Say that Amazon.com operated using a thick-client model (it doesn’t, but pretend it does for argument's sake). Imagine, then, the nightmare that the head of the company’s operations would face if tasked with making sure that every customer could visit, navigate, and shop on the site successfully following even a minor revision or upgrade.
Keith J. Dreyer, DO, PhD
The problem with this scenario is that changes affecting all users can only be implemented by changing the software residing at each user's computer. Amazon's users number many millions, so in order to pull off just one revision, Amazon would have to dispatch an army of technicians to the homes and offices of shoppers around the globe.
This is the same nightmare awaiting an imaging center or hospital radiology department choosing to revise or upgrade its advanced visualization capabilities in an environment where all image-manipulation calculations and processes take place at the workstation level.
Keith J. Dreyer, DO, PhD, is director for medical imaging, Partners Healthcare System, Inc, Boston, and vice chair of radiology, informatics division, Massachusetts General Hospital, Boston. "The traditional, thick-client approach to viewing both 2D and 3D images is to load the image data onto a PACS workstation that hosts software for displaying those images," he says. "If, for instance, you’re acquiring CT images, they will be transmitted down and received in the axial plane. If you need to do anything else with that image set—such as reformatting into a coronal or sagittal plane, or performing 3D analysis—all of that must transpire at the workstation. Unfortunately, this requires fairly sophisticated software, and the heavy payload of data that must be transmitted down to the workstation each time the user wants to manipulate the image can cause potentially significant network traffic issues."
Impact on Radiologists
The solution to these problems is to forego the thick-client model in favor of a thin-client model, also known as server-side processing. Malte Westerhoff, PhD, chief technical officer of Visage Imaging, Berlin, says, "With the thin-client approach, all of the image manipulations take place on a central server before being transmitted to the workstation. The key difference, from the user’s perspective, is that he or she can look at any data available on the server, virtually latency free—meaning you can start viewing images almost immediately. An initial view is transmitted; then, as you work your way through, the remaining dataset is streamed in the background to your viewing station. It doesn’t matter whether you slide through or rotate in volume rendering.”
Westerhoff continues, "Meanwhile, the key difference from the organization’s perspective, is that software updates affecting all users can be handled centrally—and automatically, because the next time a user logs in from his or her workstation or personal computer, the latest software is what will be there."
For radiologists and referring clinicians, the advantages of the thin-client model are substantial. "It provides a unified means of giving image access to all locations," Westerhoff says. "It’s not just the radiology department that needs access to 3D images nowadays. It’s also the operating room, for surgical planning, and it’s the emergency department, the ICU, and the referrers in the hospital or out around the community."
Dreyer agrees. "One of the things thin-client does for us here at Massachusetts General Hospital is make possible advanced visualization at 400 radiology-department workstations, and it extends this capability to a large number of users spread across downtown Boston and out into the suburbs," he says. "To support such broad access to 3D images, however, requires a complex infrastructure that includes an advanced visualization laboratory.”
He adds, “We perform approximately 50,000 advanced visualization exams annually; once the images are produced and validated, they’re routed to the laboratory, where the bulk of preprocessing analysis is performed. The images then are rendered and sent to our PACS. Next, we distribute snapshots of the 3D images to the referring physicians. Currently, our referring physicians cannot engage in further analysis of the snapshot, but we’re developing a thin-client strategy designed to overcome that limitation."
A Cost-efficient Solution
The thin-client model for advanced visualization also carries with it favorable implications