The Radiology-IT Relationship: Where It Stands, Where It’s Headed

Radiology has been intertwined with information technology in a generally enthusiastic, if occasionally uneasy, embrace for the better part of four decades. Consider: “The 1981 meeting at McCormick Place showed that future advances in radiology practice would depend primarily on computer technology,” we read in History of the Radiological Society of North America.1 Few would dispute that the year could be switched to any year since without any loss of accuracy. What does this mean for radiologists right now and going forward? To hash out that question, we assembled a panel of four practicing radiologists with leadership-level expertise in IT and informatics:

Kevin McEnery, MD, professor of radiology and director of innovation in imaging informatics at the University of Texas MD Anderson Cancer Center in Houston

Curt Langlotz, MD, PhD, professor of radiology and biomedical informatics research at Stanford University, associate chair of radiology information systems for Stanford Health Care

Tarik Alkasab, MD, PhD, assistant professor of radiology at Harvard University in Cambridge, Mass., and radiology service chief of informatics and IT at Massachusetts General Hospital in Boston

James Brink, MD, radiologist-in-chief at Massachusetts General Hospital in Boston, professor of radiology at Harvard University in Cambridge, Mass., and chair of the board of chancellors for the American College of Radiology (ACR)

[[{"fid":"23192","view_mode":"media_original","type":"media","attributes":{"height":202,"width":604,"alt":" - 201708-rbj-rountable","class":"media-element file-media-original"}}]]

RBJ: As IT has ascended in influence within healthcare-provider organizations, tensions have sometimes flared between increasingly centralized IT leadership and clinical departments. Radiology has not infrequently been a flashpoint, possibly reflecting its status as a technology leader in its own right. Today, on the whole, the radiology-IT relationship seems to have hit its stride. Is that what all of you are seeing?

LANGLOTZ: I think that’s reasonable. I have gone through the transition from an independent IT operation within radiology to a more integrated operation at two different organizations, and both times, it has gone pretty well. I think you need to be careful about the services that remain dedicated to radiology, like desktop support. Our desktops are quite different from other desktops across the organization, so we need dedicated support. But other aspects of the integration are really beneficial. Co-location of our IT staff with the rest of hospital IT has been beneficial for everyone. Now that much of the infrastructure we use is essentially the same as other IT operations, a lot of the barriers to integration have been removed.

ALKASAB: This is actually one of the things we are grappling with now. Frequently, one of the goals of transitioning to more centralized IT is to decrease costs overall. And I think this transition probably can do that. But one of the challenges can be trying to withdraw resources too early. That, I think, is one of the things to be careful of. What would probably make most sense in most cases is getting the integration performed first. You establish a new equilibrium and then go about the process of figuring out how you can achieve savings. That can be better than trying to both integrate and reduce resources, which can overwhelm the process, especially before a new equilibrium is established.

BRINK: I would maybe state it a little differently. When health systems are looking for savings in IT, oftentimes it is to leverage the economies of scale of expanding or consolidating support models among different hospitals within the health system into a single IT infrastructure. And so often when that happens, it is not uncommon for those with strong IT support to suddenly have that support evaporate, owing to the need to prop up the entities that have less strong support. And there is some risk of losing ground, because what seemed like an economy of scale actually turns out to be a drain on those who already have good systems.

McENERY: We have experienced both models, and I would say that our institution now has a hybrid model of both centralized and dedicated IT support resources. Ten years ago, assigned IT personnel were exclusively dedicated to diagnostic imaging, and now the support has been merged. It is managed by central IT, but we have dedicated people whom we consider IT colleagues. So we are totally integrated into the IT network. We have the same standards as the rest of the institution when it comes to desktop applications, the email server and the security that’s necessary to do authentication. But we also have special needs when it comes to higher-performance workstations and imaging platforms that are specific to the radiologist. And I think that, to Dr. Brink’s point, you have to know where you can commoditize, but you also have to determine where you really have to have specialized support to meet the needs of a diagnostic imaging department.

ALKASAB: I agree, and finding where it makes the most sense to draw those lines is an ongoing process. Some of those lines are very clear before you begin the process, but some of them only become clear as you get into the process. That can create challenges. As one goes about centralizing IT—which can be a transition to this hybrid model, which I think is a valuable way to be doing things—there needs to be some flexibility in terms of where those boundaries lie, where it makes sense for those boundaries to live.

McENERY: Agreed. And I would say that where those lines are today is not necessarily where they were five years ago. So it has to be an open dialogue between the groups. But we have found that the more dialogue you have, the more IT understands diagnostic imaging—and diagnostic imaging understands IT—you come to a place where the lines are well defined and beneficial to both parties. One of the biggest issues we are facing is just the distributed nature of radiology. There was a time we had film, and there was a time when the reading rooms had to be close to where the imaging was performed because of the logistics of moving film. But that’s not the case anymore. And that has led to distributed situations, which absolutely do require robust infrastructure.

ALKASAB: And maybe that is an answer to the question we started with. The fact that technology has enabled us to work in a distributed way has basically required us to work in a distributed way. This creates a dependence on the technology to really deliver on that.

RBJ: How much energy should radiologists expend to stay acceptably fluent in IT-speak in order to communicate with IT people? For example, should a radiologist be able to discuss things such as standards at a moderately technical level?

