Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, China; Shenzhen College of Advanced Technology, University of Chinese Academy of Sciences, China.
Department of Ultrasound, The Second People's Hospital of Shenzhen, China.
Med Image Anal. 2017 Apr;37:1-21. doi: 10.1016/j.media.2017.01.004. Epub 2017 Jan 10.
The dynamics of the carotid artery wall has been recognized as a valuable indicator to evaluate the status of atherosclerotic disease in the preclinical stage. However, it is still a challenge to accurately measure this dynamics from ultrasound images. This paper aims at developing an elasticity-based state-space approach for accurately measuring the two-dimensional motion of the carotid artery wall from the ultrasound imaging sequences. In our approach, we have employed a linear elasticity model of the carotid artery wall, and converted it into the state space equation. Then, the two-dimensional motion of carotid artery wall is computed by solving this state-space approach using the H filter and the block matching method. In addition, a parameter training strategy is proposed in this study for dealing with the parameter initialization problem. In our experiment, we have also developed an evaluation function to measure the tracking accuracy of the motion of the carotid artery wall by considering the influence of the sizes of the two blocks (acquired by our approach and the manual tracing) containing the same carotid wall tissue and their overlapping degree. Then, we have compared the performance of our approach with the manual traced results drawn by three medical physicians on 37 healthy subjects and 103 unhealthy subjects. The results have showed that our approach was highly correlated (Pearson's correlation coefficient equals 0.9897 for the radial motion and 0.9536 for the longitudinal motion), and agreed well (width the 95% confidence interval is 89.62 µm for the radial motion and 387.26 µm for the longitudinal motion) with the manual tracing method. We also compared our approach to the three kinds of previous methods, including conventional block matching methods, Kalman-based block matching methods and the optical flow. Altogether, we have been able to successfully demonstrate the efficacy of our elasticity-model based state-space approach (EBS) for more accurate tracking of the 2-dimensional motion of the carotid artery wall, towards more effective assessment of the status of atherosclerotic disease in the preclinical stage.
颈动脉壁的动力学已被认为是评估动脉粥样硬化疾病临床前阶段状态的一个有价值的指标。然而，从超声图像中准确测量这种动力学仍然是一个挑战。本文旨在开发一种基于弹性的状态空间方法，从超声成像序列中准确测量颈动脉壁的二维运动。在我们的方法中，我们采用了颈动脉壁的线性弹性模型，并将其转换为状态空间方程。然后，通过使用 H 滤波器和块匹配方法求解该状态空间方程，计算颈动脉壁的二维运动。此外，本研究还提出了一种参数训练策略，用于处理参数初始化问题。在我们的实验中，我们还开发了一个评估函数，通过考虑包含相同颈动脉壁组织的两个块（由我们的方法和手动跟踪获得）的大小及其重叠程度的影响，来衡量颈动脉壁运动的跟踪准确性。然后，我们将我们的方法与三名医生在 37 名健康受试者和 103 名非健康受试者上的手动跟踪结果进行了比较。结果表明，我们的方法具有高度相关性（径向运动的 Pearson 相关系数为 0.9897，纵向运动的 Pearson 相关系数为 0.9536），并且与手动跟踪方法吻合良好（置信区间为 89.62µm 的径向运动和 387.26µm 的纵向运动）。我们还将我们的方法与三种先前的方法进行了比较，包括传统的块匹配方法、基于 Kalman 的块匹配方法和光流方法。总之，我们已经成功地证明了我们基于弹性模型的状态空间方法（EBS）在更准确地跟踪颈动脉壁二维运动方面的有效性，从而更有效地评估动脉粥样硬化疾病的临床前阶段。
