ELIOT L. SIEGEL, MD: As we begin our closing discussion, I’d like to start by looking at the big picture. Considering the push toward RHIOS (Regional Health Information Organizations), these information sharing networks seem to be well-suited to everyone using a single vendor. Especially if that vendor has the expertise to get many disparate systems to seamlessly interoperate. In the real world, what happens in a RHIO environment when practices or facilities want to communicate but are using differing systems? Who should take responsibility for that? Should the government take responsibilities or maybe big third-parties like Microsoft or Google? Who will enable intercommunication among these health-care consortia?
BOB COOKE: The first question is to what end are you sharing this information? If it’s for patient-care purposes, it’s one dimension to the problem vs. the sharing of information that may be abstracts of a patient record or the structured data or abstracted structured data that could be used to generate more national databases associated with decision support.
Assuming it’s the former, the degree of integration that you’re going to need between these various facilities, to automatically share the data, is complex at best. It’s based upon the construction of the system, the basic technologies that are in place at each one of those systems, and then the degree to which each one of those systems conforms to various standards. So there could very well be some value for a brokering service for the sharing of that information, so it could be abstracted and absorbed in a more meaningful, normalized way.
Assuming it’s the latter, I think we would look forward to the opportunity to publish abstractions of the information contained within our systems to several national sources. And obviously that would probably be linked to an incentive-based model that would be driven from the ability to aggregate this information and use that information to generate more meaningful or more specific outcomes measurements that facilities would need to adhere to.
DR. SIEGEL: Khan, where do you see us going in the next few years? Or what do you think the possibilities might be for creating a national infrastructure? RHIOs have a varying degree of success, but they’ve certainly demonstrated that different hospital systems can indeed share medical images and other types of information.
KHAN M. SIDDIQUI, MD: No matter what infrastructure is used, if you really want to enable information sharing, it has to be a carrot-and-stick approach. There needs to be some incentive to do the sharing, and then some penalty for not sharing. That is what will drive adoption. So when the standards are agreed upon, and when an appropriate application is chosen, the incentive/penalty model becomes the means to the end.
The bigger question becomes “what is the incentive for radiologists to share information?”
Many radiologists believe that information sharing could lead to decreased volume and less revenue. Less revenue could then force hospital CFOs to reduce the number of radiology positions. A basic level of information sharing occurred when we first started using CD-ROMs but there were many issues related to importing those CDs into our imaging workflow. When should this information be presented? In many cases, if prior studies are presented to the radiologist, it’s too late. Duplicative imaging has already happened; a patient has already been exposed to potentially unnecessary radiation. Prior studies need to be presented at the time of care, when the physician is making a clinical decision. That physician needs to know the patient’s history before ordering a new study.
One potential solution is to give patients access to their data, which would give them more control in the healthcare process. The reality of the situation today is that physicians have a limited amount of time to see a patient, talk to them, take the history, examine them, look at lab results, look at the prior discharge summaries, pharmacy information, etc. When you see this abundance of information, a physician may have 5 seconds to look at the imaging-related activity. Physicians need to have instant access to imaging results. It’s not acceptable to spend 15 minutes transferring images, by that point the patient has gone home. Now the physician doesn’t have incentive because they cannot bill for looking at those. Besides establishing the technology and methods for interoperability, we need to establish the incentives that will enable data sharing. That is what the government’s role should be.
MR. COOKE: If you’re assuming that there’s going to be some kind of national infrastructure for the maintenance of patient archives, and you’re going to solve the problem in a top-down approach, as a software engineer might solve the problem, the first thing you’ve got to solve is settling on a unique identifier for each one of those elements in the record. And, as far as I’m aware, we still don’t have any form of nationalized system or anything along those lines.
It’s not an insurmountable problem, but I do think that to really manage national data sharing effectively, some form of unique identifier for the information would be a very valuable start towards creating that model.
DR. SIEGEL: When discussing advanced visualization, I would like to know more about the idea of having a single platform that would have the intelligence to be able to, based on the context, be able to call up the image viewer. Would that be a single vendor’s solution, or would that be achieved using multiple vendors? I would love the idea of being able to pick a best-of-breed solution. Do we need a system with FDA clearance to do that?
Is it possible for a vendor to deliver the ability for me to not only pick the best vendor who can provide the best portal, but also allows me the flexibility so that, on that portal, I can take advantage of the latest algorithms, or that would incorporate tools from new, up-and-coming companies? Who builds that platform if it’s multivendor? Is that the government that builds it, or is that built by one clever vendor?
