基于混合模型的医学图像聚类的分类数的估计.doc
约31页DOC格式手机打开展开
基于混合模型的医学图像聚类的分类数的估计,1.56万字自己原创的毕业论文，已经通过校内系统检测，重复率低，仅在本站独家出售，大家放心下载使用摘要 聚类分析研究已有篅@さ睦罚┠昀矗渲匾约捌溲芯糠较虻慕徊嫘缘玫搅巳嗣堑目隙ā＞劾嗍鞘萃诰虻闹匾芯磕谌葜唬谑侗鹗莸哪谠诮峁狗矫嬗凶偶渲匾淖饔谩�...
内容介绍
此文档由会员 jiji888 发布
基于混合模型的医学图像聚类的分类数的估计
1.56万字
自己原创的毕业论文，已经通过校内系统检测，重复率低，仅在本站独家出售，大家放心下载使用
摘要 聚类分析研究已有篅@さ睦罚┠昀矗渲匾约捌溲芯糠较虻慕徊嫘缘玫搅巳嗣堑目隙ā＞劾嗍鞘萃诰虻闹匾芯磕谌葜唬谑侗鹗莸哪谠诮峁狗矫嬗凶偶渲匾淖饔谩�
医学图像聚类技术是医学图像处理与可视化的一项关键技术，而基于混合模型对医学图像进行正确聚类分析，则可以辅助医生更深入了解病患组织，从而为更好的进行临床诊断与手术规划提供了良好的理论支持。医学图像分析是医疗诊断、药物反应监控和疾病管理等最重要的辅助手段，具有速度快、非入侵、几乎没有副作用、花费低、效果好等优势。由于人体解剖结构的复杂性、软组织的不规则性，以及成像质量受到多种因素的影响，使得医学图像分析和理解成为一个难点，医学图像聚类分析是医学图像分析和理解中的重要技术。
目前，医学图像聚类算法还没有达到理想的识别效果，不能完全满足医学图像分析和理解的要求。本文试图研究医学图像分类数的估计及基于高斯混合模型的聚类方法。首先是基于In-Group Proportion(IGP）指标估计出医学图像的分类数，然后通过高斯混合模型和期望最大值（Expectation Maximization，EM）算法实现医学图像的聚类，通过得到的不同聚类图来验证IGP指标的准确性与有效性。
本方法适合用于医学图像识别以及混合模型聚类方法及其算法，通过对医学图像的指标分析，找到最合适的医学图像分类数，使得经过聚类分类处理过的图像对实际的医学分析与病理诊断产生最大最广泛的实际效用。
关键词：医学图像 高斯混合模型 聚类分析 分类数
Estimation the Number of Medical Image Classification Based on the Mixture Model Clustering
Abstract Clustering analysis has a long history, In recent years, its importance and interdisciplinary research direction to get people's recognition. Clustering is one of the important research data mining, has an extremely important role in identifying the internal structure of the data.
Medical image clustering technology is medical image processing and visualization of a key technology, and the hybrid model based on the medical image clustering correctly estimate the number of categories, you can help doctors better understanding of patient organizations, so as to better clinical diagnosis and surgical planning provides a good theoretical support. Medical image analysis is a medical diagnosis, drug reactions most important adjunct to monitoring and disease management, etc., with fast, non-invasive, virtually no side effects, low cost, good effect and other advantages. Because of the complexity, soft tissue irregularities, human anatomy and image quality is affected by many factors, making medical image analysis and understanding become a difficult, cluster analysis of medical image analysis and understanding of medical images is an important technology.
Medical image clustering algorithm has not yet reached the desired recognition results, can not fully meet the medical image analysis and understanding of the requirements. This paper attempts to study estimated the number of medical image classification and clustering methods based on Gaussian mixture model-based (In-Group Proportion, IGP) indicators to estimate the number of categories of medical images, and the maximum value (Expectation Maximization Gaussian mixture model by expectation, number of image classification EM) algorithm into the medical image has been obtained in the corresponding cluster diagram and through experiments verify the accuracy and effectiveness of IGP indicators.
This method is suitable for medical image recognition and clustering method hybrid model and its algorithm, by indicators of medical image analysis, to find the most suitable number of categories of medical images such processed through clustering classification of medical image analysis and the actual Pathological diagnosis produces the largest and most extensive practical utility.
