Rethink Imaging

EP 32 • November 20, 2025

Why Medical Physicists Are Drowning in Level 1 Tasks (And What It Means for AI Adoption)

Featured Guest

Dr. Thomas Griglock, PhD, DABR

Executive Director • LUCIR

Listen Now44 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

Dr. Thomas Griglock returns to Rethink Imaging to explain where a medical physicist’s week actually goes. Drawing on the AAPM task group that sorted physics work into service levels, he walks through Level 1: annual equipment testing, ACR accreditation support, mammography surveys, and the CT protocol optimization the Joint Commission now requires. That work is mandatory, repetitive, and growing, because imaging volumes keep climbing and every new machine adds more to check. Griglock argues the real cost is everything it crowds out: research, education, and the higher-order projects that pulled most physicists into the field in the first place.

The conversation then turns to why new technology keeps stalling at the door. Radiology AI tools are one-trick ponies, trained to catch a single finding while a radiologist scans for everything, and residents still are not taught to use them. Photon counting CT may be the biggest advance in imaging in years, yet only large academic centers have the physics support to run it. Griglock’s fix starts with the education pipeline and with leaders who ask end users before they buy. He also previews LUCIR’s research on large, recent imaging datasets, including data showing hospitals returned to 90 percent of pre-COVID volumes within months, with fewer staff.

Host

Chris St. John

Host, Rethink Imaging • Imalogix

Featured Guest

Dr. Thomas Griglock, PhD, DABR

Executive Director • LUCIR

Watch the Episode

Key Takeaways

An AAPM task group published a service-level model for medical physics work around 2017. Level 1 covers annual equipment testing, ACR accreditation support, and mammography surveys, and CT protocol optimization moved into Level 1 once the Joint Commission made it a requirement for everyone.
Today’s radiology AI tools each detect one finding, a brain bleed or a pneumothorax, while a radiologist looks for anything in the image, including the uncommon. Griglock estimates radiologists use AI in under 3 percent of their work because trainees never learn the tools.
Hospitals were back to roughly 90 percent of pre-COVID imaging volumes within three to four months, with fewer staff, according to data LUCIR plans to publish.
Equipment turnover has a real time cost: Griglock can train someone on a familiar CT interface in about 90 minutes, but an unfamiliar interface takes three to four hours per session.
A seven-figure advanced reconstruction upgrade sat in a box collecting dust because nobody asked the radiologists who would read the images. Two 15-minute conversations would have sent the whole purchase in a different direction.

Full Transcript

Rethink Imaging Podcast Transcript

GuestTom Griglock Pt 2

Host: Chris St. John

CHRIS ST. JOHN 00:00:40 For the last time officially, I want to welcome everybody back to Rethink Imaging. Moving forward, please keep your eyes and ears peeled. Due to SEO and other reasons, we will be streamlining our name, updating in the future, and simplifying—we will just be Rethink Imaging. To all the other frame-by-frame podcasts out there, my hat is off to you. We appreciate you and your contributions to cinema, photography, and other frame-based art forms.

So, for the first time ever, welcome to Rethink Imaging. Today we are joined by Dr. Thomas Griglock, former section chief in radiology and associate professor at OHSU, where he directed the Oregon Medical Physics Program. Tom’s research focused on CT dosimetry, dose optimization, and spectral CT imaging. Now the executive director at Lucier, Tom is doing research with big data to uncover insights from massive and extremely recent datasets. Keep your eyes and ears peeled for future publications from Dr. Griglock. Welcome back to the show, Tom.

I really am so thrilled to have you back. You were the first guest on our show. You really helped us kick this thing off for me as a host. It was a great first experience, and I’m glad to keep talking to you.

TOM GRIGLOCK 00:02:09 Yeah, it’s fun. The podcast is taking off, I think.

CHRIS ST. JOHN 00:02:14 Now we’re big famous on the show. Let me elevate you and get you into some new, fresh ears. For those who haven’t heard Tom’s previous episode, Tom, you’ve spent a lot of your career working at the intersection of imaging, physics, and technology. When you look at how fast new tools are coming—like AI, automations, reconstructions—what is actually changing in a department with this constant barrage of new tech?

