多模态影像学定量评价缺血性心脏病微循环障碍的深度学习技术研究

冠状动脉微循环障碍（CMCD）是导致患者心肌缺血进而心衰的重要原因，对此病的早期精确诊断具有重大意义。目前，虽有多种影像学方法用于CMCD的诊断，然而单一模态所获得的CMCD影像信息有限，而多种模态信息间又缺乏有效的融合，严重影响了影像学手段诊断该病的准确性。但是，深度学习技术的出现，为实现多模态影像学信息的高效融合提供了技术支持。为此，本项目欲利用深度学习的优势，对临床CMCD患者的多模态影像信息进行深度挖掘、多层次分析整合以及训练，构建CMCD多模态影像人工智能诊断体系，实现对CMCD患者精确诊断；在此基础上，进一步深入探索CMCD的病理机制及其功能储备，通过组学方法筛选鉴定能用于早期识别CMCD的影像标志物，设计构建多模态分子探针，联合深度学习与分子影像学方法的优势，建立分子水平的人工智能诊断体系，为实现CMCD患者的早期诊断、治疗干预及预后判断提供新的技术手段。

Coronary microcirculatory dysfunction (CMCD) is one of the most crucial causes of myocardial ischemia and heart failure, and early diagnosis of this disease is of great significance. Currently, various imaging technologies have been applied in the diagnosis of CMCD. However, information obtained by single imaging modality is limited, and lack of effective fusion of multi-modalities was still a problem to be solved. These two limitations hampering the comprehensive evaluation of CMCD. Notably, the rapid development of deep learning provide a possibility to manage an efficient integration of multi-modalities information. We aim to develop an artificial intelligence diagnosis system based on deep learning to achieve accurate diagnosis of patients with CMCD. Furthermore, we will also explore the pathological mechanism of CMCD and screen some image markers for early recognition of CMCD. Then by fabricating multi-modal molecular probes according to these screened markers to monitor the abnormal information during CMCD at the molecular level, and in combination with deep learning, we aim to construct an artificial intelligence diagnosis system at molecular level and provide a new tool for early diagnosis, therapeutic evaluation and prognosis prediction of CMCD.