基于非局部均值的血管内超声序列运动补偿方法研究

发布时间：2018-06-08 13:26

本文选题：血管内超声 + 运动补偿　；参考：《华中科技大学》2014年博士论文

【摘要】：腹主动脉瘤是最常见的动脉瘤之一,是一种严重威胁人类生命的大血管疾病。覆膜支架腔内隔绝术是目前治疗常规型腹主动脉瘤的最佳选择。然而,由于特殊型腹主动脉瘤不满足安置覆膜支架的临床要求,常规的覆膜支架腔内隔绝术会封闭重要器官的血管入口,最终导致重要器官的衰竭。而定做带有分支的支架则有价格贵、定做周期长以及临床操作难度大等问题。因此,本文首先提出了利用血管内超声(Intravascular Ultrasound:IVUS)精确定位被覆膜支架覆盖的重要动脉分支入口,然后再使用激光进行开窗以解决重要分支动脉被覆膜支架阻隔的问题。然而,在IVUS成像过程中,由于心脏搏动以及导管的旋转,IVUS序列图像往往存在运动伪影,直接影响到医生对于血管分支入口直径测量,对分支血管入口的判断产生不利影响,给医生采用激光介入治疗带来困难。另外,运动伪影也会直接影响到血管内斑块的测量以及三维显示效果。IVUS序列图像的运动补偿技术旨在解决IVUS序列图像中的运动伪影,为医生提供更加精确的血管内三维结构信息,有助于医生准确无误地判断分支血管入口。此外,运动补偿技术还可以为斑块以及血管弹性测量提供精确的信息。因此,运动补偿技术不但是本论文迫切需要解决的技术难题,而且具有广泛的临床应用价值。目前,IVUS序列图像运动补偿方法主要是基于门控、动力学建模、模板匹配或者光流等。其中,门控法在一个心动周期只能采集一幅图像；动力学建模无法提供精确的旋转角估计结果；模板匹配会受到图像灰度变化的影响；而光流法需要IVUS序列图像满足血管中心不发生位移的条件。因此在实际应用中,四种方法的结果并不理想。本文针对IVUS运动补偿技术的难点,结合医生的临床需求,研究基于非局部均值的IVUS序列图像运动补偿方法,具体研究内容如下：首先研究了基于非局部均值的IVUS图像降噪技术：针对IVUS图像中斑点噪声特性,提出了一种基于权重更新的非局部均值去斑点噪声方法。方法针对当前应用广泛的图像斑点噪声模型,采用贝叶斯理论,推导出一种适合于斑点噪声的权值计算公式,并利用主成分分析方法(Principal Component Analysis:PCA)将其投影至PCA子空间并迭代更新权值,迭代过程可自动终止。该方法能够显著地去除IVUS斑点噪声,提高了图像的对比度。其次研究了基于非局部均值的IVUS序列图像粗运动补偿技术：通过对IVUS导管运动的建模,将其分为由于导管旋转而导致的旋转运动以及由于血流脉搏而导致的导管相对图像中心的偏移,并且结合了中膜的椭圆特性,提出了一种基于中膜检测的粗运动补偿方法。该方法通过一种非局部SUSAN边缘检测算子以及对感兴趣区域的控制点进行椭圆拟合来确定中膜的位置,进而利用前后帧中膜的拟合参数来对平移和旋转参数初始值进行估计,并对时间维的参数进行一维非局部均值获取最终补偿参数,继而完成运动补偿。该方法简单、快速,并且相对于前人的方法更具有普适性,能够适应不同情况下的IVUS图像；最后研究了基于非局部均值的IVUS序列图像精确运动补偿技术：针对粗运动补偿方法对于旋转参数提取不够精确的问题,利用基于中膜的粗运动补偿结果,提出了一种基于非局部光流法的精确运动补偿方法。该方法将非局部的相似度引入到Lucas-Kanade (LK)光流之中,提高了LK光流法的抗噪能力,并且提出了一种的旋转角矢量的计算方法,最终利用精确旋转参数在粗补偿结果的基础上完成整个补偿过程。该方法成功的解决了IVUS帧间的偏向旋转问题,并在补偿精度方面大大优于传统方法。
[Abstract]:Abdominal aortic aneurysm is one of the most common aneurysms and is a major vascular disease that seriously threatens human life. Endovascular stent graft exclusion is the best choice for conventional abdominal aortic aneurysm. However, special type of abdominal aortic aneurysm is not satisfied with the requirements for the placement of covered stent. The entrance of the blood vessel of the important organs is closed to the failure of the important organs, and the stent with branches is expensive, the fixed cycle is long and the clinical operation is difficult. Therefore, this paper first proposed the use of intravascular ultrasound (Intravascular Ultrasound:IVUS) to accurately locate the important arterial components covered by the covered stent. The portal was then used to open the window to solve the problem of obstructing important branch arteries.
However, in the process of IVUS imaging, because of the heart beating and the rotation of the catheter, the motion artifact often exists in the IVUS sequence image, which directly affects the doctor's measurement of the diameter of the branch entrance of the blood vessel. It has a negative effect on the judgment of the entrance of the branch vessel, which brings difficulties to the doctors using the laser to mediate the treatment. In addition, the motion artifact will also be directly shadow. The measurement of the intravascular plaque and the three-dimensional display effect.IVUS sequence image motion compensation technique aims to solve the motion artifact in the IVUS sequence image, provide the doctor with more accurate intravascular three-dimensional structure information, and help the doctor to judge the branch blood vessel entrance accurately and unmistakably. In addition, the motion compensation technique can also be used as a patch for the plaque. And the measurement of vascular elasticity provides accurate information. Therefore, motion compensation technology is not only an urgent technical problem to be solved in this paper, but also has a wide range of clinical applications.