LANGLOTZ: What the average radiologist wants is not to have to think too much about these standards, which really serve as the plumbing underneath the systems that they use. Today, if I want to emphasize something in a report, my only tool is using all capital letters. And that limitation relates to how information is transmitted from my dictation system back into the electronic medical record (EMR).

What we need are standards that allow a richer exchange of information and that give radiologists many more options—multimedia, better formatting and structured data that can be reused when it reaches the EMR. Those are the features people care about, I think, rather than the plumbing.

ALKASAB: I agree with Curt. I think working radiologists don’t necessarily need to understand the underpinnings of these IT systems. But they do need to buy into the idea that they are creating a report that has more than just the text they are sending; they are generating data that are being used to drive decisions downstream from where they are working. That is something we radiologists need to own and be happy about.

I think this can gradually change how radiologists perceive what they are doing. This is one of the areas a lot of us are hoping to see affected by the renaissance in artificial intelligence. In that vision, the part of the radiology report that is just about documenting what the radiologist sees, the findings section of a radiology report, will become more and more something that is automatically generated for the radiologist by these artificially intelligent assistants.

RBJ: Artificial intelligence (AI) in radiology—where is radiology headed?

McENERY: The AI discussion speaks to the fact that the product the radiologists are delivering today is not necessarily the product that is necessary to implement the future vision of what radiology and imaging can be. It speaks to the value of radiologists. More and more, you are going to see systems that define the radiologist’s value not just in the words placed into reports but also in the downstream benefits of the data created for the management of patients. The current product, a prose-driven report, is probably not going to meet the needs of clinicians in the future. The efforts from ACR and RSNA that we’ve been hearing about are going to be essential to make that transition happen.

BRINK: I would also say that, as AI plays more and more of a role in what we do, we will be generating 3D databases of findings to accompany the 3D databases of images. The radiologists will need to formally do what they do informally on a regular basis, which is to filter the wheat from the chaff. What is salient to the clinical question at hand and what isn’t? This will probably need to be more formalized to ensure that we don’t muddy the waters of what is clinically relevant with a lot of information that is not relevant to the clinical question at hand.

LANGLOTZ: Much has been said about how AI systems may replace radiologists. I think that is complete baloney. But there is no question that these technologies are going to change the way radiology is practiced as much as, or maybe even more than, the advent of CT and MR decades ago.

BRINK: Yes, and one of the principles that we have been trying to promote through the ACR is that man plus machine can potentially do great things. It’s not a matter of man versus machine. It’s more man plus machine, augmenting what man can do alone.

LANGLOTZ: If you look back to the 1980s, systems have performed at the level of expert physicians in the differential diagnosis of internal medicine conditions, for example. But the man/machine system always performs better than either one alone. That’s why I like to say the question is not whether AI will replace radiologists—it’s whether radiologists who use AI will replace radiologists who don’t.

McENERY: Curt, I think you are right on. I think the focus right now is on reading images. But there is equally as much benefit, if not more, in logistics processes around what we do. Getting back to what the report is and what information we are providing to the healthcare system, these systems could be an assistant to the radiologist. I mean, I suspect the radiologist doesn’t comment inappropriately on a finding because he or she intended to. But AI systems can help make sure the information the radiologist is providing is pertinent and complete to the clinical problem and clinical presentation, providing what the clinician needs from the radiologist.

RBJ: Overall, how sanguine—or wary—do you feel about the continuing evolution of the radiology-IT relationship heading into the next three to five years?

ALKASAB: I am optimistic about where we are going. I think radiology itself is in a position now where what we are focusing more on a forward-looking vision than on looking backward. I think we have let some of our older technologies merge with the broader needs of the enterprise healthcare system. For example, the functionalities of the RIS have now been folded into the EHR. But we are looking at how we can shape what we are providing—the nature of the radiologist’s actual work product and the practice of radiology—to incorporate new tools and a more data- and evidence-driven mode of reporting into our work. And we are trying to help work with the vendors to create the solutions that are going to enable us to fulfill the promise of that vision.

BRINK: Radiology has long been regarded as a technology innovator, developing creative solutions that really show our ability to be system integrators and provide system solutions. Because of that, I think the future is bright. We bring to the table a rich history and tradition of innovating in the space. As long as we remain open-minded about solutions that serve the enterprise as well as our department, I think we will continue to play an important and forward-thinking role in the radiology-IT relationship going forward.

LANGLOTZ: Radiologists have been pioneers in information technology. And the technology has, for the most part, played a supportive role in our practices. But given the advent of artificial intelligence and many of the innovations being developed now, technology is likely to have a much more profound effect on the practice of radiology in the next few years. We are going to need those close partnerships between radiologists and their IT colleagues. I think you will see that happening more and more as these new technologies become available

McENERY: I too am excited about the future. To quote William Gibson: “The future is already here. It’s just not evenly distributed.” An important thing is that getting there is going to take collaboration—not only within each institution but also between and among institutions and vendors. The future is bright, and it is because radiology continues to innovate, as Jim said. And we also need to innovate in the context of a profession that is becoming more and more integrated into the fabric of what it takes to take care of patients. Radiology will continue to innovate as it demonstrates the contribution the specialty makes, leading to an even larger impact into the future.


Part 19: “Return to Chicago.” Online at