多参数磁共振成像纹理特征定量评价冠状动脉粥样硬化斑块易损性的研究

Coronary atherosclerotic plaque (CAP) rupture, eventually inducing thrombus formation, is the main cause of acute coronary syndrome (ACS). Studies of CAP vulnerability are essential for early diagnosis and prognosis of ACS. Magnetic resonance imaging (MRI), as an effective way for detecting CAP, is able to quantitatively evaluate CAP in terms of histological composition and morphological attributes via the combination of pathology analysis. However, there is so far still lack of reliable, non-invasive, MRI based imaging markers for automatic identification and quantitative assessment of CAP vulnerability. We propose a new method for a computer-aided diagnosis based on multi-parametric MRI texture features. We compute the static and dynamic texture features of CAP on multi-parametric MRI to capture the difference in tissue structure and function between vulnerable and stable plaques. The MRI markers can be identified through a feature selection method using variable importance projection in conjunction with principal component analysis and a classifier fusion strategy. Following the results, we investigate the quantitative correlation between the multi-parametric MRI markers and histopathology features of CAP to establish a non-invasive, comprehensive, multi-parametric MRI based computer aided diagnosis system in order to provide a supplementary opinion in the early therapeutic decision-making for prevention of ACS.

冠状动脉粥样硬化斑块(CAP)破裂进而血栓是导致急性冠脉综合征(ACS)的主要原因。加强CAP易损性的研究对于早期诊断和预测ACS具有重要意义。磁共振成像(MRI)，作为检测CAP的一种有效方法，结合病理分析，虽然能够对CAP的组织学构成、形态特征进行定量评估，但目前仍然缺乏用于自动识别和定量评价CAP易损性的可靠的、非侵入的MRI图像标志物。本项目提出了一种基于多参数MRI纹理特征的计算机辅助诊断的新方法。通过研究CAP在多参数MRI的静态和动态纹理特征，有效描述易损斑块和稳定斑块在组织结构和功能上的差异，采用变量重要性投影和主成分分析相结合的特征选取方法以及分类器融合策略，发现MRI图像标记物；研究CAP的多参数MRI图像标记物与组织病理学特征的定量相关性，最终建立一种无创的、较为全面的、基于多参数MRI的计算机辅助诊断系统，为临床早期制定治疗方案及预防ACS发生提供一种参考依据。

冠状动脉粥样硬化斑块(CAP)破裂进而血栓是导致急性冠脉综合征(ACS)的主要原因。加强CAP易损性的研究对于早期诊断和预防ACS具有重要意义。本课题完成了80例样本的冠状动脉多参数磁共振图像以及临床数据的采集，实现了通过多结构序列和动态增强磁共振图像的多种纹理特征对易损和稳定斑块进行自动分类和定量评估，初步构建了多参数磁共振图像标记物与组织病理学及病理图像特征相关性定量模型。实验结果表明基于磁共振图像的联合多种动态纹理特征能够用于评估斑块的不稳定性，为定量评价和准确预测冠状动脉粥样硬化斑块易损性风险建立一种无创的、较为全面的、有效的基于磁共振图像的计算机辅助诊断系统提供技术基础。工作发表论文7篇。已经和一家科技公司展开合作，积极进行成果转化，目前合作研究已经申请专利2项。