What are the biggest challenges within diagnostic radiology in Europe today? And what future trends will have the greatest impact over the next five years? To answer these questions, we brought together a panel of radiology experts and thought leaders from four countries in a roundtable discussion, following the ECR 2021 Online Congress and Exhibition. Here follows a summary of the key takeaways.
Prof. Dr. Aslak Aslaksen, Radiologist and Head of the Department of Radiology at Helse Bergen, Norway
Dr. Johan Henriksson, Radiologist, Director of Medical Imaging Services at Södersjukhuset, Region Stockholm, Sweden
Prof. Dr. Mark A. van Buchem, Chairman of the Department of Radiology at Leiden University Medical Center, the Netherlands
Dr. Tom Newton, Consultant Radiologist, Clinical Director for Radiology at East Lancashire Hospitals NHS Trust, United Kingdom
The roundtable was moderated by Hans Lugnegård, Global Product Manager Radiology Workflow, with co-host Lisa Lindfors, Global Product Manager Radiology Imaging. They were also joined by Fredrik Häll, VP Product Management.
While representing four different European countries and health systems, the panelists are in clear consensus from the start: today’s biggest challenge in diagnostic imaging is the ever-growing workload for the radiologists. But what are the drivers behind this massive growth?
Prof. Dr. Mark A. van Buchem, Chairman of the Department of Radiology at Leiden University Medical Center.
To start off the discussion, Prof. van Buchem takes a jump at it: “The biggest challenge we’re facing is the sheer number of reports we have to make. It’s a worldwide problem for radiologists—the number of examinations being requested has been rising every year over the past decades. In a way, it’s based on the growing number of advanced imaging modalities that provide much more data, and they keep generating even better patient outcomes. I mean, it’s hard to neglect all that information if you have access to it, for example before a surgery.”
Prof. Dr. Aslaksen agrees and adds that oncology is clearly becoming a larger part of the whole burden, as is the emergency clinic. “I completely agree that the workload for the radiologists is the biggest challenge, especially in advanced cancer care with more complicated examinations and a lot of follow-up exams of all the new targeted drugs being introduced,” he says. “We also see that the multidisciplinary team (MDT) meetings that have been introduced over the last five to ten years have increased the workload for our radiologists a lot, and also increased the need for cross-discipline collaboration. Another perspective is the emergency clinicians—they have a very urgent need for imaging examinations, especially a growing need for CT examinations, for things like stomach pain, stroke, and other diagnoses.”
It’s a heck of a lot of cancer work. All radiologists in our department have a subspeciality interest, but 90% of us will report general oncology scans simply because it’s so much work.
Dr. Tom Newton, Consultant Radiologist, Clinical Director for Radiology at East Lancashire Hospitals NHS Trust.
The situation in the UK seems to be no different, as Dr. Newton describes how their inpatient workload for CT has undergone a very gradual increase. “The oncology work is what is driving most of our demand for CT,” he says. “Not so long ago, with prostate cancer patients for example, we wouldn’t even consider doing a CT scan to follow the progress and treatment response. Now, these patients are getting scanned regularly. The scan intervals used to be once a year, then once every six months, then once every three months—and now, sometimes every six weeks as they are starting on new therapies. It’s driven a lot by the new drugs that are coming out.”
“It’s a heck of a lot of cancer work,” Dr. Newton concludes. “All radiologists in our department have a subspeciality interest, but 90% of us will report general oncology scans simply because it’s so much work.”
Dr. Johan Henriksson, Radiologist, Director of Medical Imaging Services at Södersjukhuset, Region Stockholm.
Dr. Henriksson brings another perspective into the discussion. “I agree with all that’s been said so far, but another aspect I want to bring in is that imaging is driving costs at quite alarming levels, at least here in Region Stockholm,” he says and describes how the hospital’s CEO looks a bit afraid every time he walks into his room, fearing that he will ask for “yet another machine.” He continues, “I mean, we’re using more and more advanced imaging because it’s good for the patients, but the outcome or gain in terms of population health is hard to measure from a reimbursement perspective, which is required to justify the investment needed. At least here in Sweden, the reimbursement for the care episode lags behind the resource allocated per care episode. We need to better inform and educate our surrounding colleagues and the decision makers about the importance and benefits of medical imaging and why it has to be accepted that it costs a lot of money.”
