【作者】 吴燕；

【导师】 王成儒；

【作者基本信息】 燕山大学 ， 电路与系统， 2006， 硕士

【摘要】 研制和开发计算机辅助诊断系统是医学图像处理的一个热点。为了解决目前微钙化点检测中普遍存在的假阳性高和效率低的问题,本文提出将改进的SVM算法应用到微钙化点检测问题中,帮助医疗工作者提高诊断的准确率,是非常有意义的。所作的工作主要包括以下几点: (1)针对经典支持向量机算法存在的问题,为了解决训练速度和训练样本集规模之间的矛盾以及分类速度和支持向量个数之间的矛盾,构造可拒绝的支持向量机,提出了柔性支持向量机FSVM算法。 (2)针对目前的增强算法存在的增强钙化点的同时也增强了正常的背景组织的问题,为便于医生对病灶或感兴趣区域正确地诊断,提出了基于目标特征的非线性灰度重分布图像增强算法。 (3)为了提高微钙化点检测的效率,提出了基于D-S证据理论和FSVM的感兴趣区域检测,通过逐步减少待处理目标,提高检测效率。 (4)为了提高微钙化点检测的真阳性,降低假阳性,提出了基于迭代顺序滤波子空间约束和FSVM的微钙化点检测方法。利用迭代顺序滤波法作为粗检测器,将粗检测误检的非钙化点样本用来训练FSVM,获得具有更高鉴别能力的分类器,有利于提高检测精度,同时可排除大量非钙化点样本,有利于提高检测效率。更多还原

【Abstract】 Research and development of computer-assisted diagnosis system is a hot topic in the area of medical image processing. On the purpose of solving the problem of high false positive rate and low efficiency in microcalcification detection, an improved support vector machine called flexible SVM proposed here is introduced to help the diagnosis, which is extremely meaningful. The job has been done are mainly presented as follows:(1)Aiming at shortcomings of classical SVM, in order to construct a new algorithm with rejection ability and solve the contradiction between training speed and scale of training set as well as contradiction between classification speed and numbers of support vectors, then a new flexible support vector machine is proposed which will be used as classifier for microcalcifications detection.(2)For many reasons, it is necessary to enhance the medical image and increase the contrast between focuses and background. However, general method not only enhances focuses but also normal tissue. Nonlinear gray re-distribution method for image enhancement based on feature of objects is proposed in this paper.(3)In order to improve efficiency of microcalcification detection, Method for region of interest detection based on D-S evidence theory and flexible support vector machine is proposed. A classifier based on D-S evidence theory information fusion is used as a rough detector, which can eliminate many regions that are not of interests, which made the detection algorithm more efficient.(4)To improve the accuracy and efficiency of microcalcification detection, a method based on iterative rank-order filters subspace restricted and FSVM for microcalcification is proposed. Iterative rank-order filter is used as roughdetector to eliminate large numbers of non-microcalcifications and provide more representative negative samples for training of FS VM. Then FS VM can possess better discrimination ability.更多还原