在临床实践中，“灰区”CT-FFR值诊断冠脉狭窄精度低大大限制了该技术日常临床实践中的应用。现有CT-FFR技术缺少联合斑块负荷评估冠脉功能性狭窄的的机制，这为提高CT-FFR测算准确性提出了很大的挑战。如何解决冠脉病变的解剖学信息和CT-FFR耦合问题，利用深度学习构建高效的CT-FFR测算神经网络模型是本项目需要解决的关键问题。本项目通过CTA影像数据准确重构冠脉和斑块解剖结构，利用流固耦合数值计算方法研究影像学特征与斑块解剖结构力学的映射关系，实现联合斑块负荷预测冠脉CT-FFR的方法。本项目创新性地结合流固耦合构建“灰区”冠脉病变的CT-FFR测算方法，并结合多任务物理学特征驱动深度学习方法实现高效率的CT-FFR计算。本项目提出的斑块解剖结构力学估算模型，完善了量化分析斑块解剖结构力学与冠脉血流动力学关联性的机制，有望成为改善动脉功能性狭窄诊断精度的关键技术支持。

Clinical practices had illustrated a low diagnostic accuracy in coronary artery disease with a “grey zone” CT-FFR. The drawback of CT-FFR without incorporation with plaque burden for assessing functional stenosis had posed a significant challenge to improve the diagnostic performance of CT-FFR. To develop a mechanism for introducing plaque characteristics into CT-FFR calculation is a key issue in this field. In addition, high performance deep learning framework is another key issue to obtain the efficiency of CT-FFR for daily clinical usage. We proposed a method of 3D reconstruction to obtain a geometry in combination with coronary artery lumen and plaque. The relationship between image and plaque characteristics was analyzed by using fluid-structure intervention simulation, which connects the plaque burden and CT-FFR. This project innovatively develops a protocol of using fluid-structure interaction analysis to improve the performance of CT-FFR. Also, a framework of multiple tasks physics-informed deep learning framework would be developed for more efficient calculation of CT-FFR. In this project, the structural force evaluation model of plaque burden had bridged the gap between plaque burden and hemodynamics, which could be a supportive technique to improve the functional assessment of artery stenosis.