VIKRAM SIMHA: So it’s not a single-vendor platform. It’s a platform to enable multiple vendors to integrate. So I think, if you have a platform, and it’s no different from what advanced visualization software today does with computer-aided detection (CAD) vendors. This istaking the concept one level higher. Where a vendor would provide a platform where, if you have a great liver segmentation algorithm, you shouldn’t have to worry about getting the DICOM data or displaying the image. The radiologist would just worry about the liver segmentation algorithm. The vendor will provide a place where you can plug in that segmentation algorithm and make it available to physicians. Of course, these are clinical tools and they need the appropriate testing and FDA clearance. But I think the vendor who provides the platform can provide a qualification step to ensure that new algorithms run on the platform, similar to the certification that Microsoft utilizes.
DR. SIEGEL: One of the things that we haven’t talked about a tremendous amount has been the use of CAD. And where do you all think we’re going with CAD? What is it going to look like, in the next few years? It seems to me as though it’s part of the continuum of image interpretation and reporting, and CAD, for the most part, has been defined by the FDA as being a second reader. But I think the idea of integrating CAD in a much more intelligent way is something that will be coming. For example, as a radiologist who sits down to read a CT of the chest, abdomen and pelvis, how will I interface with CAD systems now and in the future? Do you think we’ve reached a plateau as far as the technology goes?
RASU B. SHRESTHA, MD, MBA: As far as CAD is concerned, we all know there’s been a lot of research done. But from a clinical perspective, the actual usability of CAD has been questionable for certain areas. Obviously, there are applications in mammography and in chest,and there are some newer applications in colonography. But, in other areas, we’re still exploring what the usability might be. I think the critical theme no matter what area we evaluate is that we create a unified work space. We need an intelligent clinical applications platform because we need to stop treating images like a one-off series. We have to start looking at the patient as a whole. That involves not just anticipatory algorithms, but potentially intelligent algorithms that might be in place. We not only would have a definition of certain hanging protocols, but we could apply them to the CAD algorithms, so the algorithms would automatically be applied when we pull up that study.
DR. SIDDIQUI: Today, CAD is about pathology identification. I think the other end of the CAD spectrum, where I see there is tremendous value, is on the anatomical side. This is using CAD to identify the anatomy that exists in the current data set. How many times have you read an ER scan, and you can’t read the rib number, or someone disturbed you, and you went and recounted the ribs. So a system that discovers anatomy that’s present in the current scan, giving you things like rib numbers, vertebrae labeling, vessel labeling, macro- and micro-anatomy, could be powerful. When you couple that with some advanced technology like voice recognition, e.g. where the radiologist could say, “show me the right coronary artery, show me the left circumflex, etc.,” it could enable huge efficiencies in the reading process, leading to a faster diagnosis.
MR. COOKE: I think we’ve been constrained by our traditional definition of CAD. When considering the potential for CAD, just look at mammography. People are doing CAD because it’s reimbursed, for all intents and purposes. And even as reimbursement levels are constantly threatened there hasn’t been a significant drop in CAD usage or adoption in mammography because radiologists acknowledge it as means of providing a high standard of care—it is a tool that increases effectiveness. And if we can think of CAD in that perspective, as both driving workflow and increasing efficiency, I think that the potential is huge.
DANIEL L. RUBIN, MD, MS: First of all, the D in CAD is for detection. And that’s the focal point in detecting masses in mammography, detecting lung nodules, and polyps in CT colonography. And there will probably be a few more applications on the detection side that will grow. There are some difficult problems when you detect everything, however. I think a radiologists’ job security is safe in terms of our ability to detect other abnormalities, especially complex abnormalities in the mesentery or in the abdomen, for example. It may be a while before we have an algorithm that can detect every abnormality.
But the other aspect of the D in CAD that hasn’t been widely exploited is diagnosis. This is a paradigm where after detecting anabnormality, you will be given information about what the abnormality represents diagnostically. The massive amount of data that we’re accumulating in our enterprise systems, and the data we are already sharing will enable those kinds of applications to emerge. We’re on the verge of an explosion of applications that will leverage this information.
At a pragmatic level, given any of these new, exciting applications that are developed, they need to go through many FDA hurdles,and ultimately they may not be deployable. There are so many great applications or even simple applications, like screen capture, which we can’t integrate into commercial PACS systems. So part of our progress with CAD could be held up by the inability to integrate into a given PACS. No matter how much information is provided, reading on a stand alone workstation would break up the workflow.
DR. SIEGEL: We’re seeing private radiology groups, who compete with the hospital groups, acquiring PACS and some of the intelligent add-ons to the PACS. I think this is changing the notion that they are private reading groups who just do image interpretation by using the hospital PACS. They’re buying their own PACS, and then they’re competing. What do you see as the impact of that for PACS in the way we think of hospital-based PACS vs. what private groups are doing?