TOM GRIGLOCK 00:02:50 Stress levels. I think stress levels are constantly changing in departments, levels of anxiety, things like that. It depends on the department. Everybody, everywhere—what is the movie, Everything Everywhere All at Once—is talking about AI currently. We can start with that a little bit. I certainly don’t want to talk too much about it because everybody else is.

It’s coming faster than we’ve seen it before on the software side of things. Everybody’s been talking about AI for years, and there are a lot of tools out there that people can use. The degree to which people are using them varies a ton. While we’re seeing a lot more of that and there’s been an acceleration in how many of these tools are out there, I think it’s still not being implemented as much as people would think if you’re not in the field or in the clinic.

There are many reasons for that. AI, when we talk about it with medical imaging, especially with radiology and assisted detection, there are many tools, but they’re really all individual one-trick ponies. Whereas a radiologist is trained for many years to look for any large number of things in an image, AI, in the way it’s used now, is trained to look for one specific thing. We don’t necessarily have a tool that can look for all of those things. There’s just too much going on. It can help in certain ways, and it can’t in other ways. That’s kind of where we’re at.

CHRIS ST. JOHN 00:04:40 When you’re saying AI is largely looking for one specific thing, you’re talking about models that are looking to make diagnoses, right? If you’re looking at a bunch of lung CTs looking for a pneumothorax, or you’re looking at a bunch of head CTs for a cranial hemorrhage, is that what you mean?

TOM GRIGLOCK 00:05:00 Yeah, a brain bleed or something like that. A radiologist isn’t looking for one thing. A radiologist is looking for anything and everything that might be in the image, or something that could be surprising or uncommon. That’s the difference, I think, between human intelligence and artificial intelligence. Artificial intelligence isn’t going to look for everything. It’s going to look for one or maybe, on a good day, a couple of very common things, and it likely would not see the uncommon things that a radiologist would pick up on.

CHRIS ST. JOHN 00:05:34 We don’t need to get into the nuance of all these different tools; I’m more curious about our relationship to these tools overall right now. Folks from outside of physics might assume, “Okay, we have new tech. We can figure this out; it must automatically make things easier.” But what is actually happening clinically when another system or feature is getting added into your workflow?

TOM GRIGLOCK 00:06:03 What happens is that it takes a whole lot of time and a whole lot of effort for the most part. If it’s part of a workflow and not a piece of equipment, you have a change to what people are used to doing. There’s a lot of training, implementation, acceptance, anxiety, and stress that goes along with that. Oftentimes there’s a lot of pushback on those things, so if it’s not done well, it’s not going to be accepted or utilized. It comes back to the conversation we had that started the entire idea for this podcast: just because the technology is there doesn’t mean that it’s being used.

CHRIS ST. JOHN 00:06:50 There are still different pieces of technology being installed as on-prem options, right? Non-cloud-hosted software is still getting installed all over the place. So even if you are getting new tech, you’re almost trapping yourself into having to repeat that pattern again. At the rate that this technology is picking up pace, you install an on-prem system, and then it’s out of date immediately to some degree.

TOM GRIGLOCK 00:07:25 To a certain degree, that’s very much the case. When you get into technology in medical imaging—whether you’re talking about hardware or software, MRI tech advancements, CT, or interventional fluoroscopically guided procedures—there are a lot of people that have to agree on what you’re changing to. The radiologist, physician, or technologist is used to using the equipment they’re used to using.

I often go to an analogy of cars. If every day you went out and there was a different car that you didn’t necessarily know—I mean, you have a gas pedal, brake pedal, shifter, steering wheel, and push button, which is great, but all the other interfaces are different—there’s a frustration that goes along with that. Sometimes people don’t even change cars just because they’re so used to having what they have.

