A preprocessing technique can increase the compressibility achievable from lossless compression of thin-section chest CT images, according to research published online in Radiology. The technique could help sites handle the large data volumes produced by MDCT scans.
Using a method that maximizes data redundancy outside of the region of interest, a research team led by Kil Joong Kim of Seoul National University found it could significantly improve lossless compression ratios on low-dose and standard-dose CT images, as well as with both 2D and 3D lossless JPEG 2000 compression.
"Our proposed preprocessing technique may allow us savings (without concern about the degradation of diagnostic information) in terms of the system resources required for the storage and transmission of thin-section chest CT images," the authors wrote.
Even with improvements in PACS software performance, image compression is still necessary to handle the data explosion triggered by the introduction of MDCT systems, co-author Hackjoon Shim, PhD, said. This is particularly true for lung CT studies, which include large numbers of slices and are frequently performed for screening purposes.
"Since compression in nature takes advantage of redundancy in the data, it occurred to us that compressibility could be improved if redundancy outside of the thoracic body is increased, assuming that information necessary for diagnosis is limited in the thoracic body," Shim told AuntMinnie.com.
To maximize data redundancy, the researchers' preprocessing technique automatically segments pixels outside the region of interest and replaces their values with a constant value. This value corresponded to the median CT number of all of the pixels outside the region of interest throughout the study, according to the authors.
The team tested its method on 50 standard-dose CT and 50 low-dose CT studies, which were performed on either a Brilliance 16-channel or 64-channel scanner (Philips Healthcare). The 100 studies received lossless compression both before and after receiving the preprocessing technique using Accusoft Pegasus JPEG 2000 2D and 3D algorithms (Radiology, February 15, 2011).
The technique was implemented using Visual C++ (Microsoft) and a Windows XP-based computer platform with a 2-GHz dual-core processor and 3 GB of main memory.
The researchers then compared the compression ratios for both sets of images by dividing the original data size by the compressed data size.
Mean compression ratios
|
The increases in compression ratios were all statistically significant (p < 0.001).
"The increase in reversible [compression ratio] with the preprocessing technique was evident across the two compression algorithms and the two scanning protocols," the authors wrote. "Nevertheless, as expected, the degree of increase in [compression ratio] varied with the compression algorithms and the scanning protocols, both of which are known to affect the compressibility of a CT image."
In other findings, two radiologists also determined that the automatic segmentation of regions of interest was correct in all 100 cases. And the preprocessing technique took a mean of 4.1 minutes for the standard-dose CT study and 3.8 minutes for the low-dose exam.
Following the technique's success in performing 3D segmentation of the thoracic body, the researchers are planning to expand its application to other body parts such as the abdomen and lower limbs, Shim told AuntMinnie.com. The study team has also made its source code available for public use by other researchers.
