冠状动脉粥样硬化进展预测创新技术研究

The development of atherosclerotic coronary artery disease is complex. Study of the relationships between multi-factorial features of local vessel and progression of coronary atherosclerosis will help insight in the natural course of plaque development. Geometrically correct 3D and 4D reconstruction of coronary arteries and accurate registration of baseline/follow-up imaging data are the precondition for accurately extraction and analysis of coronary features. This project aims at the key problems to the prediction of atherosclerosis progression: to increase efficiency and reproducibility of the geometric representations of coronary arteries in vivo, we will develop effective and robust Angiography - Intravascular Ultrasound fusion methods based on graph search algorithm by using the 4-D phase information; to accurately extract features for plaque progression, this project will develop 4-D morphological analysis and 3-D plaque component measurement methods, based on accurate registration of baseline – follow-up vessel walls; to explore the relationship between coronary features and progression of coronary atherosclerosis, we will develop multi-factorial feature optimization and classifier training methods based on machine learning approach. This project will provide technical supports for the research on natural course of plaque development, the analysis of coronary hemodynamics, and the prediction of adverse coronary events. This study would be important for both scientific research and clinical applications.

冠状动脉粥样硬化疾病的发展过程极其复杂。研究局部冠脉多因素特征和冠脉粥样硬化进展之间的关系，将有助于拓展人们对斑块发展自然病程的认知。准确获取及分析冠脉特征发展的关键前提是高真实度三维/四维冠脉重建及基线-随诊冠脉配准。本项目针对冠脉粥样硬化进展预测中的关键技术开展研究：为提高在体冠脉几何表达的效率和可重复性，基于图搜索等算法, 利用四维完整时相信息，发展高效鲁棒的四维冠脉造影-血管内超声融合方法；为准确提取斑块进展相关特征，在精确配准基线-随诊血管壁的基础上，发展四维形态学分析及三维斑块成分测量方法；为探索冠脉特征和冠脉粥样硬化进展之间的关系，发展基于机器学习的多因素特征优化及分类器训练方法。本项目的实施，将为研究斑块发展自然病程、分析冠脉血流动力学、预测冠脉不良事件等提供有力的技术支撑，具有广泛的科学研究价值和临床应用前景。

冠状动脉粥样硬化疾病发展过程极其复杂，尽管降脂类药物能有效降低冠脉斑块风险，然而目前仍然不清楚冠脉上哪些局部区域能从治疗中获益。本项目旨在提前、精准预测冠脉粥样硬化进展。主要研究内容包括基线-随诊血管内超声图像配准技术，基于统计学的斑块进展分析，和基于机器学习的多因素特征优化及分类器训练方法。本项目的开展，实现了与专家无差异的基线-随诊配准算法，以及精确到毫米级的、冠脉上未来可能出现高风险斑块的位置的预测，平均预测准确率约80%。本项目的成果，为研究斑块发展自然病程、预测冠脉斑块风险等提供了有力的技术支撑，具有广泛的科学价值和临床应用前景。