If you multiply the cost of that by 10,000 and we’re talking about medical imaging machines, a physician doing cardiac procedures in a cardiac cath lab is very used to the equipment they’re using and how it acts in there. They have a feel for it. They know how to do everything. If you change those things up—change the orientation, the pedal, the controls—there’s a lot of time that has to go into them relearning that in a different way. Even though the technology is probably a lot better, if you change all of the other pieces, which does happen when you upgrade, it’s going to mess with their workflow for a while.

Just because you can do something doesn’t mean that you should. That definitely falls into this category. Yeah, it would be great to upgrade and get all the latest and greatest technological advancements, but does everyone really need that everywhere?

CHRIS ST. JOHN 00:09:42 To add on top of that, a lot of features in a car analogy are more or less ubiquitous. There are details about how you interact with them, but it’s my understanding, especially with scanners, that a lot of this new technology coming out is vendor-specific and proprietary. So not only are you having to engage with new machines that have a different methodology to interact with them, but you have entirely new pieces of tech within that scanner. I think of a car with an auto-driver or auto-parallel park feature: in theory very simple, but jumping into that, it’s not nothing.

TOM GRIGLOCK 00:10:35 From a physicist’s perspective, we’re on these machines once, maybe a handful of times a year, so you get very used to being able to run things a certain way. You learn things from the manufacturer or from a friend or colleague when a new machine or a new manufacturer gets installed, and you get used to that.

What a lot of physicists really struggle with—and I think this goes across anybody who’s going to use this sort of high-end equipment—is that you have to relearn these things when you make those switches. You don’t know how to get it into a service mode. You don’t know where to get certain files that reside somewhere outside of the user interface. I can go in and train somebody on a CT that I’m very nimble and agile with running, and the third time they’re doing it, they’re fine—they’re done in an hour and 30 minutes. On a new machine where I haven’t seen the interface before, it’s three hours, four hours. It takes a good amount of time to do that.

So while a lot of the technological advancements are compelling reasons to upgrade, change, or get excited about, it does take a lot of time and effort on people’s behalf.

CHRIS ST. JOHN 00:11:58 I’m glad you said time. This is more or less going to end up being the conceit of the episode: the amount of time that it takes to actually engage with this tech and the cost-benefit of it. Before we started recording, I was talking about physics work and using the phrase “low-level versus high-level work” when talking about some of this stuff. I don’t believe that’s the right way to talk about it, but can you tell me the right way to talk about it and what that really means in a medical physics context?

TOM GRIGLOCK 00:12:35 The AAPM had a task group a number of years ago—I think they published in 2017, Task Group 301—talking about medical physics workforce services and trying to get a consistent nomenclature around different levels of service. In that publication, they talk about switching over to a level of service model where you go level zero through three.

What we would consider low-level or basic services would be level one. That’s your annual equipment testing, the “easier,” more regulated things that all or most physicists licensed and working in this country do. So it’s testing X-ray equipment, helping out with ACR accreditation, mammography surveys, things of that nature. Those are your level one services, as they call it, and you can put timestamps on those and use them for workforce estimates.

Then there’s level two and level three. Level two is a little nebulous; it’s a bit more of an in-between thing that some physicists have to or might have to do, like serving as a Radiation Safety Officer. A physicist doesn’t have to do it, but can do it in certain environments. There are departments all over the country where radiation safety is in a totally different department altogether, or it may be covered by physicists in the radiology physics section.

Level three is even more nebulous. Those are things like research, education, serving on committees, service in professional organizations, and things like that. Some physicists don’t do any of it, and for some physicists, that’s basically all that they do. You can’t put a monetary or time value on it because if you’re going to serve on a task group committee for AAPM, how much time does that take? It could take two hours a year; it could take 200 hours a year. So getting back to your original question, those are the different levels that this AAPM group worked to define to make sure that we’re talking about things in a consistent manner.

CHRIS ST. JOHN 00:15:25 Something like protocol optimization, what level would that be? Is it two?