We’re trying to give clinicians better information about the value of radiology and also pinpointing the problems with overdiagnosis and overtreatment. It’s a fine balance between funding, the necessary examinations and treatments, and the danger of overdiagnosis and overtreatment.
Prof. Dr. Aslak Aslaksen, Radiologist and Head of the Department of Radiology at Helse Bergen.
“On the topic of funding, I was personally very inspired by a paper that was co-published by several radiological societies (ESR, RSNA, ACR, etc.) in December 2020: Radiology in the era of value-based healthcare,” Prof. Dr. Aslaksen adds. “Here at Helse Bergen in Norway, we are now quite inspired by the so-called Choosing Wisely campaign and we are trying to implement the ideas that are driven in this campaign to reduce the workload. For example, we’re trying to give clinicians better information about the value of radiology, but also pinpointing the problems with overdiagnosis and overtreatment. It’s a fine balance between funding, the necessary examinations and treatments, and the danger of overdiagnosis and overtreatment. We see this, for instance, in orthopaedic surgery, with meniscus patients and so on. I think we easily can have a 5–10% quality improvement in the referrals from GPs, especially in the musculoskeletal area.”
Prof. van Buchem continues, “It’s a very interesting and challenging question—how to translate the whole value-based healthcare idea into radiology. According to Harvard Business School Professor Michael Porter’s famous value equation, value is calculated as the patient outcome over the money spent during the medical trajectory. For us to tease out what the influence of medical imaging is on the final patient outcome is very hard. But where we could really do an easy job is on the cost side. There are multiple angles to that—as Prof. Dr. Aslaksen mentioned, the Choosing Wisely campaign is one approach: we should make sure that we do not perform unjustified exams. But also, the whole efficiency of our workflow is extremely important; we can directly contribute to increasing value by increasing efficiency and thereby reducing costs.”
“More scan requests are driving costs; a lot of them are complex scans that take time to report,” Dr. Newton fills in. “You can’t sit down and report 20 CTs in a session if they are all CT thorax, abdomen and pelvis with disease everywhere, which needs to be measured and compared to previous scans. I guess this is going to bring us onto the question here in the roundtable discussion, but a lot of that needs to be automated in the PACS of the future. That would help us get up to a higher number of CT scans we could report in a session, and thereby improve the value and drive down the costs in radiology,” he concludes.
AI is not just for image interpretation—I mean, it’s great for that and it’s very important! Maybe less exciting, but at least as important, are AI applications in image acquisition on the one hand, and workflow on the other hand.
Continuing the topic of automation, Dr. Newton explains where he believes things are heading. “Here at East Lancashire, we’ve implemented automatic lung nodule detection that detects on the previous scan, and on the current scan it tells you the volume doubling time. It allows you to quickly conclude whether a lung nodule is benign or not. So, we’ve already got the first example of what I talked about just before. But this kind of solution needs to spread to the rest of the body, picking up bone metastasis instantly, and comparing to previous scans to speed up your reporting times. But it could do the same thing for adrenal nodules, liver lesions, nodes, pelvis, etc. A lot of that could be automated, and I think that is the way things are going.”
“I also believe the vetting process has potential to be automated to a large extent,” he continues. “I don’t know whether this is the same in other countries, but here in the UK, when we get a request through, we have to vet it and make sure that the right protocol is assigned to it, make sure whether or not to give contrast. A lot of that could be automated, and if it was, that work could be devolved to the radiographers. I probably spend three quarters of an hour per day vetting when I do clinical work, vetting the scans—if all that was automated, that would save me a lot of time.”
