Differences in compression artifacts on thin- and thick-section lung CT images.

Abstract

The purpose of our study was to show the difference of Joint Photographic Experts Group (JPEG) 2000 compression artifacts in the lung between thin- and thick-section CT images. Thirty-five thin-section (1 mm) and 35 corresponding thick-section (5 mm) images were compressed to reversible and irreversible 4:1, 6:1, 8:1, 10:1, and 15:1. In each compressed image, pixels outside the lung were replaced with those from the original image. By comparing the compressed and original images, three radiologists independently rated the compression artifacts using grades of 0 (none, the two images were indistinguishable), 1 (image differences were barely perceptible), 2 (image differences were subtle), or 3 (image differences were significant). At each compression level, thin and thick sections were compared for peak signal-to-noise ratio (PSNR) using paired t tests and for the readers' responses using Wilcoxon's signed rank tests and exact tests for paired proportions. Thin sections had smaller PSNR (p < 0.0001). Thin sections had higher grades of artifacts than thick sections, showing significant differences at compression levels of 10:1 (mean score, 0.8 vs 0.4, 0.9 vs 0.1, 0.3 vs 0.0; p < 0.009 for the three readers) and 15:1 (1.9 vs 1.0, 1.9 vs 1.1, 1.5 vs 0.6; p < 0.0001). The percentages of distinguishable pairs (grades 1-3) were greater for thin sections than for thick sections, showing a statistically significant difference at 10:1 for two readers (31% vs 3% and 74% vs 37%; p < 0.006). The lung shows more compression artifacts on thin sections than on thick sections. Section thickness should be taken into consideration when adjusting the compression level for lung CT images.