TOM GRIGLOCK 00:15:35 Protocol optimization is actually one of the things that used to be level two or level three that is now level one. If we’re talking about CT protocol optimization, it’s level one because it’s required by The Joint Commission as something everyone must do. If you’re talking about MRI protocols, then that would be level two.

There’s not really a monetary compensation model for that. A lot of times, if you’re a salaried physicist working at a hospital or clinic, that’s part of your work where you can estimate some time for it. If you’re a consulting physicist where it’s fee-for-service, you may charge an hourly rate for doing protocol work for somebody, or bid out eight hours a year. How many hours per year somebody spends on optimizing protocols is a black box—it could be two, or it could be 200 if you really get down to it.

The other part of that for the bigger picture of radiology is that protocol optimization involves physicians as well. Radiologists are involved in that, too. So if you’re running a model that takes more time, is less efficient, or lacks tools to help automate those tasks, and you’re doing 20 hours a quarter on protocol optimization, that’s a lot of money. Radiologist time doing that is time that radiologists aren’t in the reading room making money for whoever they’re working for.

CHRIS ST. JOHN 00:17:30 You got to turn out those RVUs. From this—and this might sound extremely naive; I’ve been in and around this field for a year and a half, getting on two years now—I feel like a lot of what I am talking to physicists about these days is that they’re embroiled and potentially even a little bogged down by level one tasks. Is that fair to say?

TOM GRIGLOCK 00:17:56 In many instances, yeah, I think that’s very true. Even if they’re not bogged down to a certain extent, it’s the thing that has to be done. You’re going to get inspected by the state, the ACR, an accrediting body, or The Joint Commission, so those things have to be done. Whoever is paying a physicist to do what medical physicists do is going to be pretty pissed off if they’re not getting their base work done.

It falls into that category, and you get bogged down with it because there’s just a lot of equipment out there. It’s also extremely repetitive. Sometimes it’s interesting, sometimes it’s boring. When I was doing that sort of work, I was good enough at it that I could put headphones in and listen to music while doing it, and it was fine. But it does take a large chunk of your time.

Normally that’s fine if you’re operating in one of the simpler consulting fee-for-service models: you have equipment that needs testing, I have the credentials to test it, you pay me, I test it, and send a report. That straightforward model works fine, but most physicists don’t operate under that model anymore. There are more regulations now. You have to do protocol optimization, protocol committee meetings, tracking, and dose tracking if you are accredited by The Joint Commission.

What ends up happening is time spent on these basic, repetitive level one tasks takes away from all the other things that you can, need to, or should be doing. That creates stress. What’s the definition of stress? It’s the distance between what you want to be doing and what you’re actually doing. The stress level for a lot of people over the last couple of years has increased. Something happened where a lot of people are doing stuff that causes stress because they’re further away from the things they want and should be doing.

CHRIS ST. JOHN 00:21:09 In your opinion, why is it so difficult for physicists to move away from these level one repetitive tasks and focus on higher-order work?

TOM GRIGLOCK 00:21:29 Because you only have so much time. That’s it. You can say there are 40 hours in a work week—or whatever number anybody puts on it—there are only so many hours. Those weeks turn into months, months into years, and years into a career.

If the number is 40 hours, how many of those hours are spent doing these level one activities? While I’m out doing level one activities, other things are happening in the background; I don’t work in a vacuum, so all of these other things continue to pile up.

I was talking to somebody this morning about football games over the weekend and some random things, and we ended up talking for 30 or 60 minutes. I made the statement, “What’s nice about what I’m doing now versus what I was doing a couple of years ago is that camaraderie we shared after a weekend.” Previously, I’d be sitting there thinking, “I’m going to have 50 emails piling up that I have to respond to,” so you would end up not having those conversations. You lose some of the joy of working with other people when you’re not able to do that. As healthcare facilities, physicians, administrators, physicists, and techs have gotten busier over time, it’s created an environment that is inconceivably busy non-stop.

CHRIS ST. JOHN 00:23:25 Tom, hearing all that from you, I know that you left academia. Is some of this part of the rationale for your departure?

