精确评估心肌活力对诊断心梗患者有着重要作用。针对延迟钆增强（LGE）图像评估心肌活力的缺点（半定量式，主观性强，毒副作用），本项目拟提出基于常规Cine MRI与左室曲面限制质心化沃罗诺伊分割(CCVT)提取左室几何动态特征以评估心肌活力的新方法。本项目拟着重研究三类问题：1）针对扫描过程中图像发生运动偏移，拟提出结合图像配准与点集配准的左室运动校正的新方法。2）针对图像层间距离较大导致的左室三维建模的不确定性，拟提出基于变分方法与图分割算法的重构左室曲面模型的新方法。3）计算左室曲度属性与CCVT，并提取曲度动态特征，应用无监督学习方法，建立自动评估心肌活力的新模型。该课题通过Cine MRI图像评估心肌活力的研究在国内外都属首次，将推进科学计算与机器学习在临床医学中的应用。

Accurately evaluating cardiac viability is significant for diagnosis for myocardial infarction patients. To avoid the disadvantages of late Gadolinium enhancement (LGE) images as the evaluation tool for cardiac viability such as semi-quantitative, subjective, side effect, this project will propose a novel method to evaluate cardiac viability based on left ventricle geometric features from Cine MRI and constrained centroidal Voronoi tessellation (CCVT) . Three issues will be addressed: 1) to correct the motion from scanning process, a novel method is proposed incorporating image-based and point-based registration; 2) to tackle the uncertainty to modeling left ventricle due to the large slice spacing in MRI, a novel method based on variational model and graph-cuts is proposed to reconstruct left ventricle model; 3) Compute left ventricular curvedness characteristics and CCVT, based on which curvedness dynamic features are extracted. Unsupervised learning methods are applied on the obtained feature data and a new model to automatically assess cardiac viability is to be built. This project aims to use Cine MRI to evaluate cardiac viability. This first attempt will promote scientific computation method and machine learning application in clinical practice and research.