【作者】 朱红娟；

【导师】 李芬华；

【作者基本信息】 河北大学 ， 通信与信息系统， 2006， 硕士

【摘要】 医学B超图像有图像模糊,肉眼分辨率低的特点。目前,临床上脂肪肝的诊断都是医师根据经验主观判断,这种定性诊断有时会造成误诊,因此计算机辅助诊断成为医学发展的一个方向。随着当代医学的进步和计算机的发展,医学图像量化分析成为计算机辅助诊断的一种方法。本文总结了脂肪肝研究上提出的多种量化方法包括边缘检测法、光点个数计算以及纹理分析方法,通过对比发现纹理分析方法是一种有效的图像量化分析方法。在针对脂肪肝图像上使用的纹理分析包括统计方法,小波变换法以及分形模型法,统计方法和小波多分辨率方法的研究取得了一定效果,但由于这两种方法只把图像分为正常肝和脂肪肝两类,因此准确率不高。在实际临床应用中,还需要知道脂肪肝的轻重程度,为此,本文使用分形方法和小波包方法对肝脏B超图像进行研究,将图像分为正常肝脏、轻度脂肪肝、重度脂肪肝三类。最后,为了便于临床应用,用MATLAB编制了用户图形交互界面,对80幅图像进行特征量提取,对不同病理的肝脏超声图像的粗糙度、能量等特征量进行测量,发现在使用小波包方法进行特征提取时,正常肝脏、轻度脂肪肝、重度脂肪肝的各特征量显著不同,并且随着脂肪肝程度的加重,各特征量的变化呈现一定的规律性,从而可作为医生临床诊断的定量依据。更多还原

【Abstract】 The medical B ultrasonic images are illegibility and naked eye is difficult to distinguish. At present, on the clinical, doctor diagnose fatty liver, according to experiences. Sometime,this kind of qualitative diagnosis method lead to misdiagnose, the computer assistance diagnosis became the direction of medicine development.With the development of medicine and computer , the medical images’ quantification analysis became one method of the computer assistance diagnosis .Various types of quantity methods which the people had brought forward including the edge detection, the number of spot computation as well as the texture analysis method are summarized in this article. Contrast to all the methods, we discover that texture analysis is the most effective of all. The texture analysis used to liver images includes the statistical method, the wavelet analysis method as well as the fractal model. The statistical method and the wavelet multi-resolution analysis method research have obtained some effectivity, but the rate of accuracy is not high, only dividing the image into two kinds of the normal livers and the fatty livers. In clinical application, the fatty liver’s pathological degree is needed to know, therefore, in this article, fractal method and the wavelet packet analysis method are used to research for the B ultrasonic liver images, dividing the image into three kinds of the normal liver, the mild fatty liver and the specific weight fatty liver.Finally, in order to be convenient for the clinical application, graph user interface used MATLAB has been designed .We extract roughness、energy features from 80 different pathology liver ultrasonic images .It find that when the wavelet packet method is used to extract features, the features ’ value of the normal liver, the mild fatty liver and the specific weight fatty liver are different, and along with fatty liver degree aggravating, the value of features change regularly, so doctors can take these features to diagnose.更多还原