MR. COOKE: I think the private groups are really focusing on efficiency, and really focusing on how they can be the most productive. We talk about advanced reporting algorithms as a good example, and those are the groups that will invest in that technology. At the same time, they’re willing to invest in people, instead of that technology, if it’s going to make them go faster and increase their business. Now, I’m not suggesting that either approach is the right one, but I do think it presents some unique opportunities for competition, and some unique ways to incubate more efficient ways to interpret radiology exams.
DR. SIEGEL: It creates some unique challenges as a patient, because what ends up happening is: I go to my orthopedic surgeon’s office and get a conventional radiograph of my foot. He then requests that I go to get an MRI study that’s done by radiologists who have a completely separate PACS and archive system. And then I get sent back to the hospital, before my surgery, for a chest radiograph.
MR. COOKE: And your images are sprinkled across the state.
DR. SIEGEL: And there is not a mechanism, there’s nothing that ends up tying those together. There’s not even anything that necessarily ties my reports together. I’d love to have a patient-centric model. It would be nice to have a mechanism where I could have a greater level of control, whether that’s with a third party or whether that’s with a standards-based mechanism.
DR. SHRESTHA: At UPMC, we have 20 hospitals and 30 imaging centers, scattered all over western Pennsylvania, and some are located internationally. So instead of waiting on the vendors or relying onthem to develop a radiology-specific tool, we’ve created a platform where we’re federating our imaging applications, such that we’re able to pull to a common interface, the images, the reports, and the other things we might need. And we’ve made it such that it’s standards based. So if we had a second vendor or a third vendor plugged in somewhere, we could still pull those data sets. But, again, this is not at a national level. This is more at an institutional level.
DR. SIEGEL: Do you have a single viewer?
DR. SHRESTHA: Yes.
DR. SIEGEL: But do you have different PACS at those different facilities?
DR. SHRESTHA: We have 20 hospitals, and we actually have 12 different implementations of PACS across them.
DR. SIEGEL: From different vendors?
DR. SHRESTHA: Well, from one vendor. But these are implementations that don’t speak to each other. So these are nearly the same as if they were 12 separate implementations. But we have made it standards-based, and we’ve actually plugged in other vendor applications to it as well.
DR. SIEGEL: Khan, from the Microsoft perspective, what do you think about the possibility of having patients be able to control their own records? Do you think that that’s where we’re going to be heading in the future? Is it going to be that diagnostic images are written to something that a patient has control over? And, that way, one doesn’t necessarily need to have a standard patient identifier that’s national, but patients could potentially have one or more providers, and then patients would have control. And I know there’s been some concern reflected about patients having access to their images. How do you feel, from the perspective of a radiologist, and from Microsoft’s perspective, about patients being able to literally view their own images and control the images?
DR. SIDDIQUI: I’ll answer the second question first. As a radiologist, I think it’s absolutely beneficial for a patient to look at their own images. A lot of researchers have shown that when a patient looks at their own disease or their own healthcare information, they comply much better with physician instructions.
DR. SIEGEL: I’ve talked with some radiologists. And they are getting an increasing number of e-mails from patients who are reading their reports and looking at their images, and asking “What does this mean? Your name is on this report. Could you maybe spend a half-hour, or so, and write me an e-mail back and explain to me exactly what you’re talking about, and maybe I could come over and sit down, next to you, and you can go over the images with me?”
DR. SIDDIQUI: That’s where I see the opportunity, actually, is in developing applications that now actually help the patient understand what they’re looking at. So I don’t think it’s the radiologist’s job today to create reports that are patient readable or patient understandable.
DR. SIEGEL: Although mammographers are doing that to some extent.
DR. SIDDIQUI: They are doing that. But there needs to be careful implementation of how we share images with patients, I think absolutely patients would need access to their images but it becomes a data management problem. The patient has data that is then scattered all over the place.
So this is a brand-new paradigm. I mean, a couple of years back, we didn’t have the ability to share advanced imaging with physician colleagues. So physicians are still learning. There’s this whole need for the physician to be able to explain to the patient what they’re looking at. What kind of application is it? Is it an advanced application? Let’s talk about an abdominal aortic aneurysm, with all these numbers and measurements, and all the stuff. You’ll have a general practitioner completely confused if that’s the kind of visualization they see. They need a very easy way to be able to explain to the patient what is going on. What is the impact to them?
