摘 要
乳腺癌是妇女常见高发的恶性肿瘤之一,早期的发现、诊断和治疗是其防治的关键,而微钙化点是早期乳腺癌的重要特征。
为了能够有效地检测出乳腺X线图像中的微钙化点,本文提出了一种包含三个模块的钙化点自动检测模型。首先根据钙化点在X线图像中的影像学特征,应用一种基于双正交小波和Hessian矩阵的方向差分滤波器组进行增强,从而得到相应的特征子图;将其与独立分量分析(Independent Component Analysis,ICA)的方法相结合对微钙化点簇进行描述,提取出有效的分类特征。然后,使用BP神经网络对进行分类判别,获得94%的真阳性率和14%的假阳性率。最后,利用方向差分滤波器组以及阈值自动选择算法对钙化点进行增强。
实验结果表明,这种检测方法具有良好的实用性和鲁棒性,能够有效地检测出乳腺图像中的感兴趣区域以及其中的微钙化点。
关键词:计算机辅助检测 微钙化点检测 多分辨分析 独立分量分析 人工神经网络
ABSTRACT
Breast cancer is one of the most common malignant tumor diseases among women, and the key of its prevention and curing lies in its early detection, diagnosis and treatment. Microcalcifications are the significant pathologic features of early breast cancer.
In order to detect microcalcifications in mammograms efficiently, this paper presents an automatic detection system model which consists of three modules. Firstly, based on the discrete biorthogonal wavelet and the basic concept of Hessian Matrix, a directional difference filter bank is introduced to obtain several subimages; these subimages are combined with the independent component analysis (ICA) for the descriptions of the images, so the valid features for classification can be extracted. Then, BP Neural Network is employed to discriminate experimental samples, achieving the results that the True Positive Rate (TPR) is 94% while the False Positive Rate (FPR) is 14%. Finally, using the subimages obtained from the filter bank and a method of threshold auto-selection, microcalcifications are enhanced.
The experimental results illustrate that the proposed detection modules with good practicability and robustness, can effectively detect regions of interest and microcalcifications in mammograms.
Keyword: CAD, Microcalcification Detection, Multi-resolution Analysis, ICA, Artificial Neural Network
目 录
第一章 绪论 1
1.1 课题背景及意义 1
1.2 国内外研究进展及现状 2
1.2.1 计算机辅助检测系统现状 2
1.2.2 乳腺X线图像中微钙化点自动检测的研究现状 3
1.3 论文的研究成果和章节安排 3
第二章 乳腺感兴趣区域特征提取及分类 5
2.1 基于双正交小波和HESSIAN矩阵的方向差分滤波器基本概念 5
2.1.1 基于HESSIAN矩阵的二次差分值 5
2.1.2 二次差分滤波器组的构建 6
2.2 ICA的基本原理 9
2.3 特征提取 10
2.3.1 灰度特征的提取 10
2.3.2 ICA特征的提取 12
2.4 BP神经网络的介绍 13
2.5 实验结果及分析 14
2.5.1特征提取的实验结果 14
2.5.2感兴趣区域ROI分类的实验结果 18
2.6 小结 20
第三章 钙化点增强 21
3.1基于多分辨分析和阈值自动选择的钙化点增强 21
3.2基于局部统计特征的图像增强方法 23
3.3实验结果及分析 25
3.3.1基于多分辨分析与阈值自动选择算法的增强效果 25
3.3.2基于局部统计特征图像增强方法的增强效果 27
3.4小结 27
第四章 结束语 29
4.1总结 29
4.2对未来发展的展望 29
致 谢 31
参考文献 33

您当前的位置:
