PACS viewing software may not show all image slices

What you don't know about your PACS viewing software might hurt you. A Japanese research team found that a number of viewers it tested were prone to dropping image slices when paging at a high speed during stack mode.

At paging speeds faster than half the refresh rate of the display, some viewers that did not have countermeasures against slice dropping lost more than half of the slices in an imaging study, according to the research published online in the Journal of Digital Imaging. These drops introduce a number of risks, including the potential for missing a lesion that was clearly depicted only on the slices that were dropped, said Dr. Masahiro Yakami of Kyoto University Hospital.

"All radiologists must keep this in mind," Yakami said. "There is no explanation of this in the user manuals of PACS viewers or no reports on this as far as I know."

Missing lesions?

In addition to potentially causing radiologists to miss lesions, slice dropping damages the video texture of a dataset, Yakami said.

"For example, tubular structures such as vessels and bronchi appear scattered in paging view," he told AuntMinnie.com. "This may affect a [radiologist's] ability to detect lesions and evaluate lesion features, and [may also] increase their fatigue."

At Kyoto University Hospital, all MDCT scanners acquire thin-slice data for all studies. The thin-slice data have been saved in the institution's clinical PACS since January 1, 2011.

Thin-slice images are often viewed during diagnosis because they eliminate partial-volume effect from images, thereby reducing radiologist uncertainty, he said. However, these thin-slice images require higher frame rates to maintain image viewing speed.

Paging speed is very high when viewing thin-slice images in stack mode. For example, a slice interval of 0.5 mm and viewing speed of 3 cm per second produces a frame rate of 60 frames per second.

"Certainly it is very difficult to detect slice omission with the naked eye because adjacent slices are too similar to distinguish at such a high speed," he said. "However, you may have experienced that an overlaid shape on a slice is sometimes not found on the first scroll, but only on the second or third, even though it has sufficient contrast to be noticed during fast paging."

Over time, the hospital realized that the experience of viewing images in stack mode during diagnosis varied with the use of different PACS viewers. To investigate these differences, the team superimposed a sequential number from 1 to 250 on each successive slice of a DICOM CT data series.

The original CT exam was a follow-up contrast-enhanced study of an aortic dissection that was acquired using an Aquilion MDCT scanner (Toshiba Medical Systems), with a slice thickness and interval of 1 mm (J Digit Imaging, October 6, 2012).

The researchers tested a number of PACS viewers, including Centricity RA1000 version 3.2.2 (GE Healthcare), EV Insite version 2.10.7.103 (PSP), Xtrek View version 1.1.0.1j (J-Mac System), and Synapse version 3.2.1 (Fujifilm Medical Systems). A freeware viewer for research that was developed in-house was also evaluated.

They used the viewers with a number of displays, including a RadiForce MX300W (Eizo Nanao Technologies), a MultiSync LCD1990SX (NEC), a RadiForce GS220 (Eizo Nanao), an Acer GD245HQ (Acer), a FlexScan S1721 (Eizo Nanao), and a RadiForce RX211 (Eizo Nanao). All displays were used with a standard refresh rate setting of 60 Hz.

Each viewer was studied with a number of hardware and configuration settings. Image data were displayed in stack mode in their original size of 512 x 512 pixels, and were paged using each of the viewers in automatic and manual cine mode without skipping slices. Automatic paging was performed using multiple speed options.

A high-speed digital movie camera recorded the display at a speed of 1,000 frames per second (fps) during paging. The researchers performed the recording five times for each combination of viewer and displays. These videos were then played back and paused repeatedly to evaluate each slice.

Speed risks

In PACS viewers that did not include countermeasures against slice dropping, the main risk for drops was the rate of paging speed, Yakami said.

"[A] very high rate of slice dropping, 52.4% in maximum, was observed with the paging speed higher than the refresh rate, while no dropping was observed with the speed lower than half the refresh rate," he said.

Tearing artifacts were observed in several viewers, even when the paging speed was less than 30 fps. Another area of concern was the failure of the viewer's indicator functions to detect slice skipping even when it had been observed.

Slice dropping can be prevented, however, by adopting certain programming techniques for the viewers, Yakami said.

Some advice

Yakami advised PACS users to use an appropriate viewer with the correct settings. If that's not possible, paging speed should be reduced to less than half the display refresh rate in viewers that do not employ slice-dropping countermeasures, he said.

"Automatic cine mode is reliable for controlling the paging speed appropriately, while manual cine mode requires users to move a mouse slowly and carefully by hand," he said. "Viewing each slice multiple times may also be effective. If the rate at which slices are dropped during a single viewing is 50%, for example, then the rates at which slices are dropped during double and triple viewings is 25% and 12.5%, respectively."

The researchers have discussed their findings with several PACS and display vendors, Yakami said.

"Some of them seem to have already started working on this problem," he said. "I anticipate new versions of PACS viewers with [slice] dropping countermeasures from the vendors."

Page 1 of 775
Next Page