经导管主动脉瓣置换术后新发传导阻滞（NOCD）可导致近远期死亡率增加和不良中远期预后。我们前期研究发现术后NOCD与主动脉根部复合体（AVC）的钙化定量、分布及血流动力学参数等多因素相关。现阶段NOCD风险预测模型均基于欧美人群构建，鉴于中国患者有AVC钙化重、瓣叶结合钙化占比高等特征，欧美模型对中国患者适用性存疑，此外，模型提取参数较单一，效能有待提升。本项目拟首先完成人工特征与基于CNN的深度特征提取，而后采用特征层融合策略避免有效特征丢失，采用SVM+SFS获取CT、MRI最佳特征子集，与临床信息矢量融合后经SVM实现预测模型构建。其中瓣叶结合钙化分区方式的运用、基于瓣膜间质细胞钙化量的人工提取特征筛选、压力变化曲线图辅助MRI数据降维等方法，保障有中国患者特征性的多维度信息的重点纳入，实现有中国患者针对性的快速精准无创NOCD预测。

The occurrence of new onset conduction disturbance (NOCD) after transcatheter aortic valve replacement (TAVR) may lead to substantial increases in mortality and negative impacts on late outcomes. Quantification and distribution of the calcification in the aortic root complex has been proposed as one of the determining factors. Compared to the western patients, Chinese patients are characterized by the higher calcification load, more frequent calcification in the borderline area as well as some specific genetic features. Additionally, all the present prediction models are western population-based estimates only. No attempts have been made to identify the specific impact of commissural calcium, which is common among Chinese AS patients. Radiomics features are extracted semi-automatically and automatically. Feature selection is then based on the SVM+SFS. To obtain higher prediction accuracy, we use the commissure-based dividing method, VICs calcium load-aid feature selection and WSS-based MRI dataset acquisition. The support vector machine is used to train the prediction of NOCD after the merging of three subsets. This study aims at employing the multi-modality images based radiomics analysis of the aortic root complex among aortic stenosis patients for the feature selection and data subset determination. We would then construct a predictive model that integrates Chinese characterized clinical, genetic, and imaging information to predict post-TAVR NOCD fast and effectively.