View From the Podium: Stanford MDCT Face-Off
By stipulation, there were to be no winners when eight vendors of advanced imaging workstations and their physician teams competed in the Seventh Annual Original Workstation Face-off, held May 20 in San Francisco, California, at Stanford Radiology’s 11th Annual International Symposium on Multidetector-row CT. Even so, according to two radiologists who made presentations at the competition, the stakes were high because conference attendees might later be making purchasing decisions based on workstation performance at the face-off. Both physicians were ready to go when the signal sounded to demonstrate five cases of varying anatomical and clinical complexity.
Bernhard Meyer, MD
The face-off has become a highly anticipated event. Moderator Geoffrey Rubin, MD, told participants that the face-off was originally designed to give workstation vendors a level playing field to demonstrate their wares. “In theory, it is a level playing field,” William W. Boonn, MD, says, “but there is always the question of whether some vendors do some preprocessing.” Boonn presented on behalf of Visage Imaging (Andover, Massachusetts); both he and the second Visage presenter, Bernhard Meyer, MD, say that they played strictly by the rules and postprocessed only on stage. Neither saw evidence that other vendors fudged, and say this year’s face-off went more smoothly than last year’s for all vendors, when a number of technical glitches could be observed for some of the cases.
Presenters prepare for the 2009 face-off.
“There were some differences, this year, in what tasks the vendors could do,” Boonn says. “Some did not perform the tasks within the time required, but it was hard to fault them fully. If someone makes a mistake, there isn’t time to fix it. In a real-world situation where you aren’t limited in time, that workstation would have done a fine job.” The time constraints, which were severe, and the pressure of performing in front of an audience were both reasons that a winning solution couldn’t be named, Boonn and Meyer note. The larger reason was that some workstations fit well into use niches that others might not fill. “One could not say that there was one system that performed best,” Meyer says. “Every system has its pros and cons. It was hard to compare them when you were sitting on stage waiting for your presentation. I wouldn’t say there was one that was the winner, in my eyes.” Visage Thin Client Meyer and Boonn both note that the Visage CS Thin Client/Server is different from other solutions presented at the face-off in that the Visage system relies on a server and accompanying software to perform image calculation, manipulation and processing. All viewing and postprocessing can be done on a PACS workstation, but also on an office computer like that of the typical referring physician. Whereas other vendors were making their presentations using big, high-powered workstations, the Visage server box was connected to a conventional laptop computer. The images were displayed and postprocessed on the laptop, showing how adaptable the Visage system is to a hospital or multispecialty setting where technologists, radiologists, referring clinicians, and surgeons might have differing display and reconfiguration needs.
Presenters watch as a case is demonstrated.
“Visage is server based, and I can use any PC to access the server,” Meyer says. “There’s no special hardware needed for a thin client. We had a server there on stage, and I was using a very small laptop to do all this stuff.” Boonn adds, “It was impressive that the Visage system is fully thin client. We presented on a run-of-the-mill laptop with processing done on the server, so that was a distinguishing feature. We were able to get through all the cases with time to spare.” Meyer notes, “The main advantage, with Visage, was that it was so easy to use, and there weren’t so many clicks and buttons. There were others that were quite good, but I’m not convinced I should use another system. I’m very happy with what I have.” Expertise Boonn and Meyer were asked to present for Visage because both have helped the company develop specialized postprocessing applications. Both also presented for Visage in last year’s face-off. Boonn is chief of 3D and advanced imaging and associate clinical director of imaging informatics in the radiology department at the Hospital of the University of Pennsylvania in Philadelphia. He is also an assistant professor of cardiovascular imaging at the University of Pennsylvania Medical School. Boonn says that he worked with Visage to develop a cardiovascular product, but he noted that he also worked with other vendors. Meyer, based in Berlin, Germany, is an assistant professor in interventional vascular imaging at Charité University Hospital. Meyer says that his collaboration with Visage involved postprocessing vascular-imaging datasets. “Two years ago, we also started working on bone removal tools that are now part of the commercial Visage CS product,” he notes. Meyer flew in for the face-off, and before it was over, had to catch a plane back to Berlin for a conference, so the whole experience, for him, took place on the fly. That is terminology sometimes applied to thin-client technology itself, where data-streaming techniques with the server are used to deliver