脑核磁共振图像分割算法研究

 2022-01-17 11:01

论文总字数:27318字

目 录

摘要 1

1 绪论 1

1.1 脑核磁共振图像的研究背景及意义 3

1.2 脑核磁共振图像分割算法概述 3

1.3 本文的主要研究工作和创新点 4

2 基于标准模糊C均值的脑核磁共振图像分割算法 4

2.1 K均值算法概述 4

2.1.1 K均值模型 4

2.1.2 K均值算法 5

2.2标准模糊C均值算法概述 5

2.2.1 标准模糊C均值模型 5

2.2.2 标准模糊C均值求解 5

2.3实验结果与分析 7

3 基于高斯混合模型的脑核磁共振图像分割算法 8

3.1 高斯混合模型 8

3.2 EM算法 9

3.3 高斯混合模型分割算法 10

3.4 实验结果与分析 11

4 基于非局部空间信息的FCM混合模型 12

4.1 偏移场概述 12

4.1.1 Legendre多项式模型 12

4.1.2 偏移场恢复 13

4.2 非局部均值概述 13

4.3 基于非局部空间信息的FCM混合模型 13

4.3.1 混合模型构建 13

4.3.2 混合模型求解 14

4.3.3 混合模型算法 15

4.4 实验结果与分析 16

4.4.1对比模型的实验参数设置 16

4.4.2 虚拟脑部合成图像分割实验 16

4.4.3 真实脑部合成图像分割实验 19

5 总结与展望 21

5.1 总结 21

5.2 未来展望 21

参考文献 22

致谢 24

脑核磁共振图像分割算法研究

李剑

, China

Abstract:

Recently, with the rapid development of Medical Imaging Technology, which has been successful. Especially the application of magnetic resonance imaging (MRI), with the characteristics of no intervention、not harmful、seldom effected by the motions of objection, has been used in taking pictures of medical images. Using medical image segmentation technology, clinicians can observe the lesion position of patient and analysis the shape,the border and the size of sick organs more directly and clearly, so as to make more accurate judgment and contribute to the treatment of patients.

The paper introduce the existing segmentation methods and its classification. Starting from the K-means clustering, after the introduction of the ideas and the analyzation of the advantages and disadvantages of the algorithm, Standard fuzzy C-means (FCM) is introduced to improve the segmentation ability of K-means. this thesis explores the medical image segmentation of standard FCM algorithm and studies such issues as the selection of the initial number of clustering. And the identification of the initial center and initial membership matrix. This paper also summarizes the current research of based on the Gaussian Mixture Model clustering algorithm, and research its cluster principle and the parameter estimation of EM algorithm in the process of cluster.

Due to the existence of noise and intensity inhomogeneity in brain magnetic resonance (MR) images, the existing segmentation algorithms are hard to find accurate results. In this paper, we propose a new mixture model(NLSCFCM), which incorporates non-local spatial information into the fuzzy c-mean clustering algorithm (FCM) framework, to obtain more accurate results. The method modifies the traditional regularization smoothing term by using the nonlocal information to reduce the effect of the noise. Inspired by the mechanism of the Gaussian Mixture Model, the dissimilarity function of FCM is defined by using the form of certain exponential function consisting of not only the distance but also the covariance and the prior probability to improve the robustness. Meanwhile, the bias field is modeled by using orthogonal basis functions to reduce the effect of intensity inhomogeneity. Compared with the state-of-the-art methods, experiment results obtained by employing the proposed model on real and synthetic MR images demonstrate its accuracy and robustness.

Key words: K-means clustering fuzzy clustering Gaussian Mixture Model bias field

non-local information

1 绪论

1.1 脑核磁共振图像的研究背景及意义

脑部疾病近来越来越常见,不断的腐蚀着人类身体健康,由于这些脑部疾病的存在,很多人因此丧命。科学家们通过研究核磁共振原理研究出了磁共振成像(Magnetic Resonance Imaging, MRI)技术,这种磁共振成像技术主要利用了原子核在弛豫时释放的电磁波。由于MRI特殊的成像机理,广泛用于心血管造影、心脏超声、同位素成像、断层X射线摄影术、脑核磁共振成像等医学成像技术。因而,核磁共振成像技术检查可以有助于医生们进行辅助诊断、量化分析、便于医生们发现各种隐藏的脑部疾病。其主要优点在于:1.MR对软组织的对比度最高;2.MRI能够在随便一个方向上进行分开成一层一层图像;3. MR不对人体进行侵入然后拍摄图像,因而可以认为是无害的; 4. MRI拍摄完成时参数非常多,这些参数含有丰富的人体组织的细节; 5.MRI的三维显示成像是非常的清晰的。MRI 也有其一定的缺点:1. 拍摄的过程将会持续很长一段时间;2. 在拍摄肝脏、胰腺、肾上腺时效果不是很好。

为了更好的观察人脑的内部结构以及病灶,MRI脑部扫描常分为三个方向,如图 1.1 从左到右的三个视图,它们是横断视图、矢状视图、冠状面视图。

图1.1 脑核磁共振图像的三个视图,分别是横断面、矢状面、冠状面

在临床诊断脑部疾病时,往往需要非常的精确的分割脑部MR,将其分割为人脑中最主要的三个部分,它们分别是灰质(GM),白质(WM)和脑脊液(CSF)。然而,MRI形成的大量的图像数据是人工处理、医生的个人分析所不能够接受的,而且由于人们的粗心以及个人主观因素常常会使的脑核磁的图像分割不是很准确,何况这还投入了巨大的人力,如果使得这一切都自动化,将会有助于医生,造福于人类。此外,由于MR的成像原理以及自身的缺陷,通过这种方式得到的图像往往是模糊的并且是多样的,在不同的软组织之间变的非常的模糊。这些造成了分割脑MR图像很大的困难。针对这些难点以及医学图像在现代医学领域的重要作用,图像分割算法得到了广大研究人员的研究,大量算法被人们提了出来,在脑核磁共振图像分割领域中,基于FCM聚类和基于GMM非常火热的两大统计方法。

剩余内容已隐藏,请支付后下载全文,论文总字数:27318字

相关图片展示:

您需要先支付 80元 才能查看全部内容!立即支付

该课题毕业论文、开题报告、外文翻译、程序设计、图纸设计等资料可联系客服协助查找;