DICOM working group's effort to add to compression standard derails

August 1, 1997

DICOM working group's effort to add to compression standard derailsLack of consumer market support doomed wavelet's prospectsIn a move that may delay the addition of new lossy compression algorithms to the DICOM 3.0 PACS connectivity

DICOM working group's effort to add to compression standard derails

Lack of consumer market support doomed wavelet's prospects

In a move that may delay the addition of new lossy compression algorithms to the DICOM 3.0 PACS connectivity standard, DICOM members have voted not to proceed with the development of standards that would enable new lossy compression components from different companies to work together. The failure of the committee to develop new standards could mean that compression technologies that could save PACS users millions in archiving and image transmission costs may not achieve widespread adoption for years to come.

DICOM supports the use of the JPEG (Joint Photographic Experts Group) compression standard, containing support for 18 JPEG lossless and lossy compression methods. There is general dissatisfaction among healthcare providers with JPEG lossy compression for medical imaging applications, however, outside of uses in nondiagnostic image-review situations.

This dissatisfaction drew the attention of the DICOM committee, a group of manufacturers and healthcare providers responsible for updating the DICOM 3.0 standard. The committee assigned its Working Group IV the task of developing standards to cover new lossy compression schemes reaching the market, such as wavelet compression. Wavelet compression yields a higher compression ratio with less noise than JPEG.

The problem with wavelet compression, however, is that it is not widely used in the consumer PC market, unlike JPEG. Sales to the PC market often help vendors offset R&D costs incurred for specialized markets like healthcare. In wavelet's case, such costs could include the commissioning of clinical trials to prove that wavelet's lossy algorithms do not result in the loss of a clinically significant amount of data.

Another problem with wavelet is that there are many variations of the algorithm, which makes it difficult to conduct functions such as taking an image compressed with a wavelet algorithm developed by one vendor and saving it to an archive that uses a different version of wavelet compression.

Developing a standard

Participants in Working Group IV had hoped to solve the impasse with a wavelet standard, and developed the basis for a draft standard after a series of meetings earlier this year. The next step was to request that the DICOM committee vote on whether to proceed with the drafting of a standard that could be placed under formal change control. Formal change control is a DICOM committee procedural term meaning that the draft could not be changed without official justification and documentation. When it came time to vote via a mail-in ballot on the proposal to proceed to a draft standard under formal change control, however, the committee voted against the move, effectively shelving the working group's efforts.

Sources close to the working group say that the committee's voting basically followed manufacturer vs. user lines. With a new version of JPEG, JPEG 2000, on the horizon, several vendors have apparently opted to wait for development of the new JPEG standards. Many companies believe that a lossy compression method unique to medical imaging would be too costly without the support of users in the broader consumer PC market.

Concern over the potential legal ramifications of lossy compression has also been raised by some vendors. The American College of Radiology does not have practice guidelines for data compression, and there are not many large-scale clinical studies evaluating lossy compression. The result is a chicken and egg problem, said Dr. Steven Horii, co-director of the DICOM committee.

"Manufacturers want some reassurance that lossy compression will not get them all sued or allow further scrutiny from the FDA," he said. "Unfortunately, without a standard, there is no way to compare all of the lossy compression studies on observer (image interpretation) performance, so we can't give back to the manufacturers the scientific backing to protect them in the event of lawsuits or FDA regulations."

The DICOM committee will vote on the fate of Working Group IV at its next meeting Aug. 25. Sources close to the situation believe that the most likely scenario will be that the working group will be placed on inactive status, which allows the committee to recall it to operational status as needed. This would enable the committee to avoid forming a new working group all over again. Another possible scenario is that the working group would continue to operate, serving as a liaison with JPEG's own JPEG 2000 committee.

JPEG 2000 still years away

According to JPEG's World Wide Web site (www.jpeg.org), medical imaging is one of the application areas targeted for the new JPEG 2000 standard. Submission of compression algorithms that could be incorporated into JPEG 2000 are due by Sept. 30; the international standard is scheduled for release in November 2000.

Once JPEG 2000 is available, it will still have to be evaluated by the DICOM committee prior to inclusion in the DICOM standard, which will prolong the adoption of a lossy compression standard even further. JPEG 2000 would allow, however, for quicker validation of the compression schemes when they became available, said Dr. David Clunie, lead designer for DICOM standardization at GE Medical Systems of Milwaukee and a member of Working Group IV.

"We (Working Group IV) could (develop our own lossy compression standards), but it would then take some convincing that we had done the right thing, and that it is something that providers should adopt and devote their academic resources to doing the clinical trials," he said. "In order for lossy compression to become useable, there needs to be clinical trials to validate its use in certain clinical contexts. With an effort as large as JPEG 2000, it should be a lot easier to muster the resources to get that kind of study done."

Also, additions of new lossy compression schemes to DICOM are still approximately two to three years away, even if the DICOM committee had voted in favor of wavelet compression. As a result, the delay between when Working Group IV could add new lossy compression algorithms to DICOM and when JPEG 2000 is released was likely to be a year to 18 months, Clunie said.

"Seeing that people are not clamoring at the gates for new, standardized lossy compression schemes, that seemed like an acceptable delay in order to get a more effective outcome," Clunie said.

Informal discussions are also under way between representatives of the DICOM committee and JPEG 2000 to ensure that the committee plays an active role in the development of the standard, Clunie said.

While it is disappointing that work on adding new lossy compression schemes to DICOM will be put on hold, the long-term benefit of a standard driven by the mass market should make it worth the wait, said Dr. Alan Rowberg, an assistant professor of radiology at the University of Washington in Seattle and a member of Working Group IV.

"I don't really mind waiting two or three more years, because we've waited 15 years," Rowberg said. "The JPEG group is showing all of the signs of being cooperative and helpful and there is a huge benefit to having a broader-based standard, because we're more likely to get chips and hardware that do the right things than if DICOM does something on its own."