基于DCE-MRI影像的乳腺癌早期诊断研究
发布时间：2018-11-18 16:32
【摘要】：乳腺癌是具有高死亡率的女性最常见恶性肿瘤。由于其致病因子的不确定性和病情的隐匿性，患者经常在乳腺癌发展至中晚期才被确诊，延误了治疗的最佳时期。乳腺动态核磁共振成像术（Dynamic Contrast-enhanced magnetic resonance imaging, DCE-MRI）凭其高组织分辨率近年来已作为一种重要早期乳腺癌影像诊断手段而被广泛采用。与此同时，为了提升医生影像诊断中的准确率，计算机辅助诊断（Computer-aided Diagnosis, CAD）技术作为一种辅助手段也被提出并应用于临床诊断和治疗决策。 基于DCE-MRI的早期乳腺癌计算机辅助诊断是目前CAD领域的一个研究热点，不少学者进行了相关研究。这些已见诸报道的乳腺癌CAD系统只注重于病灶区域的研究，往往先分割出病灶区域，然后根据分割结果中特定的像素点提取影像特征，最后再利用分类器进行乳腺癌诊断预测。根据近期临床研究，乳腺组织（Fibroglandular Tissue, FGT）与罹患乳腺癌的风险之间存在一定关联；在影像层面上，乳腺背景实质增强（Background ParenchymalEnhancement, BPE）特性呈现出更大的关联度。但是这些研究都基于医生的临床研究得出，其分析方法通常都是定性的，同时具有很强的医生个人主观性，无法有效应用于临床诊断实践。针对这些问题，本硕士论文创新性地对乳腺背景实质影像特性在甄别早期乳腺癌中的应用价值进行比较系统的客观和定量性研究。 本论文的研究工作主要围绕乳腺腺体背景特征的乳腺癌甄别能力和病灶与腺体背景特征互补性这两个问题来展开。具体内容包括： (1) MRI乳腺影像预处理 对DCE-MRI原始影像进行了包括降噪、配准以及研究区域分割在内的预处理。在医师标注的病灶区域的指导下，，本文使用基于解剖学结构的乳房区域分割方法在DCE-MRI影像上获取乳房区域，然后根据医生的影像标注信息进而确定病灶和乳腺腺体背景两个待研究影像区域。 (2)病灶与乳腺背景区域特征提取 针对病灶区域与乳腺背景区域的影像特征提取问题，分别利用统计方法和三维灰度共生矩阵对每例病例的提取计算了158维影像特征，这些特征能分别表征病灶区域和乳腺腺体背景区域的动态增强特性与三维纹理特性。 (3)乳腺腺体背景特征的乳腺癌甄别能力研究 为了评估每一个计算自乳腺腺体背景特征的乳腺癌甄别能力，本研究利用受试者工作特征曲线（Receiver Operating Characteristic Curve，ROC）和K-W检验计算分析每个乳腺腺体背景影像特征区分乳腺病灶的能力。实验数据显示部分乳腺背景特征对乳腺病灶的区分能力与病灶特征的相近，对乳腺癌具有较高的甄别能力。 (4)病灶与腺体背景特征互补性研究 为了了解病灶与腺体背景特征互补性情况，本研究对病灶特征集和包含全部特征的特征集分别进行特征挑选。然后根据挑选后的特征子集所对应的CAD系统分类性能来衡量该特征子集包含的有效信息量。实验结果表明同时包含有病灶和乳腺腺体背景特征的特征集比仅包含病灶特征的特征集含有更多有利于甄别乳腺癌的信息，从而证明了乳腺腺体背景特征与病灶特征间存有一定的互补性。 本论文对乳腺背景实质的影像特性在甄别早期乳腺癌中的应用价值进行了客观定量的研究，结果表明乳腺背景的影像特征确实具有高的甄别良恶性乳腺疾病的能力，可以为乳腺癌的早期诊断提供与单一病灶相互补的有效信息。此外，因为乳腺背景实质的影像特征从未被纳入到基于DCE-MRI影像的乳腺癌早期计算机辅助诊断研究中，所以如果将乳腺背景特征合理地融合至已有的辅助诊断系统中将有望提升其辅助诊断的性能和临床应用价值。
[Abstract]:Breast cancer is the most common malignant tumor of women with high mortality. Due to the uncertainty of its pathogenic factors and the hidden nature of the disease, the patient is often diagnosed with the development of breast cancer to the middle and late stage, and the optimal period of treatment is delayed. Dynamic Contrast-enhanced magnetic resonance imaging (DCE-MRI) has been widely used in the recent years as an important early stage of breast cancer image diagnosis. At the same time, in order to improve the accuracy of the doctor's image diagnosis, the computer-aided diagnosis (CAD) technique is also put forward and applied to the clinical diagnosis and treatment decision. The computer-assisted diagnosis of early breast cancer based on DCE-MRI is a hot spot in the field of CAD. The breast cancer CAD system which has been reported has only focused on the research of the focus area, and the focus area is often divided, and then the image features are extracted according to the specific pixel points in the segmentation result, and the breast cancer diagnosis is performed by using the classifier. According to recent clinical studies, there is a certain association between the breast tissue (FGT) and the risk of breast cancer; at the image level, the background of the mammary gland (BPE) has a greater effect on the characteristics of the BPE But these studies are based on a doctor's clinical study, which is usually qualitative, with a strong physician's personal subjectivity and can't be used for clinical diagnosis. In view of these problems, the master's thesis has innovatively compared the real image character of the background of the breast in the screening of the value of the application of the breast cancer in the early stage of breast cancer. The research work of this paper mainly focuses on the screening ability of the breast cancer and the complementarity between the focus and the background of the gland. The problem is to expand. Body content includes: (1) M The original image of DCE-MRI was included in the pre-treatment of the breast image of the RI, including the noise reduction, the alignment and the research. The method of the breast segmentation based on the anatomical structure is used to obtain the breast area on the DCE-MRI image, and then the focus and the gland of the breast are determined according to the image labeling information of the doctor. background two image areas to be studied. (2) the feature