项目名称: 对颈动脉疾病进行定量监测和评估的颈动脉图集
项目编号: No.81201149
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 影像医学与生物医学工程
项目作者: 赵智远
作者单位: 香港城市大学深圳研究院
项目金额: 23万元
中文摘要: 在中国,中风是引起死亡的第二大疾病 。因此,一套能灵敏地监控颈动脉斑块对各种治疗方法所作反应的技术将对中风预防产生莫大的帮助。基于三维超声图像中分割出来的颈动脉形状,本研究提出一种构建二维标准化颈动脉图集的新算法。构建此图集的动机源自于我们对局部动脉壁和斑块在治疗过程中所发生的变化的了解相当匮乏,而这些图将有助于我们加深对这方面的认识。虽然文献里描述了一些制作二维拉平图的方法,但得到的图的形状往往根据对应的三维动脉的形状而有所不同,而后者在不同患者间差别非常大。本研究提出的算法可使得不同动脉的对应区域被映射到二维标准化图集的相同位置,故能对来自不同病人的二维图进行直接的比较。另外,在二维图集里可供比较的数据点数量相当庞大。因此,很难根据全部数据点做出临床解释。针对这种情况,我们提出一套特征选择方法来刻画出不同治疗组病人的管壁厚度变化的模式。成功开发出这两种技术将会促进我国临床试验的发展。
中文关键词: 三维超声成像;二维标准模板;基于特征选择的生物标记;分割精度与方差;血管壁分割
英文摘要: Stroke is the second cause of death in China. Improved identification of patients who are at risk for stroke and sensitive techniques for monitoring of carotid plaque response to therapies will have an enormous impact on patient management and stroke prevention. This proposal introduces a novel algorithm to construct 2D Carotid Atlases based on the carotid surfaces segmented from 3D ultrasound images. The motivation for developing 2D carotid atlas derives from our recognition and acknowledgement of the utter lack of understanding of the local arterial wall and plaque changes that occur over time and in response to therapy in patients. These maps will provide a way to display localized plaque and wall thickness for specific patient groups. Previous approaches used to generate 2D flattening maps from 3D surfaces were developed, but the resulting 2D maps depended on the shapes of the corresponding 3D carotid surface, which was highly variable from subject to subject. This proposal introduces a novel algorithm to generate 2D standardized atlases, in which corresponding 3D regions on different arteries are mapped to the same location in the 2D standardized atlases, making possible the direct and quantitative comparison of the 2D carotid maps generated from different patients. Also, the number of data points availabl
英文关键词: 3D ultrasound imaging;2D standardized map;Feature-selection-based biomarker;Segmentation accuracy & variability;Vessel wall segmentation