For four hours Thursday evening, radiology workstations from seven manufacturers were tested before about 700 physician judges in the fifth annual Workstation Face-Off at the Stanford Multidetector-Row CT Symposium in San Francisco.
Dr. Chrisopher Herzog, chief of radiology at the Red Cross Clinic in Munich, Germany, performs final adjustments before the Workstation Face-off. Herzog represented TeraRecon. (Photo by J. Brice)
For four hours Thursday evening, radiology workstations from seven manufacturers were tested before about 700 physician judges in the fifth annual Workstation Face-Off at the Stanford Multidetector-Row CT Symposium in San Francisco.
The contestants were Barco, Carestream Health, GE Healthcare, Philips Medical Systems, Siemens Medical Solutions, TeraRecon, and Vital Images. They assigned operators to demonstrate under severe time pressure that the workstations and associated image reconstruction software could handle anything moderator Dr. Geoff Rubin could throw at them.
For each of the past five years, the face-off has aimed to define the limits of workstation performance, according to Rubin, an associate professor of radiology at Stanford.
"Everything I asked was doable, but there is no question that virtually everything could not have been done in year one, and probably not in year three," he said.
The task also reflected the emerging capabilities of multislice CT as a whole. A case called for imaging of the pancreatic gland, cystic lesions, and their suspected communication with the pancreatic duct. It required curved thin-slab minimum intensity projections deemed essential in a lecture presented earlier in the day by Dr. Jorge A. Soto, vice chair of radiology at Boston University.
Dr. Elliot Fishman, a professor of radiology at Johns Hopkins University, stressed in another lecture on Thursday the growing importance of client-server networks that give hundreds of users access to images and patient data at remote locations. The contestants' ability to operate in a client-server environment was severely tested Thursday night.
Illustrating the need for smoothing software, contestants were also asked to extract seven renal arteries in an extremely noisy data set. Another task required automated curved generation through multiple vessels with single mouse clicks and the automatic labeling of vascular branches.
"What is important to me is not just to give difficult tasks for the sake of being difficult but to pose real-world tasks and both demonstrate to the audience how far the workstations have come and lead them to where they need to go," Rubin said.
The thin-client face-off exercise aimed to illustrate how 3D visualization can provide an enterprise-wide solution by extending a network to remote sites on connections no speedier than four megabits of bandwidth with a cable modem. Participants were asked to place servers in an adjacent room and a PC on a folding table in front of the stage.
A dedicated network was created for each participant's workstation. It was used to load in two data sets from the server. A 2100-image 3D cardiac data set was transmitted to the workstation for reconstruction into a volumetric whole heart. The data were transferred to the PC, where the heart was displayed both beating and rotating on the monitor.
Contestants were then asked to load a second 2500-image data set from the server and perform the following manipulations during time remaining in the six-minute exercise:
Contestants explained their performance to the audience on the fly.
"One-click probing does a nice job for interrogating these vessels," said Dr. David Derr, an assistant professor of musculoskeletal radiology at the University of Mississippi Medical Center, as he searched for coronary stenoses at a Barco workstation.
"I'm stuck," said Dr. Axel Kuettner, a radiologist in the clinical radiology department of the University of Erlangen-Nuremberg, when his Siemens workstation froze.
Many contestants had trouble operating within the constraints of limited bandwidth.
"For those of you who are not networking inclined, please realize that four megabits is really very difficult," Rubin said to the audience between performances. "I give a lot of credit to the participants who have taken it up and said, 'Okay, we'll do four megabits. We'll show it as best we can.' I think that is really terrific."
Rubin hopes lessons learned from the face-off will lead to improved workstation and PACS design. Opportunities include better workflow, user interfaces, core algorithms, networking, and data processing.
"There should be a wealth of information that the producers of these workstations that should be able to take away," he said.
For more information from the Diagnostic Imaging archives:
Session decries lack of integration for workstation features
Carestream launches enhanced PACS, workstation
Workstation selection involves more than meets the eye
Report from RSNA: Survey finds growing acceptance of 3D interpretation tools
Study Assesses Lung CT-Based AI Models for Predicting Interstitial Lung Abnormality
September 6th 2024A machine-learning-based model demonstrated an 87 percent area under the curve and a 90 percent specificity rate for predicting interstitial lung abnormality on CT scans, according to new research.