extraction of the focus and the background region of the mammary gland aims at the problem of image feature extraction of the focus region and the background region of the breast, The features of 158-dimensional images were calculated from the extraction of the cases, which can be characterized by the characterization of the focal area and the back of the gland of the breast, respectively. The dynamic enhancement and the three-dimensional texture of the scene area. (3) Breast cancer screening ability of the mammary gland background feature in order to evaluate each of the calculated breast cancer screening abilities from the mammary gland background, the study uses the subject's working characteristic curve (Receiver Opera Calculation of Setting Characteristic Curve (ROC) and K-W To analyze the characteristics of the background image of each breast and to distinguish the ability of the breast lesions. The force is close to the characteristic of the focus, and has a higher effect on the breast cancer. High screening ability. (4) Complementarity between the focus and the background of the gland in order to understand the complementarity of the focus and the background of the gland and according to the selected characteristic subset, The corresponding CAD system classification performance is used to measure the effective amount of information contained in the subset of features. The experimental results show that the feature set containing the background features of the lesion and the gland of the breast is more favorable to the screening of breast cancer than the feature set containing only the feature of the lesion. It is proved that there is a certain complementarity between the characteristic of the background of the mammary gland and the characteristic of the focus. The present paper makes an objective and quantitative study on the application value of the image characteristic of the background of the breast in the screening of the early breast cancer. The results show that the image features of the background of the breast do have high screening for benign and malignant breast. The ability of the disease to provide effective information for the early diagnosis of breast cancer with respect to a single lesion. In addition, because of the substantial image characteristics of the background of the breast, It is not included in the early-stage computer-assisted diagnosis of breast cancer based on DCE-MRI images, so if the characteristics of the breast background are properly melted
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：R737.9;R445.2;TP391.72
[Abstract]:Breast cancer is the most common malignant tumor of women with high mortality. Due to the uncertainty of its pathogenic factors and the hidden nature of the disease, the patient is often diagnosed with the development of breast cancer to the middle and late stage, and the optimal period of treatment is delayed. Dynamic Contrast-enhanced magnetic resonance imaging (DCE-MRI) has been widely used in the recent years as an important early stage of breast cancer image diagnosis. At the same time, in order to improve the accuracy of the doctor's image diagnosis, the computer-aided diagnosis (CAD) technique is also put forward and applied to the clinical diagnosis and treatment decision. The computer-assisted diagnosis of early breast cancer based on DCE-MRI is a hot spot in the field of CAD. The breast cancer CAD system which has been reported has only focused on the research of the focus area, and the focus area is often divided, and then the image features are extracted according to the specific pixel points in the segmentation result, and the breast cancer diagnosis is performed by using the classifier. According to recent clinical studies, there is a certain association between the breast tissue (FGT) and the risk of breast cancer; at the image level, the background of the mammary gland (BPE) has a greater effect on the characteristics of the BPE But these studies are based on a doctor's clinical study, which is usually