Computed Radiogra p hy in Skeletal Imaging: Visual Assessment of Compressed Image Quality

Abstract

Purpose : To evaluate the effect of lossy image compression on skeletal images and to determine the compression ratio which does not lead to difficulties when images are interpreted for diagnostic purposes. Materials and Methods : Thirty-two computed radiographs (CR) of osteolytic bone tumors were obtained from Picture Archiving and Communication System. They were compressed to three different levels (Q factor 30, 70, 120) using the JPEG (Joint Photographic Expert Group) technique. Ninety-six pairs of uncompressed and compressed images were randomly ordered and then serially displayed on two high-resolution monitors. During a side-by-side review, three radiologists independently compared each pair of uncompressed and compressed images, and these were rated once using a five-category ordinal scale for tumor-related findings, linear structures, and soft tissues. The reviewers were then obliged to decide which image in each pair was of better quality, and finally, they were asked to evaluate the influence of image compression on diagnostic accuracy. Results : The reviewers found no significant difference in image quality between uncompressed and compressed images with a Q factor 30. Compressed images with a Q factor of 70 or 120, however, revealed clinically relevant degradation. Among 96 observations of compressed images, 15 with a Q factor of 70 and 35 with a Q factor of 120 were considered inadequate for clinical purposes. Conclusion : If the JPEG technique is used, compressed CR skeletal images with a Q factor of 30 are acceptable for clinical application. Compressed images with a Q factor of 70 or 120 may, however, cause diagnostic difficulty and thus cannot be used for clinical purposes.