【作者】 吴霜；

【导师】 于戈；

【作者基本信息】 东北大学 ， 计算机系统结构， 2008， 硕士

【摘要】 随着各种影像设备在医学诊断中的广泛应用,医学图像处理技术对医学科研及临床实践的作用和影响日益增大,其结果使临床医生对人体内部病变部位的观察更直接、更清晰,确诊率也更高。因此,医学图像处理技术一直受到国内外有关专家的高度重视,基于医学影像的计算机辅助诊断也迅速发展起来。计算机辅助诊断可以提高放射科医生诊断的准确率,协助医生对医学图像进行判断和识别。在特征提取的基础上进行模式分类是基于医学影像的计算机辅助诊断的重要步骤,所以本文针对图像特征提取和分类的问题展开研究。针对处于边缘上的点能很好地显示图像特性的特点,本文提出了一种边缘点检测的方法,旨在简化图像数据,找到代表性的点集来代表图像。同时,图像的边缘检测是图像处理中的重要方向之一,故本文采用分水岭算法和模糊C均值聚类相结合的方法来找到边缘点,不仅可以消除分水岭算法带来的过分割问题,还可以解决模糊C均值算法递归调用运算速度过慢的问题。实验表明这种方法取得了很好的效果。对于特征提取问题,由于特征提取的质量是决定分类性能的关键因素,所以选择一种适当的特征提取方法至关重要。描述图像特征的方法有很多种,如颜色特征、纹理特征、形状特征等。因为边缘点具有方向性,较其他方法能够更准确地描述图像特征,所以本文采用边缘点邻域方向测度的方法对图像进行特征提取。对于图像分类问题,本文将支持向量机的机器学习方法引入其中。分析了支持向量机的理论基础和数学模型,特别是支持向量机的推广能力和核函数理论,在此基础上应用支持向量机方法、采用医学图像库中的图像提取出来的特征对样本进行分类,并讨论了核函数及各个参数的选择。大量的理论分析和实验、特别是对比实验证明,本文提出的基于边缘点特征提取的图像分类方法具有良好的分类结果。更多还原

【Abstract】 Along with the wide applications of all kinds of medical devices, medical image processing technology is playing a more and more important part in medical science research and clinical medicine. It ensures the clinician a more direct and clearer observation of patient’s internal pathological change organs, increasing the accurate diagnosis rate. Therefore, domestic and foreign experts pay great attention to the technology. Medical image based computer aided diagnosis (MIBCAD) is rapidly developed. MIBCAD could help radiologist raise the accurate diagnosis rate, assist doctors to diagnose and identify medical images. Pattern classification based on feature extraction is an important step of MIBCAD. For this reason, the feature extraction and classification of medical image are mainly studied in this thesis.Because the points on the edge can show the characteristic of an image, an edge point detection method is proposed in this thesis. It simplifies the image data so that the typical points representing the image can be found. Image edge detection is also a key technology of image processing. Watershed transform and fuzzy C-means (FCM) clustering method together are applied to detect edge points. In this way, not only is the "over-segmentation problem" of watershed transform eliminated, but also the speed of FCM clustering algorithm which is recursively called is boosted. Experimental results display the utility of proposed methods.For feature extraction, since the quality of feature extraction is a crucial element of the performance of classification, choosing a proper way to extract features of image is extremely important. There are a lot of ways to describe image features, such as color, texture, and shape. Because edge points have directionality, and can describe image features more accurately comparing with other ways, accordingly the method of calculating the neighborhood’s orientation information measure is applied for feature extraction in this thesis.For classification, supported vector machine (SVM) is imported, which is a machine learning method. Firstly, the theoretical principle and mathematical model of SVM, especially the popularizing ability and kernel function theory are analyzed. Then the feature extraction results are taken as input and SVM is applied to classify the images of the image data base and the choice of kernels and parameters are discussed. Plentiful theoretical analyses and experiments along with contrast experiments demonstrate the proposed edge points based feature extraction for medical image classification has satisfying classification results.更多还原