基于NSS的超声造影分析系统对心肌梗死诊疗的精准监测

It has been a poor prognosis and long-term survival in acute myocardial infarction (AMI) patients with dysfunction of myocardial microcirculation, which leads to reduced cardiac function and enlarged myocardial infarct size. Currently, the evaluation of microcirculation and cardiac function are unavailable due to lack of noninvasive monitor methods. Myocardial contrast echocardiography（MCE）can provide the possibility for microcirculation and cardiac function. However, there are still many shortcomings during the process of the MCE image analysis, such as subjectivity, poor reproducibility and time-consuming. To solve these shortages, data mining and machine learning theories will be combined with medical information in this research, and a novel computer-aided image analysis system for MCE based on neutrosophic similarity score (NSS) will be developed in the study. Our purpose includes: ①In vitro study: Endocardium and epicardium in MCE image will be automatically identified and the myocardium will be automatically located based on NSS analysis. This novel image analysis system for MCE will successfully accomplish automatic positioning of the myocardium, reasonable partitions of myocardium, quantitative calculation of myocardial microcirculation perfusion intensity, color-coded map of the myocardial perfusion, and measurement of ejection fraction and infarct size. ②In vivo study: Myocardial microcirculation, cardiac function and infarct size will be evaluated in the SD rats model of AMI and myocardial microvascular regeneration. The accuracy and practicability of the novel computer-aided system will be validated through cardiac catheterization, immunofluorescence and Masson staining. The system will provide imaging evidence for clinical precise diagnosis and treatment of AMI.

急性心肌梗死（AMI）存在严重微循环灌注障碍，心肌灌注异常会导致心功能减低，梗死面积扩大，严重影响患者预后。目前临床缺乏客观定量评估心肌微循环的无创影像学方法。心肌超声造影（MCE）为解决这个难题提供了新方法，但MCE图像分析时存在主观性强、重复性差及费时等不足。针对这些问题，本研究将数据挖掘技术及机器学习理论与医学信息结合，研制一套新型计算机辅助MCE定量诊断系统，拟进行：①体外实验：在前期医工合作基础上，研究中智相似积分（NSS）算法自动分割心肌；进而研发包括自动合理分区、定量测定与彩色编码心肌微循环灌注、自动计算射血分数及梗死面积等多功能一体化MCE图像分析系统；②体内实验：建立AMI和心肌微血管再生动物模型，应用新辅助系统评估心梗诊疗中心肌微循环、心功能及梗死面积等变化，利用免疫荧光微血管染色、心导管及Masson染色等方法验证其准确性与实用性，为临床精准诊治AMI提供影像学依据。

心肌梗死（MI）一旦发生会造成心脏机能不可逆改变，严重影响患者健康。尽早发现心肌微循环灌注异常，准确计算梗死面积和心功能是临床医生在MI诊疗过程中最关注的问题和亟待解决的难题。心肌超声造影（MCE）为解决这个难题提供了新方法，它不仅有助于测量左室大小，计算左室射血分数（LVEF），还能较准确评价心肌微循环。然而分析MCE图像时存在耗时、主观性强及重复性差等缺点。针对这些缺点，本课题组开发了一套基于中智相似积分（NSS）算法的MCE定量分析系统，开展以下研究：①NSS算法对MCE图像中左室心肌的自动分割。先将左室MCE图像转换为中智图像，自动计算超声中智图像相似度消除图像中不确定信息，利用中智C均值聚类初始化超声图像区域分为背景区域、组织区域和心腔区域，用心腔区域作为水平集的初始区域开始寻优迭代，寻找心肌内外边界，即心内膜与心外膜，进而识别心肌区域。结果显示NSS算法较传统淘汰粒子群优化聚类和无边界活动轮廓模型分割效果更佳；②自动测量MI大鼠LVEF变化。结果显示NSS法与双平面Simpson法和心导管法测量LVEF偏倚分别为0.36±2.14%和0.06±1.69%，测量最大差值分别为4.98%和3.50%，说明NSS法与两种公认方法所测量LVEF具有较好的一致性；③识别缺血再灌注大鼠的梗死心肌和存活心肌。通过自动计算每个心肌节段室壁增厚率（WT）和造影剂强度（CI），再将危险心肌节段分为三个区域：危险中央区（WT<0.3且CI＜-54Pix）、危险周边区（WT<0.3且CI＞-54Pix）和危险恢复区（WT>0.3且CI>-54Pix），病理显示危险周边区和恢复区均为存活心肌，危险中央区为梗死心肌，说明MCE分析系统能自动判断存活或梗死心肌；④定量评估转染血管内皮生长因子的骨髓间充质干细胞治疗MI大鼠后心肌微循环改善情况。系统通过计算危险心肌节段CI值，实现了动态监测心肌微循环变化，结果显示：治疗组梗死心肌周边再生血管较对照组增多，NSS系统测得危险心肌CI值与免疫组化微血管密度呈正相关（r=0.81）；系统计算的梗死面积与Masson染色的梗死面积呈正相关（r=0.83）。综上，本研究首次研制的基于NSS算法的MCE图像分析系统实现了自动测定LVEF、定量计算心肌微循环及梗死面积等多功能一体化的目标，在MI诊疗精准监测方面具有重要的科学意义。