Key words: Medical image, Gaussian mixture model, Cluster analysis, Classification Number
目 录
第一章 绪论 1
1.1 课题研究背景 1
1.2 国内外发展现状 2
1.3 论文的内容安排 3
第二章 聚类分析 4
2.1聚类的基本概念 4
2.1.1 聚类的定义 4
2.1.2 聚类分析的特征 5
2.1.3 聚类分析的应用 6
2.1.4 聚类算法的要求 7
2.2 聚类的方法 8
2.2.1 划分方法 9
2.2.2 层次方法 9
2.2.3 基于密度的方法 9
2.2.4 基于网格的方法 10
2.2.5 基于模型的方法 10
2.3聚类的准则函数 11
2.3.1 类内距离准则 11
2.3.2 类间距离准则 11
2.4 本章小结 12
第三章 医学图像分类数的估计和聚类算法 13
3.1 高斯混合模型和EM算法 13
3.1.1.高斯混合模型 13
3.1.2 EM算法 13
3.2 模型的选择 15
3.3 基于IGP指标的医学图像聚类研究 16
3.3.1 IGP指标相关知识 16
3.3.2 基于IGP指标的分类数的估计 17
3.3.3 基于IGP指标的医学图像聚类 20
3.4 本章小结 24
第四章 结论 25
4.1 工作总结 25
4.2 后期工作的展望 25
致 谢 26
参考文献 27
1.56万字
自己原创的毕业论文，已经通过校内系统检测，重复率低，仅在本站独家出售，大家放心下载使用
摘要 聚类分析研究已有篅@さ睦罚┠昀矗渲匾约捌溲芯糠较虻慕徊嫘缘玫搅巳嗣堑目隙ā＞劾嗍鞘萃诰虻闹匾芯磕谌葜唬谑侗鹗莸哪谠诮峁狗矫嬗凶偶渲匾淖饔谩�
医学图像聚类技术是医学图像处理与可视化的一项关键技术，而基于混合模型对医学图像进行正确聚类分析，则可以辅助医生更深入了解病患组织，从而为更好的进行临床诊断与手术规划提供了良好的理论支持。医学图像分析是医疗诊断、药物反应监控和疾病管理等最重要的辅助手段，具有速度快、非入侵、几乎没有副作用、花费低、效果好等优势。由于人体解剖结构的复杂性、软组织的不规则性，以及成像质量受到多种因素的影响，使得医学图像分析和理解成为一个难点，医学图像聚类分析是医学图像分析和理解中的重要技术。
目前，医学图像聚类算法还没有达到理想的识别效果，不能完全满足医学图像分析和理解的要求。本文试图研究医学图像分类数的估计及基于高斯混合模型的聚类方法。首先是基于In-Group Proportion(IGP）指标估计出医学图像的分类数，然后通过高斯混合模型和期望最大值（Expectation Maximization，EM）算法实现医学图像的聚类，通过得到的不同聚类图来验证IGP指标的准确性与有效性。
本方法适合用于医学图像识别以及混合模型聚类方法及其算法，通过对医学图像的指标分析，找到最合适的医学图像分类数，使得经过聚类分类处理过的图像对实际的医学分析与病理诊断产生最大最广泛的实际效用。
关键词：医学图像 高斯混合模型 聚类分析 分类数
Estimation the Number of Medical Image Classification Based on the Mixture Model Clustering
Abstract Clustering analysis has a long history, In recent years, its importance and interdisciplinary research direction to get people's recognition. Clustering is one of the important research data mining, has an extremely important role in identifying the internal structure of the data.
Medical image clustering technology is medical image processing and visualization of a key technology, and the hybrid model based on the medical image clustering correctly estimate the number of categories, you can help doctors better understanding of patient organizations, so as to better clinical diagnosis and surgical planning provides a good theoretical support. Medical image analysis is a medical diagnosis, drug reactions most important adjunct to monitoring and disease management, etc., with fast, non-invasive, virtually no side effects, low cost, good effect and other advantages. Because of the complexity, soft tissue irregularities, human anatomy and image quality is affected by many factors, making medical image analysis and understanding become a difficult, cluster analysis of medical image analysis and understanding of medical images is an important technology.
Medical image clustering algorithm has not yet reached the desired recognition results, can not fully meet the medical image analysis and understanding of the requirements. This paper attempts to study estimated the number of medical image classification and clustering methods based on Gaussian mixture model-based (In-Group Proportion, IGP) indicators to estimate the number of categories of medical images, and the maximum value (Expectation Maximization Gaussian mixture model by expectation, number of image classification EM) algorithm into the medical image has been obtained in the corresponding cluster diagram and through experiments verify the accuracy and effectiveness of IGP indicators.
This method is suitable for medical image recognition and clustering method hybrid model and its algorithm, by indicators of medical image analysis, to find the most suitable number of categories of medical images such processed through clustering classification of medical image analysis and the actual Pathological diagnosis produces the largest and most extensive practical utility.
Key words: Medical image, Gaussian mixture model, Cluster analysis, Classification Number
目 录
第一章 绪论 1
1.1 课题研究背景 1
1.2 国内外发展现状 2
1.3 论文的内容安排 3
第二章 聚类分析 4
2.1聚类的基本概念 4
2.1.1 聚类的定义 4
2.1.2 聚类分析的特征 5
2.1.3 聚类分析的应用 6
2.1.4 聚类算法的要求 7
2.2 聚类的方法 8
2.2.1 划分方法 9
2.2.2 层次方法 9
2.2.3 基于密度的方法 9
2.2.4 基于网格的方法 10
2.2.5 基于模型的方法 10
2.3聚类的准则函数 11
2.3.1 类内距离准则 11
2.3.2 类间距离准则 11
2.4 本章小结 12
第三章 医学图像分类数的估计和聚类算法 13
3.1 高斯混合模型和EM算法 13
3.1.1.高斯混合模型 13
3.1.2 EM算法 13
3.2 模型的选择 15
3.3 基于IGP指标的医学图像聚类研究 16
3.3.1 IGP指标相关知识 16
3.3.2 基于IGP指标的分类数的估计 17
3.3.3 基于IGP指标的医学图像聚类 20
3.4 本章小结 24
第四章 结论 25
4.1 工作总结 25
4.2 后期工作的展望 25
致 谢 26
参考文献 27
