT. CT, the ideal PSS, the organ dose, different way to calculate effective doses. And we found out that even effective dose, the same metrics calculated different in different ways can give us difference up to 200%. So what are we talking about here? I mean, we should try to establish methods that are standardized and consistent. I think the CMS, if they want to that’s that’s the problem that they’re having. That’s why they are postponing, you know, the implementation because we don’t know what kind of methods we need to use. But we should learn from from our past and say, okay. This is the method, and now we can compare all the cities that we want, all the institutions that we want, all the vendors that we want. Otherwise, again, it’s really difficult to make comparison that makes sense. And allow me to say it’s very easy to cheat. Mhmm. Because I can make a software that show that your noise is almost two. I go make the noise in some air. You know? Like, fine. I will do something good for the patient. Maybe I I don’t think so. No. Like so we want to be as specific as possible, I think, and, um, try to create some some standards that are also robust and based on the current technology. Because if we are implemented today something that is based on 2014, I mean, we are already fifteen years late, and we need to use these things for the next fifteen, twenty years. So I wonder my students in twenty years from now, I’ll say, okay. What were you doing ten years ago when they implemented these things? You know? Like

[00:37:01] Chris St John: Yeah. And the I mean, just the rate at which scanner technology is improving as well. Right? It’s just, like, it’s just getting faster and better, and we have more and more tools to improve those machines. Like, it’s you know, we’re starting to climb vertically up an exponential curve here. So we got it’s hard to stay on top of things. Let’s start to wrap up here. So this framework allows optimization and justification to become, you know, more patient specific rather than just protocol driven. So if this model were start to be implemented clinically, what starts to change first? Right? The the scanner, the protocol, or, like, the decisions made in the conversations to have them?

[00:37:41] Franceso Ria: So first of all, I think that this would be on the nerd side. This would be a great tools for improving the consistency. We always talk about improving consistency, but we want to be consistently good. And different scanners because right now, they’re very sophisticated. You know? They have a lot of automatic tools, etcetera. And we don’t exactly know what happens when a scanner is modulating the radiation dose to achieve consistency, match quality. And they work differently. They have different strategies. But because we can measure the outcome, we can say, okay. Let’s let’s try to change the protocol on this scanner to be closer to the protocol on the other scanner because, I mean, at the end of the day, we want to keep the patient the same the same quality the same quality of care. Also, what I was mentioning before, we can go towards something a little bit more patient or population specific. Again, let’s talk about liver cancer because I talked with the gastroenterologist about this. If we have specific ethnic groups that are more sensitive to liver cancer and they also have a cirrhosis or they have other risk factors, perhaps we can design a protocol specific for this kind of situation. And then on the non nerd side, on the visionary and fantasy world, I I would like to see in the future some kind of shared decision making between practitioners, but also between practitioners and patients. Mhmm. Because patients are smarter than we think. You know? Every once in a while, I get ibuprofen because I have headache or, I don’t know, Iran. I have bad knees pain, whatever. And then I go read on the ibuprofen, drug fat, and say it can cause severe stomach bleeding, heart failure that can be fatal. And people get ibuprofen. We we do our own risk to benefit. Do I want the headache, or do I want the risk of severe stomach bleeding? You know? We do. So I think that, again, this is like a vision, the fantasy world, whatever. But we can perhaps aim to the same approach also in in radiology where we can discuss with the patient and say, okay. This is the risk. This is the risk of not taking the exam. Mhmm. What do we want to do?

[00:39:39] Chris St John: Doctor Francesco Ria is a medical physicist once again from Duke University. Doctor Ria Francesco, thank you so much for for joining us today on Rethink Imaging. It’s been just delightful to have you here today.

[00:39:50] Franceso Ria: Thank you, Chris. And I’m I always follow your podcast. It is really a pleasure to be here.

[00:39:55] Chris St John: Thank you so much. Frame by Frame Rethink Imaging is brought to you by Imologix. Here, you’ll find engaging interviews with thought leaders, experts, and patients sharing stories that showcase the transformative power of medical imaging. To discover how Imologix is rethinking imaging in health care, visit imologix dot com. Be sure to subscribe to Frame by Frame Rethink Imaging on Apple Podcasts, Spotify, or wherever you listen. And from all of us here at Imologix, thanks for tuning in.

Questions from This Episode

What is total risk in CT imaging?
Total risk is a way of weighing every risk attached to an imaging decision, not just radiation. It includes the clinical consequences of a missed or delayed diagnosis, unnecessary repeat scans, and exams that never happen because of radiation fear. Dr. Ria argues that when you account for all of it, clinical risk often outweighs radiation risk.
Does lowering radiation dose always make a CT scan safer?
Not necessarily. Lowering dose reduces one risk, but if it lowers image quality enough to compromise the diagnosis, it raises another. Ria frames every imaging decision as a risk-to-risk comparison, similar to driving slower to lower crash risk while raising the risk of being late. The aim is the dose that delivers diagnostic confidence for that patient.
Why does Dr. Ria say patients are not phantoms?
Physicists measure dose and image quality on phantoms, which are standardized test objects. Those measurements describe scanner performance, but a real patient’s risk depends on which organs are exposed, their age, the disease in question, and the clinical task. Optimizing for patients means adding that context rather than relying on phantom metrics alone.