TOM GRIGLOCK 00:23:35 Yeah, I think so. It got to a point where I was working with the people I wanted to work with and doing what I thought were the things I wanted to do, and I realized one day that there were so many ancillary things that went along with that that I didn’t want to do them anymore. I didn’t have time to do the things I wanted to do. I have some research publications under my belt, I love working with graduate students, and I still like testing medical equipment—it’s fun, enjoyable, and keeps your brain sharp, especially when you find a problem. But I got so far away from those things because I got pulled into hospital committees and meetings, and I was running a graduate program.

I wanted to do research, but got to do very little. At one point I looked around and thought, “Why am I not able to do some of the things that the people working with me are doing that I would enjoy?” I had the realization that something had to change. Once you get to that point, you can’t put that toothpaste back in the tube. You look at it and go, “This is actually where I’m at. If I’m going to be honest, this is not how I’m going to spend the rest of my time here.” I have an estimated number of years left in this career, on this planet, or in this life, and I made a move to do something I want to be doing.

CHRIS ST. JOHN 00:25:45 What did that experience teach you about how the healthcare system overall values or undervalues time?

TOM GRIGLOCK 00:25:23 Everything comes down to people and culture, so it really depends on where you’re at. What I can say is that after I transitioned out of my previous role into this role where I’m able to talk to other physicists, radiologists, and administrators from all over the country, my story is not unique. Hearing the same thing from almost everyone—not enough time, too many things to do—made me feel like I wasn’t crazy.

I think many organizations or individuals within organizations don’t make the business connection that time and money are intertwined. Making a transition now to the industry side of the physics and medical imaging spectrum changes the way you look at things.

I used to say to my boss—we had a great working relationship and he was super supportive—when he would ask for information on something: “How accurate do you need it to be? Do you need me to sit down, do a calculation, figure out all the variables, and get really nitty-gritty?” Physics nerd mind goes to, “Does my number match the actual true value listed on the back of my physics book? Is the answer 17.4?” Is that how accurate you need me to be, or do you want my opinion on a rough estimate? I can give you my rough estimate now or in five minutes of thinking, and it’s going to cost three dollars of my salary. Or do you want me to collect data, talk to other people, and do calculations? In which case, it’s going to take hours and cost a couple thousand dollars to get the answer. Which one do you want, and how impactful is my answer?

We need to start thinking about things in that manner. A lot of smart, well-trained, kind people whose main goal is helping patients are getting beat up by these things. That’s why you saw an exodus of people. When the world goes crazy, people get out. I don’t know if they’re getting back in, but we have to figure that out at some point.

CHRIS ST. JOHN 00:30:00 We’re at a point where we’ve come back to the beginning in my mind. I opened up talking about the rate at which new technologies are coming and how they impact workflow and time clinically. Then you have all of these physicists and other folks whose time is occupied with level one and mandatory work—something where they effectively have no agency in making that choice. You have patients coming in, and you have to deal with everything that needs to get done.

TOM GRIGLOCK 00:30:35 Keep in mind that not only are we seeing more and more people on a daily, weekly, and monthly basis in these places, we are also doing more things on them because we’re capable of doing more to provide better care. So these things don’t go away—it’s increasing on its own.

CHRIS ST. JOHN 00:30:58 Right, so we almost have this Catch-22 moment of “technology will save us,” but do we have the time to implement it? In hindsight, it’s crazy, but I was very lucky that my restaurant shut down during the pandemic for a few months. After we made it through, I felt quite grateful for that shutdown period because I was able to use that time to rehash my systems, install a new POS, test new recipes, and come up with new cocktails that could be batched more efficiently. I had a period of latency to step back, assess what I’d learned, and have the space to implement those changes so we were a step ahead when we came back. With healthcare, that’s not the case. It’s not like a facility can pause for a year, install new software, and come back way snappier.

