快速水平集方法：磁共振血管造影图像分割新策略

6 浏览量更新于2024-08-28 收藏 450KB PDF 举报

"MRA图像分割：一种改进的水平集方法" 在医学成像领域，无监督的体积数据分割是一项极具挑战性的任务。水平集方法近年来因其独特的特性在图像分割中得到了广泛应用。它结合了全局平滑性和拓扑改变的灵活性，使得在处理复杂图像结构时比传统的统计分类方法更具优势。然而，标准的水平集方法通常需要大量的迭代过程，导致计算效率较低。针对这一问题，研究人员提出了一种基于分水岭算法的快速水平集框架，用于解决从磁共振血管造影（MRA）图像中分割出3D人脑血管结构的问题。MRA是一种非侵入性的成像技术，能够显示血管的详细结构，但其图像中的血管几何形状和强度模式复杂，使得自动分割变得非常困难。分水岭算法是一种经典的图像分割方法，通过模拟地貌地形，将图像看作地貌表面，寻找像素强度值的局部极小值，将其作为“流域”，从而实现分割。将分水岭算法与水平集方法相结合，可以在保持分割精度的同时，减少计算复杂度，提高分割速度。论文中所提出的改进方法首先利用分水岭算法对图像进行预处理，快速地找出潜在的分割边界。然后，这些边界被引入到水平集框架中，作为初始化的演化曲线。在随后的演化过程中，水平集函数被优化，以适应图像特征并精确地追踪目标结构。这种方法巧妙地利用了分水岭算法的快速性，同时保持了水平集方法的灵活性，有效地解决了血管结构的复杂性问题。实验结果证实，该方法在处理MRA图像时，能实现快速且准确的分割，对于复杂的3D人脑血管结构，其分割效果显著。这对于进一步的血管分析、疾病诊断和治疗规划等具有重要意义。同时，这种融合不同算法的策略也为其他领域的图像分割问题提供了新的思路。关键词：生物医学测量，水平集方法，磁共振血管造影，分水岭算法

1316 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 56, NO. 4, AUGUST 2007

Segmentation for MRA Image: An

Improved Level-Set Approach

Jiasheng Hao, Yi Shen, Member, IEEE, and Qiang Wang, Member, IEEE

Abstract—Unsupervised segmentation of volumetric data is still

a challenging task. Recently, level-set methods have received a

great deal of attention, which combine global smoothness with

the ﬂexibility of topology changes and offer signiﬁcant advan-

tages over conventional statistical classiﬁcation. However, level-set

methods suffer from heavy computational burden because of a

lot of iterations. We present a fast level-set framework based

on the watershed algorithm for the segmentation of complicated

structures from a volumetric data set. The driving application

is the segmentation of 3-D human cerebrovascular structures

from magnetic resonance angiography, which is known to be a

very challenging segmentation problem due to the complexity

of vessel geometry and intensity patterns. Experimental results

show that the proposed method gives fast and accurate excellent

segmentation.

Index Terms—Biomedical measurements, image segmentation,

level set, magnetic resonance angiography (MRA), watershed

algorithm.

I. INTRODUCTION

EGMENTATION is a fundamental problem in almost all

computer vision and medical image analysis, and unsuper-

vised segmentation of volumetric data is still a challenging task.

Recently, the level-set method has received a great deal of atten-

tion, particularly for volumetric data segmentation [1]–[3]. The

particular interest in this paper is to reduce the computational

complexity of level-set methods to segmentation.

The level-set method was introduced by Osher and Sethian

[4] to track the evolution of interfaces and was found to

overcome many limitations of the active contour methods

(ACMs). ACMs (also called snakes) were originally deﬁned by

Kass et al. [5] as an energy-minimizing spline and since,

then have, been widely used for extracting the object of in-

terest. However, traditional snake models suffer from sev-

eral limitations: 1) They are sensitive to the initial contours;

2) they are nonfree parameters; 3) they cannot handle changes

in topology; and 4) they have a tendency to produce degenerate

contours or self-intersections. The level-set method is appealing

for its ability to handle topological changes automatically.

Its numerical implementation is also straightforward for any

dimension. While the level-set method has many advantages,

Manuscript received June 15, 2006; revised April 3, 2007. This paper was

presented in part at the IEEE Instrumentation and Measurement Technology

Conference, Sorrento, Italy, April 24–27, 2006.

The authors are with the Department of Control Science and Engineer-

ing, Harbin Institute of Technology, Harbin 150001, China (e-mail: haojs@

hit.edu.cn).

Color versions of one or more of the ﬁgures in this paper are available online

at http://ieeexplore.ieee.org.

Digital Object Identiﬁer 10.1109/TIM.2007.899839

its implementation, which is based on the solution of certain

partial differential equations (PDEs), results in a signiﬁcant

computational burden, which bars its application in complex

segmentation such as magnetic resonance angiography (MRA).

On the other hand, most of the iterations of the level-set

evolution in those segmentations may be unnecessary, in fact,

because it evolves quite inside the boundary. For example, the

vessels are always very long. Thus, the question is whether we

can minimize the iterations of the level-set evolution without

losing accuracy. This is achieved in this paper.

Our driving problem is the segmentation of vessels within

MRA images. There are two major classes of these approaches.

One class of methods is based on a statistical model, which

classiﬁes voxels within the image volume into either vascular

or nonvascular class for time-of-ﬂight MRA [1]. Another

class of segmentation is based on intensity threshold, where

points are classiﬁed as either greater than or less than a

given intensity. This is the basis of the isointensity surface

reconstruction method [6]. This method suffers from errors

due to image inhomogeneities; in addition, the choice of the

threshold level is subjective. An alternative to segmentation

is axis detection, which is known as skeletonization process,

where the central line of the tree vessels is extracted based

on the tubular shape of vessels [7]. Reﬁnements using other

techniques such as morphological methods for segmenting the

vascular have improved, yet most require an interventional

step like specifying seed locations, intensity values, and other

interventional parameters [8].

In this paper, we propose a fast level-set framework based on

the watershed algorithm for MRA segmentation, allowing a sig-

niﬁcant reduction in computation complexity and the promise

of real-time implementation. Thus, the overall cerebrovascular

segmentation turns into an efﬁcient, nearly real-time process

with intuitive and usefully restricted user interaction. This

approach is applied on different types of MRA data sets and

shows good results.

The rest of this paper is organized as follows. Section II

sets the scene by brieﬂy describing level-set methods and

watershed algorithms. Section III presents the new approach. In

Section IV, we apply the framework to the MRA data set and

give the results. Finally, conclusions are drawn in Section V.

II. P

REVIOUS WORK

A. Level-Set Evolution

The central idea of the Osher–Sethian level-set method [4] is

to track the motion of an interface by embedding a propagating

下载后可阅读完整内容，剩余5页未读，立即下载

weixin_38581308

粉丝: 2
资源: 893

快速水平集方法：磁共振血管造影图像分割新策略

Metropolis-SA算法在脑部MRA图像分割中的应用

概率混合模型在MRA血管分割中的应用

改进的MRF方法在脑部MRA三维血管分割中的应用

MATLAB图像分割：图像分析的基石，掌握图像分割的奥秘

一种改进的MRA图像脑血管提取方法

mra_hsi:用于MRA报告的HSI分析

mra-protocol:Mail.Ru 代理协议实现

mra-portal:模拟定向医疗保健福利消费者门户

卫星经纬高matlab代码-MRA-Serial:多分辨率近似（MRA）的串行实现

卫星经纬高matlab代码-MRA-Parallel:多分辨率近似（MRA）的并行实现

最新资源