Replacement of single- or dual-slice spiral CT with multidetector-row scanners undoubtedly increases the level of imaging data generated. But how exactly is this growing CT data burden affecting PACS performance? Austrian researchers compared different
Replacement of single- or dual-slice spiral CT with multidetector-row scanners undoubtedly increases the level of imaging data generated. But how exactly is this growing CT data burden affecting PACS performance? Austrian researchers compared different solutions.
Innsbruck University Hospital is relatively unusual in having three different PACS operating concurrently. The hospital's two radiology departments are each served by an off-the-shelf PACS from a well-known vendor. A third home-grown PACS, developed with a local IT company, handles long-term archiving and distribution of exam data to individual wards.
"They all work together," said Dr. Raimund Vogl, imaging information manager at Innsbruck. "Referring physicians in any ward or outpatient department have transparent access to all data produced in either of the radiology departments."
The hospital's switch to MDCT started four years ago, with the upgrade of a single-slice scanner to a quadslice CT unit. Innsbruck radiologists gained two more quadslice scanners and two 16-slice scanners between February 2000 and December 2002. The dramatic rise in the number of CT studies, images per exam, and absolute data volume prompted the IT team to investigate whether their systems could handle the load, Vogl said.
He presented the study's results at the Management in Radiology meeting, held in Basel, Switzerland this month.
The hospital's existing hardware was found to be perfectly capable of handling the increased data load, and switching to storage area networks was not necessary. But differences between the departmental PACS did emerge. The on-demand PACS, where radiologists pull imaging studies onto their selected workstation, coped with the MDCT data volume much better than the autorouting solution, which had to push data from its archive to multiple workstations, Vogl said.
The increased number of images per series posed problems for both PACS. Calculations revealed the maximum transfer rate from modality to archive to be approximately 75 images per minute, even with a 100-Mb/sec Ethernet link.
"If you have an imaging exam with 1000 or 2000 images, this calculation shows that it will take 10 to 20 minutes to send the data to the archive, and then be available to the workstation," he said. "If you have an urgent case - for example, a suspected pulmonary embolism - if it takes 15 minutes to have the images available for reporting, that's too long. So the radiologists sit directly by the console of the CT unit and do the very urgent reporting right there."
The study indicated the benefits of using lossless compression to transfer CT images. It also showed how upgrading PC workstation processor power to at least 3 GHz, to speed up image decompression, would have more impact than upgrading the workstation link.
"Changing from a 100-Mb/sec to a 1-Gb/sec Ethernet, you would expect to see performance improve by a factor of 10," Vogl said. "But in a PACS context, the actual performance only increases by 50%, owing to the time it takes for image decompression."