地中海贫血症是一种严重危害人类健康的血液遗传病，利用心脏磁共振T2*参数方法准确地测量心脏铁沉积对于指导与监控地贫患者的除铁治疗具有重要的临床意义。这就急切地需要建立一种全自动的心脏铁沉积测量方法，然而目前尚无不受噪声影响的反映心铁空间分布的T2*图和室间隔区域自动提取模型。.本课题拟采用基于磁共振弛豫衰减曲线距离约束的非局部均值滤波和噪声补偿模型以快速精确去除非零背景噪声对T2*估计的影响进而准确求解T2*图；并在此基础上拟采用基于多图谱配准和基于KL距离加权和局部邻域信息的CV模型融合分割方法以实现左心室全心肌组织自动提取，结合经典的形态学操作、局部模糊聚类、阈值分割等方法自动提取右心室血池，最后利用解剖位置信息提取室间隔区域以计算室间隔区域的代表性T2*值。.本项目研究将为临床地贫患者提供一种全自动心脏铁沉积测量模型，辅助临床医生对地贫患者进行铁含量分级诊断和除铁治疗。

Thalassemia is a kind of hereditary hematonosis. Cardiovascular magnetic resonance (CMR) T2* is the method of choice for the assessment of myocardial iron, which is needed to detect iron early and guide chelation therapy in hospital. It is highly importmant for estalish a fully automated myocardial iron concentration (MIC) measurements for the major thalassemia. Barriers to CMR T2* assessment exist. The T2* map, which reflects the spatial distribution of MIC, cannot be efficiently and accurately caculated due to the impact from noise. There is no reliable and easy-to-implement interventricular septum extraction method for the whole cardiac T2* measurement which indicates the total MIC and has been shown to be more reproducibile..In this project, we plan to extend the nonlocal means algorithm to myocardial T2* mapping introducing a novel distance measure to include similarities between exponential decay curves based on a noise correction model for T2* relaxometry. After that, we will use the multi-atlas registration and level sets, using KL distance and local neighborhood information, for the myocardium extraction.By combination of the classical image segmentation methods(thresholding, local fuzzy clustering, connected region labeled and etc), we will segmente the interventricular septum taking spatiallocationinto account to exclude the pixels which are likely affected by the partial volume effect for calculating the representive T2* value. .This project would provide improved, reliable and easy-to-implement T2* measurment methods for the quantificaiton of MIC, and improve the therapy management of beta-thalassemia major patients.