2024年6月14日发(作者：)

摘 要

随着计算机与影像技术的不断发展，图像分割被应用到各个领域并发挥着

巨大的作用。近年来，医学图像处理作为医学的一个重要辅助工具逐渐成为研

究热点。核磁共振图像具有软组织成像效果好，空间分辨率高的特点，因此被

广泛应用于医学领域，给患者的诊断和治疗带来了很大的帮助。脊柱核磁共振

图像包含着脊柱的重要信息，对脊柱疾病的诊断有着十分重要的意义，这毫无

疑问使得它成为重点的研究对象。本文根据脊柱核磁共振图像的特点，充分利

用在核磁共振图像中椎间盘与椎骨灰度的鲜明对比，提出了一种改进的标记符

分水岭算法对椎间盘进行分割。在分割出椎间盘之后，通过构造初始演化轮廓，

然后利用水平集曲线演化理论完成了对椎骨的提取。

本文主要工作和研究内容如下：

（1）首先分析了常用的图像处理算法，以及它们的优缺点。然后研究了几

种常用分割算法在脊柱磁共振图像上的应用情况，通过仿真结果表明这些算法

的处理效果不是很理想。最后结合脊柱磁共振图像成像特点，给出了本文算法

选择方案，提出采用分水岭和水平集算法分别对椎间盘和椎骨进行分割提取。

（2）在分析分水岭算法的基础上，完成了基于控制标记符的分水岭算法。

针对形态学去噪效果不理想的情况，通过加入基于区域面积属性的限制条件，

获取到更加精确的内部标记，成功提取出了椎间盘区域，克服了过分割的问题。

并与基于CLAHE的分水岭算法作对比，证明了本文方法的优越性。

（3）在分析水平集算法的基础上，引入距离正则化的水平集模型，并针对

该模型中边缘检测函数对噪声敏感的问题，提出新的表达形式进行改进。通过

加入区域灰度信息，改善了噪声和弱边界影响，成功提取出了椎骨的轮廓。并

与基于GVF snake形变模型算法进行对比，证明了本文方法的精确性。

本文创新点如下：

（1）在改进的控制标记符分水岭算法对椎间盘的分割中，通过计算区域面

积并给定限制范围，获得了相对精确的内部标记。

（2）将距离规则化的无须初始化水平集模型应用于脊柱MRI图像的椎骨

分割，并提出新的边缘检测函数改善噪声和弱边缘的影响。

关键词：核磁共振，脊柱，分水岭，水平集

I

Abstract

With the development of computer and image technology, image segmentation is

applied to various fields and plays an important role. In recent years, medical image

processing which is an important auxiliary tool of medical science has gradually

become a hot research. Magnetic resonance image has a good effect on soft tissue and

has high spatial resolution, therefore, it is widely used in medical field and gives great

assistance to the diagnosis and treatment of diseases. The spine magnetic resonance

image contains important information of spine, which is of great significance to the

diagnosis of spinal diseases, there is no doubt that it will become the focus of the

research object. Based on the characteristics of spine MRI, this paper makes a full use

of the gray contrast of the intervertebral disk and the vertebral in the MRI images, and

proposes an improved marker watershed algorithm to segment the intervertebral disc.

After segmenting the intervertebral disc, the level set curve evolution theory is used to

extract the vertebrae based on the initial evolution contour.

The main work and research contents are as follows:

(1)Analysis the commonly used image processing algorithms and their

advantages and disadvantages. And then do research on the application of several

commonly used segmentation algorithm in spinal MRI image, the simulation results

show that these algorithms are not satisfactory. Finally present the algorithms used in

this paper, propose a watershed and level set segmentation algorithm to segment the

intervertebral disk and vertebral based on the characteristics of spine imaging.

(2)Complete the marker controlled watershed algorithm based on the analysis of

watershed algorithm. In view of the unsatisfactory morphological denoising result, add

a limitation based on the area property to obtain more accurate internal markers, so the

intervertebral disc region was extracted successfully, and overcome the severe over

segmentation problem. In contrast to the watershed algorithm based on CLAHE, this

method is more superior.

(3)Introduce the distance regularized level set mode based on the analysis of

level set algorithm. Propose a new form of expression for the noise sensitivity of the

edge detection function in this model. By adding the region gray information, the

II