Using irreversible compression in digital radiology: a preliminary study of the opinions of radiologists

Computed Tomography Image Compressibility and Limitations of Compression Ratio-Based Guidelines

“…In 2006, a survey of radiologists' opinions on compression [3] revealed that lossy compression was already used in the USA for both primary readings and clinical reviews, while Canadian institutions remained much more cautious about irreversible compression. In the USA, two radiologists out of five reported using lossy compression before primary reading and all reported using lossy compression for clinical reviews.…”

“…The impacts of compressibility differences between different modalities on image fidelity are widely known [2,3,9], but as variations within the same modalities are not as widely acknowledged, guidelines are often only defined on a modality basis. However, these variations can be fairly significant.…”

“…Recommendations like those from well-established organizations are essential, but compression ratio, on which these ratios are usually based, is poorly correlated with image fidelity [3] because deterioration levels depend highly on image information [16] and noise [17]. The CAR acknowledged this by providing different guidelines for different scenarios, but it is still only a coarse approximation and image fidelity cannot be guaranteed for a given compression ratio.…”

“…Papers such as [4,11,22,23] on diagnostic image quality have used fixed values of interest for their evaluation. However, diagnosis may require unpredictable settings [3] and a narrower window can make distortions more apparent, while a wide widow can mask them. Observers should be allowed to freely modify [14] the value-of-interest setting as they would in their practice; otherwise, their observation may be skewed.…”

“…Images with seemingly similar characteristics that are compressed using identical compression parameters can result in very different reconstruction fidelity; some can preserve all their diagnostic qualities, while others may become completely unusable. Because of liability issues raised by possible diagnostic errors caused by lossy compression, radiologists generally are not inclined to use compression techniques that would produce visually lossy results [3]. Compression guidelines were introduced to enable the use of lossy compression, but variations [4,5] in image compressibility suggest that broad guidelines allow only for conservative and suboptimal compression.…”