qualitative, with a strong physician's personal subjectivity and can't be used for clinical diagnosis. In view of these problems, the master's thesis has innovatively compared the real image character of the background of the breast in the screening of the value of the application of the breast cancer in the early stage of breast cancer. The research work of this paper mainly focuses on the screening ability of the breast cancer and the complementarity between the focus and the background of the gland. The problem is to expand. Body content includes: (1) M The original image of DCE-MRI was included in the pre-treatment of the breast image of the RI, including the noise reduction, the alignment and the research. The method of the breast segmentation based on the anatomical structure is used to obtain the breast area on the DCE-MRI image, and then the focus and the gland of the breast are determined according to the image labeling information of the doctor. background two image areas to be studied. (2) the feature extraction of the focus and the background region of the mammary gland aims at the problem of image feature extraction of the focus region and the background region of the breast, The features of 158-dimensional images were calculated from the extraction of the cases, which can be characterized by the characterization of the focal area and the back of the gland of the breast, respectively. The dynamic enhancement and the three-dimensional texture of the scene area. (3) Breast cancer screening ability of the mammary gland background feature in order to evaluate each of the calculated breast cancer screening abilities from the mammary gland background, the study uses the subject's working characteristic curve (Receiver Opera Calculation of Setting Characteristic Curve (ROC) and K-W To analyze the characteristics of the background image of each breast and to distinguish the ability of the breast lesions. The force is close to the characteristic of the focus, and has a higher effect on the breast cancer. High screening ability. (4) Complementarity between the focus and the background of the gland in order to understand the complementarity of the focus and the background of the gland and according to the selected characteristic subset, The corresponding CAD system classification performance is used to measure the effective amount of information contained in the subset of features. The experimental results show that the feature set containing the background features of the lesion and the gland of the breast is more favorable to the screening of breast cancer than the feature set containing only the feature of the lesion. It is proved that there is a certain complementarity between the characteristic of the background of the mammary gland and the characteristic of the focus. The present paper makes an objective and quantitative study on the application value of the image characteristic of the background of the breast in the screening of the early breast cancer. The results show that the image features of the background of the breast do have high screening for benign and malignant breast. The ability of the disease to provide effective information for the early diagnosis of breast cancer with respect to a single lesion. In addition, because of the substantial image characteristics of the background of the breast, It is not included in the early-stage computer-assisted diagnosis of breast cancer based on DCE-MRI images, so if the characteristics of the breast background are properly melted
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：R737.9;R445.2;TP391.72
【参考文献】
相关期刊论文 前10条
1 娄路馨;彰俊杰;时高峰;;MR动态增强及减影成像对乳腺良恶性疾病的诊断价值[J];临床放射学杂志;2007年02期
2 陶卿;姚穗;范劲松;方廷健;;一种新的机器学习算法:Support Vector Machines[J];模式识别与人工智能;2000年03期
3 许光中;李凯;封国生;;3种影像学检查方法在乳腺癌早期诊断中的作用[J];首都医科大学学报;2009年03期
4 司淑芳;贾红岩;;多种影像学联合检查诊断早期乳腺癌的临床价值[J];山东医药;2010年21期
5 郭阳,张继武,张祥德;MRI图像信噪比提高技术[J];生物医学工程学杂志;2002年03期
6 孙强;;乳腺癌的早期诊断[J];实用医学杂志;2007年01期
7 吴炯,张秀彬,张峰,门蓬涛,孙志e
本文编号：2340542
本文链接：https://www.wllwen.com/yixuelunwen/fangshe/2340542.html
