Data compression is among the ‘hot topics’ in radiology. Although there are numerous solutions available, among radiologists there is some uncertainty as to their exact application and the limitations of compression.
New guidelines issued by the European Society of Radiology (ESR) summarise the most important findings on irreversible data compression and present an important decision-making aid for users as well as for the industry.
Interviewed by Visus VIEW, Professor Peter Mildenberger, radiologist at the Clinic and Polyclinic for Diagnostic and Interventional Radiology , Johannes-Gutenberg University Hospital in Mainz, and Chairman of the ICT Subcommittee at the ESR, explained the meaning of irreversible data compression and the content and implications of the guidelines
Prof. Mildenberger: Unlike the reversible option, with irreversible data compression the user accepts a certain inevitable loss of information. After compression, an image is therefore not displayed identically, pixel by pixel. What is important is the content of the information, which obviously must not deviate from the original. In other words, the loss must not have a negative impact on the safety of the diagnosis.
The ESR policy document deals exclusively with the irreversible procedure and summarises the results of a Canadian study as well as the German and English consensus conferences on this topic. All three were based on the question to what degree compression is diagnostically acceptable. So far, assessments have been purely subjective as there are currently no objective, mathematical procedures available to determine this. This has led to a situation where the German consensus conference suggested a compression rate of 1:10 for a normal chest X-ray whereas the Canadian study suggested a figure of up to 1:30!
What other rules should be observed?
The benchmarks for compression on the one hand depend on the modality and on the other on the part of the body being examined. It depends on whether low-contrast images such as a head CT or high contrast images, such as in the case of bone fractures, are being compressed. The extent of the greyscale in the image is also important: In the case of mammography images with a detailed greyscale, the potential for image compression is relatively high. According to the German consensus conference the following compression rate recommendations result from this: Mammography 1:15, head CT 1:5 and other CT scans between 1:8 and 1:10.
What do you hope to achieve from those guidelines?
Our objective is to give users more security for data compression to ease the increasing number of telemedical processes, and especially to facilitate utilisation within the PACS. We are aiming to give users decision aids for questions such as: Do I primarily want to work with original images, or only compress the long term archiving? Or: What must be observed with different body regions and examination procedures? The topic is also of interest for the industry, as it allows it to better adapt its products to market requirements. Finally, we also want to highlight issues that have so far remained unresolved, such as the effects of compression on CAD applications. We still need to find answers here.