【作者】 魏欢欢；

【导师】 王济浩；

【作者基本信息】 山东大学 ， 生物医学工程， 2008， 硕士

【摘要】 随着医学影像设备如超声、放射、CT等大量进入医院,数字医学影像数据的广泛应用,PACS(Picture Archiving & Communication System)系统应运而生。它是一种用于医学图像采集、存储、管理和传输的图像存档及通信系统(Picture Archiving & Communication System,PACS)。在PACS系统中运行的医学图像必须有统一的标准,美国放射学院和国家电气制造商协会共同制定了DICOM(Imaging and Communication in Medicine)标准,即医学数字成像与通信标准,包括医学的数字成像和通信传输两个方面,成为现在PACS系统中通用的图像标准。医学图像数字量大,对存储设备和传输设备的性能都提出了极大的挑战。因此,对于医学图像压缩算法的研究非常必要。本文基于医学图像的DICOM格式、提升小波变换和BP网络算法原理,结合临床实践的可操作性,对医学图像压缩算法进行了较为深入的研究和探讨。论文的主要工作如下:(1)论文在提升小波变换的基础上,研究了一种新的有效的图像压缩算法。它使用系数为有理数的9/7提升小波滤波器,同时融合了BP神经网络和人眼视觉系统。为了提高重建图像的精确性,本文算法采用改进的系数为有理数的滤波器,解决了小波变换后的系数为浮点数造成的压缩过程不可逆的问题;进一步的,在提升小波变换基础上,针对其滤波器系数进行了大量实验,最终选取到一组有理系数,在数值上接近原来的无理系数,在压缩性能上与原滤波器相差无几。(2)小波变换后根据人眼视觉系统的原理,对小波系数进行按分解级划分,根据频带的重要性选择不同BP网络进行自组织学习压缩。同时,针对BP网络中不同的训练算法进行多组仿真实验和对比,通过得到的仿真结果选取了最优的算法,为后续章节中软件系统的开发打下了基础。(3)基于Windows环境,采用Visual C++可视化编程开发了医学图像压缩软件系统。主要完成本文所提出算法和经典EZW编码压缩算法以及DICOM文件与BMP文件的转换。可以通过可视化拉条来选择图像压缩质量,在当前窗口下观察到不同PSNR值情况时的图像质量来决定最终的压缩比。(4)开发了DICOM文件操作软件。DICOM文件作为一种运行在PACS系统中的特殊的文件格式,有其独特的文件头和数据格式构成方法,在普通电脑Windows系统中不能浏览;同时,一般医学图像像素深度差别大,正常显示范围经常不能包含其全部信息,因此,本文在详细介绍DICOM文件格式的基础上,实现了从DICOM文件格式到位图文件格式的转换以及把窗口区域的图像数据线性的转换到显示器的最大显示范围内,动态调整窗宽和窗位,观察医学图像全部信息,进而使对DICOM文件的浏览不止局限于PACS系统工作站中,而可以在任何一台普通电脑上进行。更多还原

【Abstract】 As a great number of medical image equipments,such as ultrasonic instruments,radiation and CT,have entered hospitals,digital medical image data have been widely used and the PACS emerged.PACS is a kind of picture archiving and communication system(Picture Archiving & Communication System,PACS)for medical image sampling,storing,administration and transmission.So the medical images in PACS must abide by uniform standards. American radiation institute and national electrical manufactures association established the DICOM(Digital Imaging and Communication in Medicine) standard,i.e,medical digital imaging and communication standard which consists of medical digital imaging and communication transmission.And this has become the general images standard in PACS now.The quantity of medical images is so large that it is a great chalienge to the performance of the storage equipments and the transmission equipments.On the premise of not affecting the clinic diagnosis,it is quite necessary to compress the medical images.Combining with clinic operability,this paper does deep research and discussion on medical images compression algorithm on the basis of DICOM formation of medical images,lifting integer wavelet and neural network algorithm.The main work of this paper is as follows(1)This paper mainly focuses on a new kind of image compression algorithm,which takes advantage of a 9/7 lifting wavelet filter with rational coefficients and combines BP neural network and human vision system,on the basis of lifting wavelet transformation.To improve the quality of reconstructed image,the algorithm in this paper adopts improved filter with rational coefficients,and it overcomes the irreversibility of the compression process caused by float coefficients.After wavelet transformation,the wavelet coefficients are blocked by frequency band and imported to BP network to do self-organizing learning compression.Then the validity of different reconstructed images via different BP networks algorithms was discussed and the optimal algorithm was chosen by simulation.Due to the preprocessing,the wavelet coefficients of which the numerical value is in close agreement and the energy is concentrated,which would benefit network learning.At last,the reconstructed image is of good vision effect and better peak signal to noise ratio(PSNR),and the algorithm is easy to operate.(2)Using Visual C++ visual programming technology,medical images software compression system in windows operation system is developed.It mainly achieves the algorithm proposed in this paper, the classical EZW coding compression method and the transformation between DICOM and BMP files.Users can choose different image compression quality with visual scroll bar,and can decide the final compression ratio via previewing images of different PSNR value.(3)A DICOM files operating software is developed.As a DICOM file includes unique file head and data format,it can’t be browsed in general computer.Meanwhile,medical images are of great difference in pixels depth,so the normal displaying range always can only reflect parts of the information included by the images.Therefore, this paper gives a method to transform the DICOM files to bitmap files format,and linearly transform the images data in the windows area into the largest displaying area of the display,and adapt the width and position of the window,in order to observe all information of the medical image.This method makes it possible to browse the DICOM file in any common computer更多还原