• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

J4 ›› 2011, Vol. 33 ›› Issue (7): 62-66.

• 论文 • 上一篇    下一篇

基于DICOM标准的医学图像有损压缩的研究

祖研1,帅仁俊1,陈平2   

  1. (1.南京工业大学计算机系,江苏 南京 210009;2.南京市卫生信息中心,江苏 南京 210003)
  • 收稿日期:2010-09-23 修回日期:2010-12-29 出版日期:2011-07-21 发布日期:2011-07-25
  • 作者简介:祖研(1986),男,江苏南京人,硕士生,研究方向为图像处理、数据分析和医疗信息化。帅仁俊(1962),男,江苏南京人,硕士,副教授,研究方向为数据挖掘、数据分析、图像处理和医疗信息化。陈平(1962),男,江苏南京人,硕士,高级工程师,研究方向为图像处理和卫生信息架构。
  • 基金资助:

    十一五国家科技支撑计划项目(2010BAI88B00)

ZU Yan1,SHUAI Renjun1,CHEN Ping2   

  1. (1.Department of Computer Science,Nanjing University of Technology,Nanjing 210009;
    2.Nanjing Health Information Center,Nanjing 210003,China)
  • Received:2010-09-23 Revised:2010-12-29 Online:2011-07-21 Published:2011-07-25

摘要:

本文采用DICOM标准所规定的两种有损压缩方法对单帧胸部CT图像进行有损压缩。实验形式分为两种:(1)采用基于离散余弦变换的JPEG算法在不同压缩比下对50幅单帧胸部CT图像进行有损压缩。三位放射科专家对压缩后的医学图像与原始图像进行对比评片。根据评片结果进行受试者工作特性ROC分析和t检验,得出在压缩比10:1以下的胸部CT图像对临床诊断无影响的结论。(2)采用基于小波变换的JPEG2000算法对单帧胸部CT图像进行有损压缩。通过客观保真度准则判定,在压缩比为20:1以下时胸部CT压缩图像依然具有诊断价值。

关键词: JPEG, JPEG2000, 离散余弦变换, 小波变换, CT图像, 有损压缩, 受试者工作特性, t检验

Abstract:

In this paper, two lossy compression methods have been used on the singleframe chest CT images, which belong to the DICOM standard. There are two experiments:1. Discrete Cosine Transformbased JPEG algorithm has been used on 50 singleframe chest CT images with different compression ratios.The compressed medical images are  compared with the original images by 3 radiologists. According to the result of the process, a Receiver Operating Characteristics analysis and a ttest are performed to ascertain the clinical performance. The conclusion is that the chest CT images with the compression ratios as high as 10:1 can be utilized without affecting the clinical diagnosis.2. Wavelet transformbased JPEG2000 algorithm has been used on the single frame chest CT images for lossy compressions.The compressed chest CT images still have the diagnostic value, when the compression ratio is less than 20:1.The result is  judged by the objective criteria fidelity.

Key words: JPEG;JPEG2000;DCT;wavelet transform;CT image;lossy compression;ROC;ttest