elements of an imaging study only where and when they are needed. Reliance on minimal data streams is what makes a thin client fast, from an end user’s point of view—and one reason Boonn and Meyer say that they easily met the time limits imposed at the face-off. The Cases Two weeks before the event, Meyer was sent slides of the five cases to be presented at the face-off and was allowed to load and practice on the datasets (as were all participants), but no postprocessing could be retained. At the face-off, Meyer says, all competitors started with the same raw datasets and had to do the postprocessing in front of the audience within the given time periods. Of course, if they practiced, as the Visage presenters both did, they knew how to achieve the results that they sought quickly. Boonn and Meyer divided the demonstrations between them, based on expertise. On one case, both worked as a team. The first case was a skeletal-vascular work-up for surgical planning on a patient with chronic osteomyelitis who was undergoing a myocutaneous flap transfer. Meyer calls this case easy. “We do a lot of CT runoff studies,” he says. “I do them every day.” The bone removal in the case was done automatically by the Visage software. “It worked perfectly, and I didn’t have to do any postprocessing afterwards,” Meyer says. Although he was given three minutes for the demonstration, Meyer needed only two. “The rendering is really fast on the Visage,” he notes, “with a cool graphics card in real time. The physician was the limiting factor, in terms of speed.” Meyer also did the second case, a carotid/intracranial study with a two-minute time limit that involved setting minimum seed points to extract a centerline from the aortic arch to the right middle cerebral artery. Meyer calls this an easy case, too. Last year, he says, a similar case had been difficult. “This year, we needed to count the seed points,” he says. “I used only six seed points, and that worked very well. I was very familiar with the vascular cases, so it wasn’t that challenging. I was able to rotate around the centerline and demonstrate perfect alignment. We ended before the time threshold.” On the third case, Boonn and Meyer worked together to complete some complicated postprocessing and reconfiguration. Meyer pretended to be a radiologic technologist in a 3D laboratory, assigned to create a model for a genitourinary vascular study. In creating the model, he was instructed to exclude (as though inadvertently) the most inferior right renal artery originating from the distal aorta. Boonn then played the part of the radiologist receiving the study in the reading room. He had to restore the missing renal artery without adding extraneous tissue or vessels. Boonn then had to create a new volume rendering showing four right and two left renal arteries. The team used two different accounts on the laptop to demonstrate this. “Boonn restored from another Windows account and showed he was able to re-edit everything under the time limit,” Meyer says. Meyer performed the fourth case, which involved measuring volume change in a lung nodule. He says that the key was using the same segmentation threshold for both parts of the study to ensure accurate measurement. “I did this case, but I didn’t see the [other vendors’] results,” Meyer says. “I had to leave to catch my plane.” The final case, which involved renderings on a heart patient with existing bypass grafts, was done by Boonn. “The presence of metal made extraction of the coronary artery more challenging,” Boonn says. “Almost all the vendors can do a fantastic job on the easy cases. Part of this case was to show how to deal with streak artifacts and more challenging datasets.” Boonn says that he finished in the three minutes allotted. Boonn praises the face-off sponsors for selecting cases that “would help to highlight various differences between the applications and challenge the workstations to perform the various different tasks.” How did the Visage system compare? “I think it performed well,” Boonn says. “It depends on which system would be optimized for your particular needs.” Boonn says that part of the interest generated by the face-off was being able to study workflow as the presenters worked with the different systems. “The differences are more than if they just get the job done,” he notes. “The differences in workflow can be more important than whether they actually do the job.” Boonn also singles out automation as a variable among competing workstations. “Some vendors had more automation in special analysis that required fewer clicks than the system I used, but the end results were similar,” he says. “Automation often comes with a price, though. If you automate the procedure too much and extract the wrong thing, it takes longer to go back and fix what the computer did than to do it manually from the beginning. It’s a balance.” Boonn and Meyer make it clear that they are physicians willing to demonstrate the Visage system’s capabilities and advantages, but are not affiliated with the company’s sales or marketing departments. “The people from the company said I did well and they were happy,” Meyer says. “Before the face-off I told them, ‘I will do the job my way.’ I’m a radiologist, not a salesman. I just show how I prefer to work with the solution. I think I did it OK.”