TOM GRIGLOCK 00:32:25 That’s potentially the difference between what happened in healthcare versus other parts of the world at that time. Healthcare workers didn’t get to stay home and not go into work. It didn’t get much slower. People assumed that hospital volumes slowed down for an extended period because of restrictions, delayed surgeries, and delayed care. But the data—and we have data we’ll be publishing in the next little while—shows that average hospital imaging facilities and outpatient centers were back to 90% of pre-COVID volumes within about three or four months. That’s not a long time to be down. Keep in mind that facilities returned to those volumes with fewer people because a lot of people quit. It dealt a harsh blow to key patient care personnel.

CHRIS ST. JOHN 00:34:18 I would love to touch on the physics workforce right now. As we’re having this conversation about bringing in technology to help solve these problems, every solution still needs oversight and validation from physicists. How are we supposed to navigate these workforce shortages and staffing models, and what is actually possible right now?

TOM GRIGLOCK 00:34:50 I don’t know that it’s necessarily limiting what’s possible; I think it creates an environment where people just have more stuff to do when things come on board. You don’t get extra time just because somebody wants to implement new tech.

Photon-counting CT is a good example of this. It’s a great technology with a lot of promise, and it’s really the biggest advancement in medical imaging that we’ve seen in a couple of years and will likely see moving forward. But there’s a lot of work that has to be done to make it clinically relevant and feasible. Large academic facilities specifically can probably do it right now, but almost nobody else has the ability because they don’t have enough support to put a good program in. There are a lot more tweaks, understanding, and training required. We need to figure out some of these things if we really want to do what’s best for patients and give them access to these emerging technologies.

The field of theranostics is another example. You take relatively high doses of targeted radiopharmaceuticals, and you do dosimetry to figure out how much of a radiopharmaceutical to inject into a patient to target the cells you want to kill. These are lifesaving and life-extending procedures. We don’t always think of this as technology because it’s not hardware or software, but it is a very important field.

There’s a lot of training that goes into it. OHSU still has a very good program in this. We worked together with a lot of people there to implement it with radiologists, nursing, and techs. It’s a big program, but it’s limited in the number of people who can actually do it, so there are a limited number of places offering it.

If you look at that as an advancement in healthcare technology—which I argue it is—it’s not going to become widespread and available to everyone who can benefit until it gets into the education pipeline. When I was at OHSU, we put theranostics into our graduate program and imaging physics curriculum. Until you start to train people on it, it’s not going to become widespread because graduates won’t know how to set it up elsewhere.

The same thing goes with photon-counting CT. Places that have it now need to educate as many people as possible on how to use it. That’s how technology gets pushed forward. Otherwise, a non-academic hospital, outpatient clinic, or community hospital doesn’t have people who can spend weeks figuring it out. You need to train people on emerging technologies.

It’s the same thing with AI. People go to RSNA every November and talk about AI, but how many radiologists at these facilities are using AI more than 3% of their time? Why aren’t they doing it? Because trainees aren’t trained on AI and don’t know how to use it. Once these tools become accepted and residents are taught how to use AI to become more efficient or more accurate, both benefit patients through higher throughput or better outcomes and earlier detection. But you need to get it into the education systems first.

CHRIS ST. JOHN 00:410:10 Well said. Honestly, for those listeners who are interested in theranostics, pay attention to our Rethink Imaging feed because we have an episode coming out soon enough. So the education starts right here at home.

TOM GRIGLOCK 00:40:25 All of these things feed into workforce issues. There are only so many people who can do what needs to be done. If you want more things done, you either need more time—which we can’t create—or more people, which we can create.

The key comes down to whether we have enough people becoming interested in healthcare. In medical imaging and medical physics, do we have enough people interested in medical physics who want to go into the field or even know about it? Or are we doing something that artificially keeps people out of the field, or not graduating enough people because we don’t have enough residency programs? AAPM has said that for years, and ACR has said that.

This comes down to physicians as well. We need to make sure that if we’re trying to get more people into these roles, there are pathways for them to do it. We need to analyze why we’re not getting enough people, look at admissions rates and barriers to entry, and see if we need to reconsider how we’re doing things. Right now, this is a major issue in healthcare, and it’s important for healthcare to deal with this reality now, not 20 years from now, because there’s a freight train coming and it’s not going to get better on its own.

