Innovations offer radiologists opportunity to remain relevant

Dr. Paul J. Chang sees a silver lining in medical informatics innovations that are transporting diagnostic imaging practice to places where many radiologists fear it will go.

In an interview, Chang described his vision of how PACS and workstation performance will evolve in the next three years. A professor and vice chair of radiology informatics at the University of Chicago Pritzker School of Medicine, Chang predicts that thin-client server communications models will address technical problems posed by the giant imaging files typically produced during multislice CT imaging. Even with limited bandwidth, thin-client server technology will move these files rapidly from place to place.

Thin-client servers give hundreds of physicians in remote locations rapid access to these massive files. They also give physicians access to advanced visualization software needed to manipulate images before rendering a diagnosis.

Many radiologists may see this trend as a threat, but Chang believes these and other web-based innovations actually offer radiologists an opportunity to strengthen their position in clinical practice.

A solution can be found in the type of web-based communications tools that are immensely popular with tech-savvy teenagers, he said. They can make radiologists as ubiquitous as their images are destined to be. In the process, radiologists can reestablish the kind of consultative relationships with referring physicians-albeit on a virtual basis-that typified radiology before the age of digital imaging.

Diagnostic Imaging: What will be the main challenges for workstations and PACS in the next three years?

Dr. Paul Chang: One of the big problems is that these imaging devices are generating massive amounts of data. We did not anticipate how quickly the data sets have increased with the 64-plus detectors on CT.

The problem with these very large data sets is not storage, which is cheap, but transferring them from one place to another and displaying and manipulating them. That is why at the Stanford MDCT course in June, you saw the movement toward thin-client servers where most of the heavy lifting is done server-side, so you don't pay the penalty of sending these very large data sets around through a network.

The fact that people still need advanced or dedicated 3D labs is actually a reflection of our failure to develop a fully mature model. If postprocessing tools were more intelligent, we wouldn't need specialized labs staffed with dedicated technologists to preformat images. Segmentation, for example, should be automatic, like window and level. One of the biggest challenges we face is how to make advanced visualization become mainstream.

DI: What is the server-client model? Will it eventually win out over alternative models?

Chang: The storage of thousands of slices does not bother me. What bothers me is that a thick-client model requires you to send everyone those thousands of images through a network. This is a big problem, especially if I'm trying to send that study from the reading room to my home or to referring clinicians.

Let's say I do have a lot of bandwidth and send these thousands of images. Now what? I'm also going to need a very sophisticated and powerful workstation to handle those images. But what if my referring clinician has a laptop?

Another problem is that these massive data sets are going to be larger than what is theoretically possible to address in a 32-bit processor, which means we have to upgrade to 64 bits. There is no way we can afford to have everybody's laptop or PC upgraded to 64 bits. Even if we could afford that expense, we wouldn't want to do that because most of our other applications wouldn't run on a 64-bit platform.

When you go to the thin-client model, you only have to upgrade one unit-the server-to 64 bits. Everybody else can stay at 32 bits. The heavy lifting, the addressable stuff, is done on the server. The only thing you send to the client is the image, the result of the calculation, which is relatively low resolution and doesn't represent all these thousands of images. It is just the nice pretty final image.

DI: Are you giving referring physicians access to advanced visualization software?

Chang: Exactly. This is why thin clients will predominate. By using a thin client, I get improved performance. We are doing this with our partners now. We have people at home using cable modems doing advanced visualization with limited bandwidth on their PCs or laptops. They are getting fabulous results doing virtual colonoscopy, doing stroke imaging with perfusion. But the heavy lifting isn't being done through the network or the local clients, which is how the older models work. The heavy lifting is done on the server, making it easy to work remotely.

DI: Wasn't it apparent that vendors are having problems with thin-client applications from their performances at the Stanford MDCT conference workstation face-off?

Chang: They had huge difficulties, but they'll get there. Next year, you'll see more technically mature thin-client models from all of the vendors, but that doesn't mean you are going to be able to buy it next year. Vendors need to figure out how to make it work from a business perspective. The vendors are used to selling $70,000 workstations, with service maintenance agreements and upgrades. Now, they are selling a web-based model involving a single server. They are going to need a different business model.

DI: What additional functionality will the thin-client model provide radiologists and referring physicians?

Chang: The big advantage is that advanced visualization will be ubiquitously available. The future of the server-client model will be that anybody anywhere on any type of computer will be able to do advanced visualization. That is a huge advantage.

DI: Won't this drive radiologists crazy?

Chang: That is the topic I've been asked to address at the opening session at this year's RSNA meeting. The topic will be very sobering and works off themes from the book The World is Flat: A Brief History of the Twentieth Century. The idea behind Thomas Friedman's book is globalization. His point is that if your service is a commodity and not value added, you will be outsourced by somebody who will provide that commodity service cheaper.

