Workstation face-off gauges 3D capability

September 2, 2004

So many images, so little time.That conundrum, kindled by the large image data sets multislice CT creates, has spurred radiologists to integrate 3D techniques into their diagnostic routine. Identifying the best workstation for the job can be

So many images, so little time.
That conundrum, kindled by the large image data sets multislice CT creates, has spurred radiologists to integrate 3D techniques into their diagnostic routine. Identifying the best workstation for the job can be daunting.

In June, organizers of the sixth annual Stanford University Symposium on Multidetector-Row CT attempted to make the task easier, staging a face-off among workstations from GE, Philips, Siemens, TeraRecon, and Vital Images. During the three-hour event, physicians selected by each vendor put workstations through their paces, interpreting five cases that ranged from relatively simple to complex.

"We should do this more often," said Dr. Thorsten Fleiter, a radiologist with University Hospitals of Cleveland, who demonstrated the use of Philips' 3D workstation. "You can only see how things really work when you compare them. And we all need to see the different ways these images can be read."

The event yielded no winners or losers. The standing-room-only crowd of 600-plus attendees at the Ritz Carlton in San Francisco were left to draw their own conclusions about the performance and capabilities of each system. That result is by design, said program director Dr. Geoffrey Rubin, chief of cardiovascular imaging at Stanford.

"The face-off is a large amount of work for many people, not the least of which are the five vendors with workstations participating here," Rubin said. "The idea is to give a representation of how a true user interacts with these data sets."

Vendors received anonymized DICOM data sets of the five cases about three weeks in advance of the event. They then identified physicians and technologists familiar with their workstations. No prerendering was permitted.

Data sets were reviewed live from a stage crowded with cables and network connections for the five workstations. Interpretation of each case was subject to a time limit of five to 10 minutes, depending on the complexity of findings. Audience members followed along by viewing a large-screen projection of each workstation's monitor.

Prior to the event, the crowd received key words of advice from Dr. Elliot K. Fishman, director of diagnostic radiology and body CT at Johns Hopkins Hospital. When it comes to evaluating 3D CT workstations, he said, "look, listen, and test drive."

Generally, workstations and software that support the shift from axial to 3D interpretation should allow radiologists to view volumes in a range of formats, including volume rendering, maximum intensity projection (MIP), minimum intensity projection (MinIP), and multiplanar visualization.

MIPs provide a valuable overview, particularly in vascular cases, said Dr. Axel Kuettner, a radiologist at Eberhard-Karls University in Tuebingen, Germany, who demonstrated the Siemens 3D workstation.

Such rendering tools allow primary analysis of data volume in real-time, which is a necessity because reviewing CT slices on film or at a review station is not practical, Fishman said.

"We're no longer in the world of axial slices," he said. "But our goals as radiologists are the same: to optimize detection of disease; define extent of disease for triage, planning, and management; and improve how we treat our patients."

Three-D CT supports all those goals, he said.

"CT is far more than looking at axial images and spinning them around. CT is motion and function," Fishman said.

But many radiologists have yet to make the leap into volumetric interpretation. Lack of training is the biggest reason, he said.

User complaints abound. Three-D workstations can be clunky, with software that is difficult to use. The time to process images may be lengthy, and image quality varies among systems. Integration with workflow is often poor, he said.

"Everything turns back around to a lack of training. But when you have the right system, and the training is there, there's no doubt that workflow is faster and interpretive capability is increased," Fishman said.

Tailoring features

While key differences remain, commercially available 3D workstations appear to be moving closer together in terms of functionality. Image display quality, user-friendliness of the interfaces, and availability of automated tools are a few of the indicators by which radiologists can measure performance.

The first determination to make in selecting a 3D workstation is how it will be used, Fishman said. Will 3D be the primary mode of interpretation for most CT studies? Or will it be a supplemental tool used only occasionally or to generate images for referring physicians? The answers will help dictate the range of function and features needed.

Other questions concern who will actually perform the 3D volume renderings-radiologists or technologists. At Hopkins, radiologists perform all 3D renderings. But at Stanford, as at many major medical institutions, technologists in 3D labs process images according to preset protocols.

Because they perform multiple postprocessing tasks, technologists' workstation needs differ from those of radiologists, said Laura Logan, manager of the Stanford 3D laboratory, during a lecture that preceded the workstation face-off.

