精确量化评估冠脉狭窄对诊断冠心病有着重要作用。本项目拟提出基于多维几何特征的冠脉狭窄量化评估的新方法。该方法相比传统的一维冠脉狭窄评估，将具有更好的相关性。对应冠脉CT造影（CTA）图像处理与量化评估冠脉狭窄这两个步骤，本项目拟着重研究两类问题：1）针对CTA图像结构细小、灰度分布不均匀等困难，拟提出各向异性的图分割算法的冠脉分割的新方法。应用海森矩阵生成血管测度与向量场，向量场的各向异性被加权至图分割算法的能量函数。2）相比传统的一维冠脉狭窄评估方法（假阳性率高达50%），拟提出基于多维几何特征的冠脉狭窄评估的新思路和新方法。在冠脉模型上定义基于二维横截面的局部狭窄函数，并沿中心线曲线积分得到区间狭窄函数。通过对区间狭窄函数衍生的能量模型的最优化，可检测冠脉狭窄位置并量化评估狭窄程度。本课题通过多维几何特征量化评估冠脉狭窄，在国内外都属首次，将推进科学计算在临床医学和研究中的应用。

Precisely quantifying and evaluating coronary stenosis is significant for diagnosis for coronary heart disease. This project will proposal a novel method to quantify and evaluate coronary stenosis based on multi-dimensional geometric characteristics, which is supposed to have better correlation with clinical gold standard than traditional diameter stenosis evaluation. As for coronary CT angiogram image processing and coronary stenosis quantification, this project aims to address two main issues: 1) to tackle difficulties such as tiny structure, inhomogeneity of intensity in CTA images, a novel graph-cuts method with anisotropy will be adopted. Hessian matrix is used to generate vesselness measure and vector field. Anisotropy from vector field is introduced to the energy functional used in graph-cuts. 2) Rather than the conventional one dimensional coronary stenosis evaluation(false positive rate up to 50%), an novel evaluation method based on multi-dimensional geometric characteristics is proposed. Local stenosis function based on sectional area is defined on the coronary model. This stenosis function is then integral along the centerline of coronary model to obtain the stenosis function of a certain span. Through optimizing the energy function derived from the span stenosis function, the stenosis position could be localized and the stenosis severity could be quantified. This project aims to use multi-dimensional geometric characteristics to evaluate the coronary stenosis. This first attempt will promote scientific computation in clinical practice and research.