“Here in Leiden, we actually apply AI at multiple levels already—on image acquisition, on image interpretation, but also on the workflow,” Prof. van Buchem describes. “On the acquisition side, we’re actively working on a project where we use AI to speed up MR acquisitions. One of the limiting factors in the workflow in our department is the duration of the scan time for the MR—if we can speed that up, it will reduce the cost per scan and increase our capacity. So far, it looks very promising—speeding up MR acquisitions by up to four to eight times. We’re now expanding this project from knees, where we started, into other body parts.”
He continues, “In the image interpretation phase, there are a number of techniques that we also use for automatic analysis of the data, of the scans. But I also expect a lot from AI in the workflow. Patient scheduling is a good example where I believe AI will add value. For example, we’ve seen interesting data from a group in Sweden; we always think, in an ideal world, that a time slot for a given examination should be entirely determined by the length of the sequences that you use, in the MR scenario. But they have shown that there’s a wide variation in the time it takes for a time slot for a given patient. For cooperative, young patients it can be shorter, but for an older or invalid patient it can take much longer time. By using AI, you can predict which patients need a certain amount of time for a given examination. That’s a good example of how AI could improve the radiology workflow.”
“To summarize,” Prof. van Buchem says, “AI is not just for image interpretation—I mean, it’s great for that and it’s very important! Maybe less exciting, but at least as important, are AI applications in image acquisition on the one hand, and workflow on the other hand. Also, the easiest one to start with from a legal perspective is workflow. Because you don’t need all kinds of legislation and regulations to pass. There’s no need for that certification.”
“In regards to AI, what we do here in Bergen is a lot of research, especially within mammography screening, to validate that new AI-based tools really do what they are supposed to do,” Prof. Dr. Aslaksen says. “In addition, what we’re also working on is developing research repositories and archives together with Sectra. And also using ‘AI as a service’ for the researchers to build a biobank, so to say, for research studies, so they don’t have to develop their own AI system, but can ask us to do the work for them as a service.”
“Another interesting trend that I would like to bring up,” he continues, “is the increasing collaboration between radiology and nuclear medicine: SPEC-CT, PET-CT, and we also have a PET-MR in our department, especially used for pelvis, prostate, and CNS imaging. The trend of increasing collaboration between oncologists, radiologists, and nuclear medicine physicians, both in diagnostic and therapeutic imaging, is something I think will accelerate in the future.”
Dr. Henriksson takes over and can’t hide his enthusiasm for the topic discussed. “An immensely fascinating subject—the amount of improvement potential in every workflow step that we do is immense, so we could easily talk about this for two days straight,” he says and laughs. “For me and my colleagues, one of the most important things to address is the need for deep workflow integration so that the mechanisms, be it machine learning or whatever, are integrated back-end and don’t come with additional buttons for us radiologists. Everything should be as automatic as possible, even though there are smart orchestration engines in the background that do a lot of magic for you—the radiologists don’t want to see it, they don’t want to choose on/off, they just want it configured in their worklists and as simple as possible. I think that is a significant driver of efficiency that has potential to help us help more patients per hour—combined with trust, of course. You need to know that it actually works.”
“And speaking of automation, one thing that I hope for, going forward, is that we can make the diagnostic system aware of our different competence profiles and link that to the different exams, so that we can automatically route the exams to the right competences or profiles—that would be a big win,” Dr. Henriksson continues. “I have 50 radiologists with very different profiles; some of them are very general, some are very deep in niched areas. Now, they are scrolling through these worklists trying to understand which exams are suitable for their respective competence, where they can create the most value.”
One additional aspect that both Dr. Henriksson and Dr. Newton agree is very time consuming is sifting through the patient history to find all the relevant priors—not least for patients with cancer who have sometimes undergone 30–50 exams in the past. “If there was some sort of automated tool, maybe using really basic AI just to work out to exactly what type of study it was, and then use that to retrieve the relevant studies, I think that would be a quick win,” Dr. Newton says.
The trend of increasing collaboration between oncologists, radiologists, and nuclear medicine physicians, both in diagnostic and therapeutic imaging, is something I think will accelerate in the future.
