基于超声图像的颈动脉运动估计

发布时间：2018-03-31 10:28

本文选题：动脉粥样硬化　切入点：颈总动脉　出处：《哈尔滨工业大学》2017年硕士论文

【摘要】：世界卫生组织报告显示,心血管疾病在2008年就已经超越肿瘤性疾病成为危害世界人民健康的头号杀手,其发病率和死亡率居于全球第一。心血管疾病是累及全身血管的病变,即使使用目前最先进、最完善的治疗手段,仍然不能完全治疗该疾病。引起心血管疾病的主要原因是动脉硬化,动脉硬化的早期表现是血管壁内中膜的变硬和增厚,从而降低血管弹性。血管壁二维运动的检测是血管弹性检测的基础,因此血管壁的二维运动在近年来被认为是心血管事件的重要标记。本文提出估计超声图像的劲动脉运动方法,从而对血管的健康状况提供有用信息,帮助我们尽早对亚临床高危患者进行有效治疗。本文的方法从临床患者B超声图像中评估颈总动脉壁的径向和纵向运动。为了估计心动周期内选定点的轨迹,本文提出了一个块匹配方法,当涉及到一个时间更新的参考块时,使用一个像素级的非线性无损卡尔曼滤波器,并且滤波器使用初始灰度级作为控制信号以避免由于累积误差而造成发散。在这项研究中,本文不仅与以前研究方法结果进行比较,并且对比了块匹配方法中不同的匹配准则,以及无损变换中不同的sigma策略,同时对初始值进行整体的实验结果对比。在结果对比中,又比较了不同身体状况、性别、以及年龄等受试者的情况,来分析不同情况受试者之间的微小差异。其中,运动追踪结果与手动跟踪结果之间的结果比较主要在以下几个方面进行分析:相关性,一致性以及与误差。通过线性回归研究相关性,该分析方法可以根据Pearson相关系数,测量误差和系统误差三个指标。该分析方法可以测量出由本文的方法和手动跟踪方法执行颈总动脉壁的运动轨迹之间的统计关系。本文已经在103个受试者的超声图像序列中测试了本文的算法,并将本文算法得到的结果与以前方法的结果以及两个超声医生手工得到的结果进行对比。实验结果表明,本文的方法具有与超声医生的手动结果有很高的相关性(纵向方向相关系数是0.9957,径向相关系数是0.9959)和一致性(95%的置信区间宽度对于纵向方向运动是0.8871 mm,径向方向运动是0.4159 mm),并且比以前的方法准确度更高。这可以证明本文的方法在颈总动脉壁运动跟踪的有效性。
[Abstract]:The World Health Organization (WHO) reports that cardiovascular disease has surpassed tumor disease as the world's leading killer of people's health in 2008, with morbidity and mortality ranking first in the world.Cardiovascular disease is a disease involving systemic blood vessels. Even with the most advanced and perfect treatment, it is still unable to treat the disease completely.The main cause of cardiovascular disease is arteriosclerosis. The early manifestation of arteriosclerosis is the hardening and thickening of the medial membrane of the vascular wall, which reduces the elasticity of the blood vessel.The detection of two-dimensional motion of vascular wall is the basis of vascular elasticity detection, so the two-dimensional motion of vascular wall is regarded as an important marker of cardiovascular events in recent years.In this paper, a method of estimating arterial motion in ultrasound images is presented, which can provide useful information for the health status of blood vessels and help us to effectively treat subclinical high risk patients as soon as possible.Methods Radial and longitudinal motion of common carotid artery was evaluated by B-ultrasound images in clinical patients.In order to estimate the trajectory of selected points in cardiac cycle, a block matching method is proposed. When a time-updated reference block is involved, a pixel level nonlinear lossless Kalman filter is used.And the filter uses the initial gray level as the control signal to avoid divergence caused by accumulated errors.In this study, we not only compare the results with the previous methods, but also compare the different matching criteria in the block matching method and the different sigma strategies in the lossless transformation. At the same time, the experimental results of the whole initial value are compared.In the result comparison, we analyzed the slight differences among the subjects in different physical condition, sex, age and so on.The results of motion tracking and manual tracking are compared in the following aspects: correlation, consistency and error.The correlation is studied by linear regression. The method can be based on three indexes: Pearson correlation coefficient, measurement error and system error.This method can measure the statistical relationship between the motion trajectory of the common carotid artery performed by the present method and the manual tracking method.In this paper, we have tested our algorithm in 103 subjects' ultrasound image sequences, and compared the results obtained by our method with those obtained by previous methods and by hand by two ultrasound doctors.The experimental results show that,The method presented in this paper has a high correlation with the manual results of the ultrasound doctor (the longitudinal correlation coefficient is 0.9957, the radial correlation coefficient is 0.9959) and the consistent confidence interval width of 95% is 0.8871 mm for the longitudinal movement and the radial direction.The motion is 0.4159 mm and is more accurate than previous methods.This can prove the effectiveness of this method in common carotid artery wall motion tracking.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R445.1;TP391.41

【参考文献】