We are going to talk about the commoditization of radiology. In many ways, the fact that we have PACS means people don't talk to us anymore. We make it difficult for people to talk to us. Radiologists are viewed as simply the keepers of visual images and reports, and this makes us an easy target for outsourcing.

You are not going to outsource your primary-care physicians because they have to lay hands on you to treat you, but you definitely could outsource radiologists, especially if we are viewed as fat cats who sit alone in darkened rooms interpreting images on a PACS. Advanced visualization/thin clients could drive a nail in our coffin. If you give those tools to surgeons, they might say, "We don't need you damn radiologists anymore."

There is hope if we become more than the keepers of images and their interpreters. We have to add value. We have to add consultation capabilities. We need to leverage our tools to create virtual collaborations.

DI: How can this be accomplished?

Chang: We are Web 1.0 right now, but we need to become Web 2.0 to fully leverage thin clients and advanced visualization. What do I mean by Web 1.0 and Web 2.0? When people in my generation use the Internet, they use the browser and look at websites to consume information. And that is how PACS operates. We stay where we are, and information comes to us.

My kids don't use the Internet in that way. They are Web 2.0 guys. The next generation of web user-my 17-year-old son and my 13-year-old daughter-uses applications such as Skype, instant messaging, YouTube, and MySpace. These tools are not used to passively consume information. They are used to leverage the Internet and electronic communication to foster virtual collaborations.

That's our future. To realize it, we need to use and leverage our PACS tools and advanced visualization capabilities not just to send images to our clinicians, but to transport ourselves in a virtual sense as well. When a clinician is in the OR, clinic, ward, or in her office, I'm there, too.

In the old days, if clinicians wanted to examine film, they had to come to the reading room and see me because that film was physically hanging from my alternator. They talked to me, they learned from me, and they recognized that I added value.

To survive as a profession, we need to use the equivalent of Web 2.0 tools to make consultation easy again, and we have to respond more quickly with our reports to remain relevant. What is the relevance of my report in a Level 1 trauma case when my report comes in three hours after imaging, even with speech recognition? It is zero. I am not relevant if my reports are not timely. If I am not engaged, if clinicians don't talk to me, if they don't consult with me because they read the images themselves, then I have no opportunity to affect the clinical outcomes.

DI: What progress do you expect in the next three years to better integrate the communication of imaging study results into the RIS and the electronic medical record (EMR)?

Chang: You are going to see more work in the next three years to optimize communications models. It will have to go well beyond simply integrating to the EMR and RIS. When you are talking about integrating to the EMR/RIS, you are just talking about a report. Most of the time that report is useful for billing but is frequently available too late to influence clinical decision making.

We are going to have to spend much more time developing systems for dashboard alerts. In other businesses, e-mail is sometimes not good enough. People are communicating using other kinds of methodologies. They use instant messaging, wireless local area networks, and wikis as virtual repositories for exchange of information.

It is mind-boggling that it has taken us so long to understand this, because outside of medicine we live this every day. That is why the BlackBerry is so popular. That is why we have instant messaging or other synchronous and asynchronous tools. It is because different kinds of messages require different kinds of delivery mechanisms.

Yet, in radiology, we still concentrate on the phone or report-and putting the report into the RIS and EMR. This is inadequate. We have to understand what we do is not dissimilar from other businesses where rapid communication of results or information requires many types of optimized transmission mechanisms, not just one.

DI: How might the work of the Integrating the Healthcare Enterprise and other such efforts lead to automated distribution of postprocessed imaging studies?

Chang: I have a love-hate relationship with IHE. I love its integration profiles-the use cases, the specifications-and its understanding of work flow for communications and advanced visualization. IHE has also done a good job of marketing itself. Everybody knows about it.

But the fundamental issue of IHE is interoperability. In order to support complex work flow, systems that ordinarily don't talk with each other must communicate with one another to support the workflow.

I hate how IHE proposed to solve the interoperability problem. They depend on edge protocols, such as HL7 and DICOM, and that is a nonstarter from an engineering and architectural viewpoint.

But guess what? The problem is not unique to medicine. All industries face it, and without exception, they have all locked on to the same solution. It used to be called "web services." Now it is called service-oriented architecture.

We should be less arrogant, a little more humble, and adopt solutions that have proved successful elsewhere. That means addressing the interoperability problem by using distributed object approaches, middleware, web services, service-oriented architecture, and an enterprise service bus. It doesn't matter what you call it. It is the same type of approach.

Mr. Brice is Senior Editor of Diagnostic Imaging.