As Prof. Dr. Aslaksen mentioned earlier, there has been a substantially increased demand for MDT meetings in recent years, which has also affected the role of and the burden on the radiologist. When raising the topic again, he stresses the importance of having a diagnostic imaging system that facilitates the MDT meetings and rounds arranged by the radiologists, to really make them more efficient.
Prof. van Buchem supports Prof. Dr. Aslaksen’s statement that MDT meetings are increasingly taking up more of their time. “They are very time consuming. Often the radiologist is leading the meeting, that’s what I hear from the people in my department—that the clinicians often sit back and relax and the radiologist has spent hours preparing for the meeting and setting everything in order. So, everything that facilitates the preparation, but also the execution, of such meetings from the radiologist’s perspective is important.”
One issue with these MDT meetings that all participants agree on is that it seems common that they lack information regarding other diagnostic procedures and that they need to postpone decisions due to not having the full picture.
“You’re wasting a lot of specialists’ time when things are not ready for the MDT conference,” Dr. Henriksson confirms. “With the gradual maturation of digital pathology, for instance, there’s an obvious opening for a greater level of integrated diagnostics, I believe, at least in the university hospital setting. The rad–path correlation could be done beforehand, in a step before the actual tumor board. A consensual statement between the radiologist and the pathologist or cytologist, before the meeting, is something I think would be greatly beneficial.”
Dr. Newton concurs with Dr. Henriksson. “I’d say that probably about 15–20% of the patients need to be recalled to the next MDT because you’ve not got that full picture. We installed Sectra’s solution in November 2019, and it has really made the efficiency of our MDTs a lot better, I’d say. But I think that rad–path correlation idea is fantastic, to have that before the tumor board meeting.”
You’re wasting a lot of specialists’ time when things are not ready for the MDT conference. […] A consensual statement between the radiologist and the pathologist or cytologist, before the meeting, is something I think would be greatly beneficial.
Before rounding the discussion off, all participants were asked to predict the next five years—what are the most prominent changes that have happened that have made them all more efficient?
Prof. van Buchem is the first to respond. “What I really hope is that AI will make it—as I always say—from PowerPoint to practice,” he says. “That it will be introduced in clinical practice in a substantial way, so that for all the multiple examinations that we do, we’ll have AI solutions for the more mundane tasks that really will help us tackle the workload.”
“Firstly,” Dr. Henriksson continues, “I think we will see better decision support for our referring physicians, connected to the ‘choosing wisely’ discussion that we had earlier. Secondly, we’ve already mentioned shorter scan protocols and efficiency in terms of the acquisition of images—but then also the integration of various machine learning tools in the daily workflow that will increase the sensitivity, specificity, and on many levels. And thirdly, better back-end integration with our EMR, where the report ends up, where it creates value—this can be greatly improved, and I think that will happen within a couple of years.”
“I have two trends that I think we’ll see in the next few years,” Prof. Dr. Aslaksen adds. “One is that in five years, I hope that radiologists are more outward-looking and are collaborating more with the referring physicians. We must not only rely on electronic decision support. The other, which I think we need to discuss more, is the information to and preparation of patients, giving them access to electronic boards, etc. so that we don’t have to send physical letters. Here in Norway, we must still send paper letters to our patients. I hope, in the future, that this information to prepare and to schedule patients will be more efficient and link to other IT systems that we use in the healthcare system.”
“I think when you try to predict five years’ time—and everyone in the room probably knows this—it’s going to be adoption of stuff that’s already here,” Dr. Newton replies. “So, everything we’ve talked about today, if that’s implemented well, all the incremental gains will really improve the workflow. So those are things that need to be worked on.”
“I think we will for sure have major changes, because we need them, just to be able to deal with the increasing numbers that we started the conversation with,” Prof. van Buchem adds as a final remark.
We want to thank our customers for joining us in this post-ECR roundtable discussion on April 14, focusing on European trends in diagnostic radiology. Sectra will arrange more of these virtual roundtables in the future, deep diving into different key themes. Stay tuned!