At present, the motion compensation methods of IVUS sequence images are mainly based on gate control, dynamic modeling, template matching or optical flow. Among them, the gate control method can only collect one image in one heart cycle; dynamic modeling can not provide accurate results of rotation angle estimation; template matching will be affected by the change of image gray level; and optical flow method needs to be used. The IVUS sequence image satisfies the condition that the blood vessel center is not displaced. Therefore, in practical application, the results of the four methods are not ideal. This paper, aiming at the difficulty of the IVUS motion compensation technology, combines the clinical needs of the doctors, and studies the motion compensation method of the IVUS sequence image based on the non local mean. The specific research contents are as follows:
In this paper, a non local mean based IVUS image denoising technique is first studied. A non local mean speckle reduction noise method based on weight updating is proposed for speckle noise in IVUS images. A Bayesian theory is used to derive a right for speckle noise based on the widely used image speckle noise model. The value calculation formula, and using the principal component analysis method (Principal Component Analysis:PCA), projection into the PCA subspace and iterating the weight value iteratively, the iterative process can automatically terminate. This method can significantly remove the IVUS speckle noise and improve the contrast of the image.
Secondly, the rough motion compensation technique of IVUS sequence image based on the non local mean is studied. By modeling the motion of the IVUS catheter, it is divided into the rotation motion caused by the catheter rotation and the deviation of the relative center of the catheter caused by the blood flow pulse, and a kind of the medium is combined with the elliptical characteristics of the middle film. The method of coarse motion compensation for membrane detection. This method determines the position of the middle film by a non local SUSAN edge detection operator and the ellipse fitting to the control points of the region of interest. Then the initial values of translation and rotation are estimated by using the fitting parameters of the middle and back frames, and the parameters of the time dimension are non local. The method is simple and fast, and it is more universal and adaptable to the IVUS image in different circumstances.
In the end, the accurate motion compensation technology of IVUS sequence image based on non local mean is studied. In view of the problem that the rough motion compensation method is not accurate enough for the rotation parameter extraction, an accurate motion compensation method based on the non local optical flow method is proposed, which is based on the coarse motion compensation of the middle film. It is introduced into the Lucas-Kanade (LK) optical flow to improve the anti noise ability of the LK optical flow method, and a calculation method of the rotation angle vector is proposed. Finally, the whole compensation process is completed on the basis of the coarse compensation results. The method successfully solves the problem of deflection rotation between the IVUS frames, and the compensation precision is also solved. It's much better than the traditional method.
【学位授予单位】：华中科技大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：R543.16;R445.1

【参考文献】