CHRIS ST. JOHN 00:42:10 What responsibility would you say leadership and vendors share in making sure technology doesn’t just add work, but actually helps give time back?

TOM GRIGLOCK 00:42:28 Leadership has to do their due diligence to make sure they’re asking the right questions of the right people as they do these things. I’ve been in situations where, because of what one person wanted, a decision was made that impacted a hundred other people in the department. I’m not saying it was the wrong decision, but if they had talked to a couple of key people in key roles, they wouldn’t have made the decision to spend money on something, only two years later to say, “Wow, I really shouldn’t have spent all that money on it because nobody wanted it in the first place.”

This was a software upgrade that went awry in a previous life a long time ago, and it was an expensive seven-plus-figure lesson to be learned. Leadership needs to make sure that the things they choose to implement are the correct things in the correct way at the correct time.

Vendors and manufacturers can benefit a lot of people if they put more time and effort not only into creating better products and innovations, but making sure the people using them know how to use them correctly, efficiently, and effectively. People working in the field cringe when I say “apps training” because it can be a pretty painful process. You transition from Manufacturer A to Manufacturer B, you have your first photon-counting CT system coming in, you spent $2.4 million on the machine, and construction for the entire project cost $8 million—and the manufacturer has two people on site for a week, and then it’s, “Good luck everybody, we’ll see you later.”

That sucks a thousand different ways and can doom an upgrade from the get-go. If manufacturers listen to this podcast, take that lesson home with you: be a lot more supportive of the people you’re selling this stuff to.

CHRIS ST. JOHN 00:45:10 I want to throw in one additional follow-up because you were talking about that software upgrade where the decision-makers weren’t talking to the right people further down the line. Who are these decision-makers that you’re talking about, and if they’re listening, who should they actually be talking and listening to?

TOM GRIGLOCK 00:45:44 When it comes to hardware or software, you need to talk to the people who are going to be using the equipment. You need input from the technologists and the managers of that equipment—not from everybody, because you don’t need a hundred different voices, but you need to see the problem from a couple of different points of view to ensure your single point of view isn’t blind to something.

In the instance I was talking about, an advanced reconstruction upgrade was purchased, implemented, and went through applications. But nobody really talked to the radiologists who were reading the images reconstructed in that way, and they just hated the images. So we had all this great technology sitting in a box collecting dust at a very high price tag. If two 15-minute conversations had taken place, the entire thing would have gone in a different direction.

CHRIS ST. JOHN 00:47:09 Nobody has enough time, but take those two minutes! Thank you so much. This has been our first episode of Rethink Imaging. This has been Dr. Thomas Griglock, executive director at Lucier, former section chief in radiology and associate professor at OHSU. Tom, thank you so much for joining us back on the show. It’s truly fitting that you’re on the first episode under our renamed title—episode one once again.

TOM GRIGLOCK 00:47:37 Thank you, Chris. It’s been great.

Questions from This Episode

What counts as a Level 1 task in medical physics?
Level 1 is the mandatory, regulated baseline: annual equipment testing, ACR accreditation support, and mammography surveys. CT protocol optimization now sits in Level 1 too, since the Joint Commission requires it of every accredited facility. These tasks can be time-stamped, which is why AAPM uses them for workforce estimates.
Why is AI adoption in radiology so slow?
Current tools are trained to find one specific thing per model, while a radiologist looks for everything that could be in an image. Griglock puts radiologist AI use at less than 3 percent of their time and says that number will not move until residents are trained on the tools during their education.
How should leadership buy technology that gives time back instead of adding work?
Talk to the people who will use it first: technologists, equipment managers, and the radiologists reading the output. Griglock describes an expensive reconstruction upgrade that went unused because radiologists hated the images, a lesson two short conversations would have prevented. Vendors also owe buyers far more than two applications trainers on site for a week after an 8 million dollar project.