Yamamoto S, Johkoh T, Mihara N, Umeda T, Azuma M, Nakanishi S, Narumi Y, Naito H, Nakamura H
Department of Radiology, Osaka University Hospital, Suita, Japan.
Radiat Med. 2001 Nov-Dec;19(6):321-9.
The goals of this study were (1) to evaluate the quality of compressed lung CT images obtained using high resolution CT (HRCT: 2 mm slice thickness) for degree of compression and conventional CT (10 mm slice thickness) images by using physical and subjective evaluations, and (2) to analyze the distortion of density distribution on lung CT images using histogram analysis for each compression ratio. The coding method was performed according to the Joint Photographic Experts Group (JPEG). We physically evaluated the quality of compressed lung CT images using the peak signal-to-noise ratio (PSNR) as given by the square root of the ratio of the peak value of the gray level squared to the mean square error (dB) and subjectively evaluated the CT images using the mean opinion score (MOS). The acceptable compression ratio for diagnosis was about 1:6 to 1:7 for conventional CT images and about 1:4 to 1:5 for HRCT images as determined by MOS. The PSNR corresponding to acceptable compression ratios was about 50 dB. The difference in density distribution between HRCT and conventional CT was statistically significant (Friedman test: p<0.02) in histogram analysis. Results suggested that, in comparison with conventional CT, a high compression ratio was not suitable for HRCT.
（1）通过物理和主观评估，使用高分辨率CT（HRCT：层厚2mm）和传统CT（层厚10mm）图像来评估压缩肺部CT图像在压缩程度方面的质量；（2）针对每个压缩比，使用直方图分析来分析肺部CT图像上密度分布的畸变。编码方法按照联合图像专家组（JPEG）进行。我们使用峰值信噪比（PSNR）对压缩肺部CT图像的质量进行物理评估，PSNR由灰度级峰值平方与均方误差之比的平方根给出（单位：dB），并使用平均意见得分（MOS）对CT图像进行主观评估。根据MOS确定，传统CT图像用于诊断的可接受压缩比约为1:6至1:7，HRCT图像的可接受压缩比约为1:4至1:5。与可接受压缩比对应的PSNR约为50dB。在直方图分析中，HRCT和传统CT之间的密度分布差异具有统计学意义（Friedman检验：p<0.02）。结果表明，与传统CT相比，高压缩比不适用于HRCT。
