Hard-copy (film) versus soft-copy (CRT) reading performance between compressed and uncompressed images. SOLs in abdominal CT images

The objective of this study was to examine the feasibility of soft-copy (CRT) reading and suitable compression. Forty abdominal CT images with a space-occupying lesion (SOL) in liver and 40 normal images were selected for receiver-operating-curve (ROC) analysis. Each image was compressed by JPEG extended mode into 1/10 its original capacity, and then an expanded image was printed on film. Ten radiologists evaluated the presence of liver SOLs (primary and secondary tumors) on soft-copy (CRT) and hard-copy (film) images. Each radiologist reviewed four types of images (original and compressed hard-copy and original and compressed soft-copy images). Values of the area under the curve in the various ROC analyses were 0.858 (FILM) and 0.842 (CRT) for original images and 0.879 (FILM) and 0.846 (CRT) for compressed images. The results of ROC analysis showed better reading performance with hard-copy than soft-copy images, but the difference was not statistically significant. Compressed images showed a higher value (0.879) than original images (0.858), a difference that was statistically significant (p<0.029) by the paired t-test but not by the jackknife method. The results indicate that soft-copy reading is a clinically acceptable alternative to hard-copy reading. We have had no difficulty in reading abdominal CT images compressed to 1/10 of the original size by the JPEG method. This study was supported in part by a grant from the Japanese Ministry of Health and Welfare. (author)