基于心脏磁共振成像的卷积神经网络深度学习模型在肥厚型心肌病辅助诊断及危险因素分层的方法研究

肥厚型心肌病（HCM）是具有高度异质性的遗传性心肌病，HCM的准确诊断和危险分层有助于患者的精准化管理。多参数心脏磁共振成像（CMR）是HCM临床诊断和预后随访的重要影像手段，但CMR图像后处理繁琐耗时，海量数据与人工处理导致不同程度主观性偏差，限制了其在HCM临床的应用和推广。卷积神经网络（CNN）作为深度学习的代表性算法，有望实现标准化的CMR数据处理，自动化提取大量高维数据中的关键特征，智能化深度发掘影像和临床中的关键信息。项目组拥有CMR大数据库，前期采用DeepLab V3+深度学习模型初步实现对CMR电影图像的心室轮廓识别和分割。本研究拟进一步构建多分类CNN算法模型，研发智能化CMR处理技术，实现HCM辅助智能诊断；并结合多变量临床指标，建立HCM不良预后危险因素分层的综合辅助影像决策系统，为HCM的临床诊疗决策提供新策略和科学依据，也为普及CMR的应用提供帮助。

Hypertrophic cardiomyopathy (HCM) is a highly heterogeneous inherited cardiomyopathy, accurate diagnosis and risk stratification of HCM are helpful for accurate management of patients. Multi-parameter cardiac magnetic resonance imaging (CMR) is an important imaging method for clinical diagnosis and prognosis follow-up of HCM, but CMR image post-processing is complex and time-consuming, massive data and manual processing lead to subjective deviation in certain degrees, which limits its clinical application and promotion in HCM. As a representative algorithm of deep learning, convolutional neural network (CNN) is expected to achieve standardized CMR data processing, automatic extraction of key features from a large number of high-dimensional data, intelligent deep exploration of key information in image and clinical. The project team has a large CMR database, and used Deeplab V3 + deep learning model to preliminarily achieve the recognition and segmentation of CMR cine images. In this study, we intend to further build a multi-classification CNN algorithm model, develop intelligent CMR processing technology, and achieve the intelligent assistant diagnosis of HCM; combined with multivariate clinical factors, to establish a comprehensive image decision-making system of risk factors stratification for HCM, which provides new strategies and scientific basis for clinical diagnosis and therapeutic decision-making of HCM, and helps to promote the application of CMR.