MPEG-4 compression an option for combat teleradiology

In combat zones, it's not always feasible to send large cross-sectional imaging studies over the low-bandwidth networks often found in remote areas. But this critical task can be performed reliably by converting DICOM files into MPEG-4 movies, according to an article published online in the Journal of Digital Imaging.

In a retrospective analysis, researchers from Walter Reed National Military Medical Center and the U.S. National Institutes of Health found that MPEG-4 compression ratios as high as 171:1 could be feasible for preliminary image interpretation purposes.

"Such technology would be useful in combat operations or austere environments where images are acquired in remote locations with limited upstream bandwidth," wrote the study team led by Dr. Gabriel Peterson of Walter Reed (J Digit Imaging, June 22, 2012).

Seeking to evaluate the utility of MPEG-4-compressed CT studies at combat hospitals when guiding major treatment regimens, the researchers converted 25 CT chest, abdominal, and pelvic exams of combat casualties from DICOM to MPEG-4 movies at compression ratios of 171:1, 86:1, and 41:1. The studies used a single window/level setting and were 5 x 5-mm or less contiguous axial reconstructions generated from scanners from a variety of vendors, according to the researchers.

Images were compressed using VirtualDub compression software. In the first step, DICOM images were converted to AVI files at a frame rate of 10 frames per second. They were then compressed using the DivX v6 MPEG-4 codec using a single-pass compression method.

A software program called Avi4Bmp (Bottomap) converted the AVI files back to bitmap (BMP) images, which were then reconverted to DICOM by another program called K-PACS. At that point, they were sent to PACS for review.

The average file size of the original DICOM datasets in the study was 73 MB, which yielded compressed file sizes of 0.4 MB (250 kbps MPEG-4 compression), 0.8 MB (500 kbps), and 1.7 MB (1,000 kbps). The average compression ratios were therefore 171:1 (range, 157:1 to 180:1) for 250 kbps, 86:1 (77:1 to 94:1) for 500 kbps, and 41:1 (33:1 to 49:1) for 1,000 kbps.

Variability in compression ratios was attributed to the single-pass compression method used in the study.

"Had additional compression passes been made, the compressed file sizes would be averaged together and would demonstrate less variability," the authors wrote. "Since efficiency was a primary factor this study aimed to address, we chose to accept this variability because of the time saved during the compression process."

Viewing the studies on Impax PACS software (Agfa HealthCare), three board-certified radiologists with more than five years of combat trauma interpretation experience reviewed each of the studies, which were presented in random order and with a minimum of one week separation between reading sessions. They were asked to note the presence or absence of predetermined emergent findings: free air in the chest or abdomen; major vascular injury; spinal cord injury or impending injury; rupture of solid organs; free fluid/blood in the chest, abdomen, or pelvis; and foreign bodies in critical locations.

Reader performance compared with noncompressed image
171:1 86:1 41:1
Combined sensitivity and
95% confidence interval
90% 94% 100%
Combined specificity and
95% confidence interval
100% 100% 96%

The results demonstrate that this compression technique could be useful for deployed military medical corps, according to the researchers.

"The authors do not advocate use of this compression system for primary review, rather as a means of obtaining timely consultation and backup transmission when existing systems are not capable," they wrote. "The proposed compression technique is intended for preliminary rather than diagnostic interpretations, to gain information quickly for damage control surgery and also [to] provide comparison images for subsequent studies. We believe having a CT exam rapidly available, albeit lower quality, is more advantageous than higher quality CT data that is not available (unfortunately, a common scenario)."

The authors said they believe the 500 kbps bit rate produced the optimal balance of high compression (approximately 86:1) and resultant image quality.

In 69 of the 75 cases, findings from the compressed images were consistent with those from the original noncompressed images.

"In other words, 92% of the time, the reviewing radiologist agreed with the presence or absence of emergent findings on all three compression ratios as compared to the original series," they wrote.

Page 1 of 775
Next Page