基于多组学预测Stanford B型主动脉夹层腔内治疗术后远端重构的机器学习模型构建及验证

远端重构不良是制约Stanford B型主动脉夹层腔内治疗远期疗效的瓶颈之一。术后早期识别重构不良的高危患者，有助于制定个体化的随访方案，尽早干预。申请人前期发现联合电子病史和影像学构建的机器学习模型可以用来预测远端重构，但准确度尚不够。同时发现基于术中DSA后处理获取的血流动力学参数与主动脉腔内修复术后的结局密切相关。因此，申请人结合文献证据，提出主动脉形态学，管壁的生物学特性以及血流动力学三者相互作用，共同影响夹层术后远端重构的科学假说。本课题拟利用患者的术前、术后CTA影像，外周血清和术中DSA造影，分别从影像组学、蛋白质组学和血流动力学，多组学角度探讨夹层腔内治疗术后远端重构不良的潜在预测因素和相关机制。再者，联合多组学信息用机器学习的方法构建风险预测模型，并在前瞻性队列中进行验证。本课题如获成功，将有助于深入理解夹层远端重构的机理，改善患者的远期疗效，具有重要的临床及社会价值。

Negative aortic remodeling has adverse impacts on the long-term outcome of endovascular treatment for Stanford type B aortic dissection. Early prediction of patients with high risk of negative aortic remodeling helps set an individual follow-up plan and benefits for early intervention. Previously, our team found a machine learning model based on the electric history records and radiological features could be used to predict the aortic remodeling. However, the accuracy is not high enough. Additionally, we also found hymodynamic parameters obtained from the postprocessing of intraoperative DSA images were associated with the outcomes of endovascular treatment for aortic diseases. Thus, considering the evidence from several literatures, we hypothesize that the morphology and biological activities of aorta and hymodynamics interact with each other, which all have great effects on the distal aortic remodeling after endovascular treatment. This project aims to determine the potential predictors of negative aortic remodeling and related mechanisms using the pre-and post-operative CTA images, serum and intraoperative DSA angiography from the view of radiomics, proteomics and hymodynamics, respectively. Furthermore, a risk predictive model combining multiomic information is constructed through machine learning, and validated in a prospective cohort. This project would help further understand the mechanism of distal aortic remodeling in aortic dissection, improve the long-term outcome of patients, which have great clinical and social value.