基于ACM和EMD的超声造影对心肌微循环方法学研究

It has been a poor prognosis and long-term survival in patients with dysfunction of myocardial microcirculation. Early diagnosis of micro- circulation abnormality leads to optimal treatment and favorable long-term clinical outcomes. Myocardial contrast echocardiography（MCE）is a novel imageology method for myocardial microcirculation. However, there are many shortcomings in the MCE image analysis, such as time-consuming, subjectivity, time-intensity curves are vulnerable to interference and poor reproducibility. In response to these shortages, data mining and pattern recognition theory will be combined with medical information in the study. ①Based on section method of active contour models(ACM), endocardium and epicardium in MCE image will be automatically identified and the myocardium will be automatically located. ②Based on empirical mode decomposition(EMD) and Hilbert-Huang marginal spectrum method, the time-intensity curve will be processed and the diagnostic signal will be extracted and separated automatically. ③A novel computer-aided image analysis system for myocardial contrast echocardiography based on active contour model combined with empirical mode decomposition will be developed. Automatic positioning of the myocardium, a reasonable partition of myocardium, quantitative calculation of myocardial microcirculation perfusion intensity and color-coded map of the myocardial perfusion will be successfully accomplished. Several limitations in MCE analysis will be solved and integrative multifunction will be accomplished, including analyzing automatically, calculating quantitatively myocardial perfusion and drawing the color-coded map of ischemic myocardium.

缺血心肌存在严重微循环灌注障碍，灌注异常严重影响受累心肌功能恢复，影响患者预后和长期生存率。早期诊断微灌注异常是临床亟待解决的问题，心肌超声造影（MCE）为解决这个难题提供了新方法。然而在MCE图像分析中存在许多缺点，如耗时、主观性强、时间-强度曲线干扰大及重复性差等。针对这些缺点，本研究将数据挖掘技术及模式识别理论与医学信息结合进行：①基于活动轮廓模型（ACM）分割法自动识别MCE图像中心内膜和心外膜，定位心肌区域；②基于经验模态分解（EMD）及边际谱方法处理时间-强度曲线，提取分离出有诊断意义的信号；③整合以上研究方法，研制新型计算机辅助的MCE图像自动分析系统，对心肌准确分区，定量标注各节段心肌灌注强度，彩色编码显示心肌灌注情况，优化灌注曲线。此系统顺利完成能弥补和修正以往MCE图像分析的不足，可望完成计算机自动分析、快速获得心肌灌注定量参数及彩色编码危险心肌等多功能一体化的目标。

急性冠脉综合症常存在严重微循环灌注障碍，灌注异常严重影响受累心肌功能恢复，影响患者预后和长期生存率。早期诊断微灌注异常是临床亟待解决的问题，心肌超声造影（MCE）为解决这个难题提供了新方法。然而在MCE图像分析中存在许多缺点，如耗时、主观性强、时间-强度曲线干扰大及重复性差等。针对这些缺点，本研究将数据挖掘技术及模式识别理论与医学信息结合进行：①超声造影图像分割：心内膜分割基于二值形态学的原理，采用膨胀、腐蚀、闭合运算以及基于对象标记的连通域搜索的方法；心外膜分割是基于Mumford-Shah模型的水平集活动轮廓模型（ACM）图像分割法，并结合最小二乘圆拟合对边界进行处理，将边界溢出的部分进行修剪接合。对于高质量和中等质量的图像，自动分割获得心内外膜轮廓与人工分割的轮廓具有良好的吻合度，但是对于低质量的图像，心内膜吻合度较高，而心外膜分割误差较大；②计算机辅助MCE图像分析系统可以实现对心肌准确分区，自动测量左室面积变化率（AVF），定量标注各节段心肌灌注强度，彩色编码显示心肌灌注情况及优化灌注曲线；研究显示：计算机获得的面积变化率（AVF）与射血分数（EF）呈正相关(r = 0.934，P<0.001)；术后3周时，前壁危险节段CI值与免疫组化微血管密度呈正相关(r = 0.852，P<0.01)；术后3周MI组梗死节段明显变薄，呈大片红色渲染，其它节段显示为绿色和蓝色混杂，与微球染色及TTC染色明显相似；③基于经验模态分解（EMD）和希尔伯特变换（HHT）分别得到各分区曲线的本征模态函数和残差曲线，经过HHT后得到瞬时特征量及边际谱，进而得出一种诊断心肌梗死的新方法。此系统顺利完成弥补和修正以往MCE图像分析的不足，完成了计算机自动分析、快速获得心肌灌注定量参数及彩色编码危险心肌等多功能一体化的目标。