【作者】 张永亮；

【导师】 张耀楠； 郑海荣；

【作者基本信息】 东北大学 ， 生物医学工程， 2015， 硕士

【摘要】 乳腺癌是女性最常见的恶性肿瘤之一,其发病率已居女性恶性肿瘤首位,并且呈逐年上升趋势。因此,对乳腺癌的诊断和治疗已经引起了医学界的广泛重视。报告显示早期乳腺癌多数是可以治愈,其五年生存率可以达到90%以上,因此早期发、早治疗就显得很必要。近年来,乳腺癌的死亡率却有明显的下降。原因是超声弹性成像技术的应用以及乳腺计算机辅助诊断系统(CAD)的介入,更加有效的发现潜在肿块,大大提高乳腺疾病的诊断精度,对乳腺癌的治愈有着重大意义。本文主要研究了超声弹性成像技术在乳腺肿瘤分类中的诊断方法。乳腺肿块是乳腺肿瘤的重要症状,因此本文提取肿块的弹性特征及纹理特征,作为判断肿块良恶性的重要依据。本文设计一套乳腺肿块特征的自动提取方法,利用支持向量机(SVM)分类器完成分类,实现乳腺肿块的诊断。研究思路是:第一,获取图像数据,并完成乳腺图像的预处理;第二,对肿块弹性信息进行图像重建;第三,运用水平集方法实现肿瘤区域自动分割;第四,实现肿块特征提取,根据乳腺肿块图像弹性方面信息,本文提取五个弹性特征来描述肿瘤的属性;根据乳腺肿块的纹理特征,提取了肿块的四个灰度共生矩阵特征;根据肿块的形状变化情况,提取了圆度特征;总共提取十个特征,作为肿块分类依据;第五,对SVM方法进行研究,包括SVM的理论基础、核函数的介绍,将SVM技术应用到乳腺肿块分类中,完成对乳腺肿块的诊断。采用上述方法对195幅乳腺图像进行检测,应用交叉验证的方法对SVM进行训练识别,使得到的SVM分类结果具有较高的可靠性,并且最终取得了有一定意义的实验结果,为进一步的研究奠定了较好的基础。更多还原

【Abstract】 Breast cancer is one of the most common malignancies in women,and its incidence rate is the first place in the women malignant tumor,and shows a rising trend.Therefore,diagnosis and prevention of breast cancer is given great attention to the medical field.The report shows that the majority of early breast cancer can be cured,its five-year survival rate can reach over 90%,so early detection and early treatment becomes very necessary.In recent years,the mortality of breast cancer is decline.The reason is that the application of ultrasonic elastography technique and computer-aided diagnosis,and identify the potential mass more effective,and greatly improve the diagnostic accuracy of breast disease,and has great significance for the cure of breast cancer.This paper studies the diagnostic method of the ultrasonic elastography technique in breast tumors classification.Breast mass is the most common symptoms of breast tumors,therefore,this article extracts the elasto graphic features and the texture features of the masses,which is an important basis for distinguishing the benign and malignant breast masses.The research methods:firstly,acquire the image data,and complete the pre-processing of the breast images;secondly,reconstruct the elasticity information of the masses;thirdly,apply the level sets method to achieve the automatic segmentation of the tumor;fourthly,achieve the feature extraction of the masses,this article extract five elastographic features to describe the properties of tumor according to the elasticity information of the breast masses;extract four GLCM features according to the texture features of the breast masses;extract the roundness feature according to the shape of the tumor;extract a total of ten features to be used as basis of mass classification;fifthly,SVM is studied,including theoretical foundation of SVM,introduction of the kernel function,and the SVM technology is applied to classification of breast masses,and to complete the diagnosis of breast masses.195 breast images are diagnosis by above method,training and identifying SVM by using cross-validation method,in order to make the results of SVM classification with high reliability,and the significant experimental results are achieved,which lay a good foundation for further studies.更多还原