And here I see an opportunity for RSNA and ACR to actually build tools to be able to explain to the patient what they’re looking at. The American Heart Association has great tools to explain cardiac-related activities to patients. That’s where maybe ACR and RSNA have a role to provide that kind of patient-focused education. Maybe one day patients could upload their chest X-rays and get an explanation of the diagnosis. If a radiologist or a radiology group is planning to make images available to the patient, and enable communication back to the radiologists, then we need to build a structure to manage that. In the meantime, be aware that you will get calls and communication directly from the patient.
WILLIAM W. BOONN, MD: We have definitely seen some changes in the target audience for enterprise distribution of 3D tools and the way that audience is using those tools. The people who have wanted to do the most detailed manipulations actually have been surgeons, or oncologists, or people who are either doing surgical planning or therapy planning. But as this technology continues to simplify, there will be tools out there that can lower the barrier in terms of accessibility and user interface, making it easier to distribute advanced visualization tools to the general practitioner.
Not to say that, at this point, they aren’t being used. I think a lot of our general practitioners really appreciate when we create a number of 3D images, so that they can use it to communicate their findings to their patients. And even if they’re not doing the manipulation on their own, the 3D images that we create can actually be useful in their clinical practice. But, again, as things improve, as we move to simplify some of the user interfaces and offer Web-based solutions, I think that it would be great to be able to offer, to all of your referring clinicians, the ability to do 3D.
DR. SHRESTHA: I also think it’s an opportunity for some entrepreneur to build a Radiology-to-English interpretation dictionary. A system that could spell out: what the diagnosis means, what the treatment is, and how it is going to be fixed, in common English, could be a real winner.
DR. RUBIN: I view this as a new opportunity or frontier for radiologists to get their face in front of the public. I mean, there was a big push at previous RSNA conferences with regard to patient-centric radiology. And there have been a number of editorials, advocating the role of radiologists communicating directly to patients. We need to be careful of how it’s implemented, because a radiologist can only be deluged with so many patients, although internists are now e-mailing back and forth with their patients. If this is done with thought and care, it could benefit radiologists to be more directly in contact with patients.
DR. SIEGEL: It would certainly increase the visibility of radiologists. If you watch shows like House or Gray’s Anatomy, you think that it’s the surgical residents who not only scan the patients but do the interpretations. And there is no such thing as a radiologist on the majority of those shows. And having a mechanism where patients could see the name of their radiologist, look at the report, and then have access to the images, and have some added information, might increase our visibility, which I think is really important for our specialty.
Moving on, one of the things that came up during the discussion about structured reporting and speech recognition was the idea of the difference between academic settings and private-practice settings. And so when we talk about RadLex, AIM and searching images, is there a huge discrepancy between the penetration of the academic market with speech recognition and semi-structured reporting systems, and the private practice market. Do you see that changing? And are a lot of the things we talked about really just for the academic market?
Private practitioners are not enamored with doing the technology for technology’s sake or for publishing’s sake. You really have to make them better and faster, and you have to convince them to use it. Even if you do convince them, you still have to figure out a way to train them, because they work in settings without easy access to other colleagues. And so, from the private practice radiologist’s perspective, I wonder if some of our colleagues are sitting here and thinking, well, this works really well for other people, but they don’t really know what my practice is like.
DR. SHRESTHA: Coming from the academic side, I’d like to think that the private-practice folks would appreciate this sooner or later. The reality of the private-practice setting is that it’s about the bottom line. It’s about how efficiently you can get through your work day, while giving the level of care that’s required at that point. On the academic side, we have more luxury to explore some of the newer avenues that might be out there. So as we’re leveraging some of these newer technologies, as we’re looking at the workflows, including adopting some of the efficiencies that the private practice groups have mastered, I think very quickly the proof will be in the pudding. So the private-practice groups will be able to look at these efficiencies and adopt them in their practices.
MR. COOKE: There are many academic facilities that are obviously part of a larger healthcare business, and that are looking to leverage their academic strengths in the area of subspecialty teleradiology interpretation. To a large extent, that becomes a variation on the same theme that the private radiology groups are experiencing. They have a need to extend their reach, increase their volume and increase their efficiency. In some sense, that may be an opportunity for the problems to converge a little bit.
DR. SIDDIQUI: If you look at the teleradiology practices, they are private practices, but they have embraced IT to be really productive. They have very efficient, IT-based systems. So using IT in the right way is the solution.
DR. SIEGEL: That was such a nice summary of the theme of the morning, that I think I’ll use that as a wrap-up. I think that information technology really has evolved to become much more than something that drives your computer and drives your PACS. It is really an integral part of patient care now. And what we’re seeing is this is happening in the enterprise, and across enterprises. What we’re seeing are some really creative, innovative ways to be able to use information technology to completely change the way that we practice radiology and the way that we’re able to deliver care to our patients.