In 1993, the Baltimore VA Medical Center made medical imaging history.

It became the first health care facility in the world to feature an all-hospital PACS. Now, over the next few weeks, the hospital is looking to shake things up again: it's examining vendor proposals to replace its existing system with the next generation of PACS technology.

But although PACS have changed since the hospital made the digital transition, they haven't changed enough, according to Dr. Eliot Siegel, the mastermind behind the hospital's PACS adoption and one of the early pioneers of the digital image technology.

"I think it's almost comic, but actually tragic simultaneously, how 20 years after we initially bought the system, how little progress from an intelligence perspective there has actually been," said Siegel.

DOTmed News spoke with Siegel, who's also a radiology professor at University of Maryland, and who recently gave the Dwyer lecture on the future of PACS at the annual conference of the Society of Imaging Informatics in Medicine (SIIM). We talked about what he wants to see in next-gen PACS, bringing IBM's Watson to medicine and his hopes for a Center for Computational Medicine. Here's part one of our talk; check back tomorrow for the rest.

DMBN: In a preview of your talk at SIIM, it mentions your belief that PACS are still overly beholden to film. Why is that?

SIEGEL: I think in a lot of ways the processes in hospitals, and the way PACS works, haven't really changed very much since they were paper-based. So for example, it would be great to be able to automate the process of setting up scanning protocols with CT and MRI, based on the reason for an examination. It would be great to be able to, in a more automated way, follow up on recommendations. If I have to make a recommendation that a patient should have a CAT scan, after I see what might be an abnormality on a chest X-ray, right now we're operating just as though the systems were on paper and not digital, because there's no automated way to track that follow-up.

There's also no mechanism to be able to connect my electronic medical record with my PACS, so that when I'm on my PACS and looking at different images, I have to open a whole separate window and sign into the hospital computer system to bring up patient information. The systems act as though they're completely independent entities...

The same thing happens with radiology reports in the EMR. The reports are digitized in the sense that if I want the information, I can look up an image report, but there's no hyperlinking, there's no idea of being able to index that information, there's no idea of being able to intelligently search or take advantage of the fact that I have certain types of information available to me. For instance, if a patient has a history of breast cancer in the family, there's no automated way, if I'm reading the mammogram, to have that information brought over from the EMR.

We still are living in a world where...there's no intelligence with regard to combining information and asking questions and doing things prophylactically, having the computer automatically say, "Hey, you made this recommendation but no one ever followed up on it," or, "Hey, it looks like the impression that says there's disease is different from the body of your report, which seems to imply that everything is normal," or, "Hey, you issued a report that sounds like it has some important findings, but there's no record that anyone has opened up and read your report, much less acted on it."

Where are the vendors in this?

The problem is most of the vendors only have control of their own domain. An Epic or a Cerner doesn't really interface with a GE or a Carestream or a Philips PACS, for example. The vendors sound sympathetic, but to a large extent most of the people who buy these systems are relatively complacent, because they don't even think about the idea of having intelligence in the systems. People still think when they go to buy a PACS that what they really want is something that will, as fast as possible, display the images, allow them to generate reports and go on to the next study. But because this capability for smart PACS hasn't existed, radiology facilities do not know to ask for it. If imaging facilities made "smart" a criterion for purchasing or upgrading, it would accelerate this next generation of PACS.

Is there anyone that makes something that comes close to what you want?

The most interesting developments seem to be coming from non-traditional PACS vendors. There have been exciting advances in "artificial intelligence" systems. Apple's Siri has been improving rapidly and of course the IBM Jeopardy playing software, Deep Q&A has demonstrated the tremendous potential for software that can interpret complex questions and rapidly find answers using a dynamically changing database. Microsoft's, and now Caradigm's Amalga represented a user friendly/intelligent bridge between the EMR and other clinical information systems providing the type of functionality that I would love to see as part of a "smart" next generation PACS.

So how is the replacement process going along?

Our Chesapeake regional VA network, VISN 5, is about to make a selection of a new, hopefully next generation PACS. So our challenge is in some ways analogous to what we had with our first generation PACS 19 years ago; to select a PACS that will meet our future needs. Early PACS pioneer Sam Dwyer taught us to follow the "hockey pass" principle of not passing to where the guy is but to where he is going to be. Now we've got almost 20 years of experience, the question is where will the technology be in the next 5, 10, 20 years and how can we incorporate that into our next generation PACS. My best guess is that our best bet will be to purchase a system that takes advantage of advanced clinical as well as business intelligence in a new generation of smart and interoperable systems that will make us more productive, safer, and more effective..

Check back tomorrow for part two of our talk.