While radiologists use 3D software to interrogate data sets in real-time, technologists perform time-consuming tasks such as curved planar reformations and segmentation of large data sets.

For the technologist's purposes, a 3D workstation should be capable of postprocessing multiple modalities, such as CT, MR, ultrasound, and rotational angiography. It should produce thin- and thick-slab volumes using MIP, MinIP, averaging, and volume-rendering techniques with the ability to rotate the slab on demand, Logan said.

Other workstation features include the ability to easily edit and check accuracy of curved planar reformations, with the capability of generating curves in multiple planes from a single trace. Tissue segmentation tools should be simple and accurate, allowing addition or deletion of structures from an edited model.

Quantitative tools must offer ease of acquisition of orthonormal diameters of tortuous vessels, centerline path lengths, areas, circumference, angles, tissue volumes, ejection fractions, coronary calcium scores, and vessel flow analyses, she said.

Such functionality is wasted, however, if it is not used. At most sites, the high cost of 3D workstations relegates them to either the CT suite or the 3D imaging lab, although a few centers have invested in additional stations in the radiology reading room. Accessibility is the most important workstation feature for Dr. Todd Fibus, assistant chief of radiology at the

VA Medical Center in Atlanta. He demonstrated use of the thin-client and stand-alone TeraRecon 3D workstation during the face-off.

"The workstation software has got to be where you are reading. If you have to go into another room, or even across the room, you are not going to use it," he said. "And that's the biggest impediment to mastering software."

The thin-client solution allows Fibus to access the workstation throughout the hospital, during conference presentations, and from home via the facility's virtual private network.

DIFFERENT STROKES

Other seasoned 3D users say that speed of image loading and processing is the most critical workstation feature. There are few things more frustrating than waiting for data sets to transfer from the scanner to the workstation. The ability to load 3000 images in under 30 seconds should be an achievable goal for vendors working on improving workstations, Fishman said.

Fleiter agrees.

"Speed is very important," he said. "You want to be able to click on an image and it is there."

Ease of use also plays a big role in overall efficiency and performance. The easy access to tools on the Philips system is one reason he favors it, Fleiter said.

"You have to remember that radiology is a business. If something takes too long to do, people won't do it. There is a huge gap in practice between what people can do and what they actually do in their practices. I like to use 3D, but not everyone does," he said.

Fleiter anticipates that the practice gap will widen as new 40- and 64-slice scanners come on the market.

"Scanners are producing more slices, and vendors are trying to keep up with workstations, but they can't," he said. "So the gap is getting larger and larger. It's problematic."

Use of automated tools is a huge time-saver, pre- and postinterpretation. During the face-off, Dr. Roger Shifrin and technologist Tom Merideth, the team at the helm of GE's 3D workstation, demonstrated the power of automated protocols.

"Any of the images in these cases can be linked to a one-touch 3D protocol," said Shifrin, chief of MRI and 3D at Radiology Associates in Daytona Beach, FL. "In the future, when we want to examine this, all we have to do is double-click."

User-friendly interface, powerful processing tools, and the versatility to perform multiple tasks are the components most prized by Dr. Scott Lipson, vice chair of imaging at Long Beach Memorial Hospital in Long Beach, CA. Lipson demonstrated the Vital Images 3D workstation during the face-off.

In his practice, Lipson relies heavily on automated segmentation tools, including vessel tracking and analysis software that segments blood vessels and generates curved planar reconstructions.

"Automated 3D fly-through, lung nodule analysis, and fusion software are vital to effectively and efficiently process cases," he said.

FUTURE TWEAKS

While each of the physicians participating in the face-off clearly prefers one system over another, the event did offer a chance to see what competing workstations can do.

Fleiter liked a feature offered by the TeraRecon system that allows image reconstruction with multiple orthogonal cuts. While emphasizing his preference for the Vital Images system, Lipson noted that GE's workstation offers similarly useful automated tools for 3D processing. Fibus observed that Philips' 3D colon package is intriguing, with its unfolded view and ability to move through the colon quickly.

"All of the workstations had nice features," Fibus said. "I wish I could take pieces of each and build my own."

If the workstation face-off proved one thing, it is that 3D interpretation of large CT data sets is doable within the context of everyday practice, Rubin said.

"I hope everyone realizes that, within the context of clinical practice, you can be doing this kind of visualization," he said. "It's tremendous the number of cases analyzed during the session and a great accomplishment by these workstations and the people operating them."