While the demand for immediate transmission of digital images is high, and radiologic image compression offers a reduction in data volume, the debate continues as to whether there is degradation in diagnostic image quality. With storage prices dropping, many still question whether image compression is even necessary.
If we go back to basics: Image compression can be divided into two categories: lossless and lossy. Lossless compression uses redundancy within an image to decrease the medical image size by 2:1 to 4:1. Lossy, or irreversible, techniques can reduce images by a far greater ratio and while it does not perfectly reproduce the original image, the reproduction may be sufficient for diagnostic purposes, with no visible degradation or loss in diagnostic value.
Why is image compression still a hot topic? According to Eliot Siegel, MD, professor and vice chair of information systems, University of Maryland Department of Diagnostic Radiology, and director of imaging at VA Maryland Healthcare Systems, there is “an increasing mismatch between our clinical image production and our storage and network capacity and bandwidth.”
The critical issue is not the cost of storage of data. “It is about the time required to move data from one location to another,” he says. Most vendors today are storing the images using only lossless compression. Lossy compression is typically used for enterprise web distribution systems while lossless compression is more often used for primary review by the radiologists.
Most PACS vendors rely on ultrafast gigabit Ethernet connections to muscle images to radiology, but since that high level of bandwidth is not available everywhere in the hospital, “everyone is utilizing lossy compression either by creating a compressed copy and then later sending a lossy compressed image or by doing this on demand for studies that are being served up outside and, in some cases, inside the radiology department,” Siegel adds.
Teleradiology groups experience longer download times due to the lack of availability of bandwidth, which tends to be significantly less than is available within hospitals, he says.
As a way to combat the issue of transmission to remote locations, Augustin DeLago, MD, FACC, FSCAI, director of the cath lab at Albany Medical Center in New York, who, in collaboration with IBM, started work on a project to apply digital video compression on CT studies.
DeLago, along with a team from IBM, examined cardiac CT studies using a H264 video compression algorithm to view them as video clips because “CT cuts the body very thinly so the correlation between the successive slice is very high.” IBM sent compressed studies at different ratios and DeLago, in a blinded evaluation, determined that using a compression ratio of 15:1 still preserved image quality and for whole-body studies, 14:1 preserved diagnostic quality.
“Obviously, when we are looking at vessel size in 2.5 millimeters and down, the compression got out of whack when it got over 15 or 20, but below that, I could not tell the difference in image quality,” DeLago says.
Normally, remote facilities use a low-bandwidth link and data are acquired and stored in the central repository at the hospital. If a remote physician wants to download a study, it takes a while over low-bandwidth links. For modern CT studies, this could take 10 minutes. “From an economical perspective, that is just a waste of time. We wanted to create a way to compress data on the fly, send it in a compressed way and then decompressed at the remote workstation, in a seamless manner,” says Vadim Sheinin, manager of multimedia technologies, IBM Research, who participated in the project.
However, Siegel says that “this relatively steep level of compression is greater than the 8:1 that is usually cited in the imaging literature for CT.”
It may be acceptable for images displayed in cine or movie mode since the eye and brain integrate noise out of moving images, but image quality is likely to suffer when reviewing a single image. “Our research suggests that the proposed JPEG 2000 3D compression standard works in an analogous way to the video compression standard and makes up for the noisier images associated with very thin slices. However, 8:1 still seems to be the upper limit for compression of these noisier thin slices even using 3D compression.”
It is clear that further investigation and product development is needed so that the entire digital radiology community can reap the benefits.