Prospective sudden cardiac death risk stratification using CMR and echocardiography machine learning in mitral valve prolapse

使用 CMR 和超声心动图机器学习对二尖瓣脱垂进行前瞻性心脏性猝死风险分层

• 批准号: 10600113
• 负责人: Francesca N Delling
• 金额: $ 77.5万
• 依托单位: UNIVERSITY OF CALIFORNIA, SAN FRANCISCO
• 依托单位国家: 美国
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-05-26 至 2025-04-30
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10600113
• 关键词: Affect; Arrhythmia; Artificial Intelligence; Autopsy; California; Cardiac; Clinical; Clinical Markers; Complex; Data; Databases; Defibrillators; Development; Diffuse; Echocardiography; Fibrosis; Future; Gadolinium; Goals; Heart Arrest; High Prevalence; Holter Electrocardiography; Hybrids; Image; Image Enhancement; Implantable Defibrillators; Individual; Left; Link; Machine Learning; Magnetic Resonance; Maps; Mission; Mitral Valve Insufficiency; Mitral Valve Prolapse; Myocardial; Outcome; Patients; Phenotype; Population; Prevalence; Primary Prevention; Registries; Retrospective Studies; Risk; Role; Sampling; San Francisco; Screening procedure; Survivors; Testing; Training; Uncertainty; United States National Institutes of Health; Universities; Validation; Ventricular; Ventricular Tachycardia; cardiovascular risk factor; convolutional neural network; coronary fibrosis; cost; extracellular; hemodynamics; high risk; improved; machine learning algorithm; mortality; neural network architecture; novel; predictive tools; prognostic significance; prospective; recurrent neural network; risk prediction; risk prediction model; risk stratification; secondary analysis; stem; sudden cardiac death; tool

PROJECT SUMMARY Mitral valve prolapse (MVP) is a common valvulopathy affecting over 170 million worldwide. Every year, 0.4- 1.9% of individuals with MVP will develop sudden cardiac arrest (SCA) or sudden cardiac death (SCD), and 7% of SCDs in the young are caused by MVP. However, predictors of this devastating outcome are not readily available, and indications for a primary prevention implantable cardioverter defibrillator (ICD) in MVP are lacking. Severe mitral regurgitation explains only 50% of SCA cases in MVP. SCD/SCA risk has also been linked to a bileaflet phenotype with mild MR, mitral annular disjunction (MAD), and left ventricular focal fibrosis on cardiac magnetic resonance (CMR)-late gadolinium enhancement (LGE) images. Such imaging parameters (including LGE) have not been evaluated prospectively. Moreover, they are not consistently found in SCA survivors, and diffuse fibrosis has been proposed as an alternative arrhythmic substrate by our group and others based on CMR/T1 mapping, strain echocardiography, and post-mortem data. Overall, it is challenging to pinpoint a unique imaging phenotype, and uncertainty exists about which MVP patients should undergo CMR. Regardless of arrhythmic phenotype, complex ventricular ectopy (ComVE - defined as frequent polymorphic PVCs, bigeminy or non-sustained ventricular tachycardia) is detected in 80-100% of MVP cases prior to SCA or SCD. ComVE, commonly associated with left ventricular fibrosis on CMR, is linked to higher all-cause mortality and SCA rates (20% versus 12% if no ComVE, p < 0.05) based on preliminary cross-sectional data. Our central hypothesis is that MVP patients with ComVE, because of the higher prevalence of either LGE or abnormal T1 mapping, represent ideal CMR candidates regardless of leaflet involvement or MAD, and can be rapidly identified by an automated “surveillance” tool within a large echocardiographic database. Moreover, we hypothesize that fibrosis is the strongest predictor of SCD/SCA in an unprecedented, multi-center effort to longitudinally assess clinical and CMR parameters of arrhythmic risk in MVP. Specifically, we aim to 1) Assess the role of CMR as a screening tool for fibrosis in MVP with ComVE incorporating T1 mapping in addition to LGE in an unselected MVP sample; 2) Develop an echo-based machine-learning algorithm to detect MVP with ComVE, test its association with myocardial fibrosis on CMR and longitudinal SCD/SCA risk; and 3) Build a novel prospective SCD/SCA risk prediction model in MVP. Better selection of CMR candidates and development of a SCD/SCA risk prediction tool inclusive of fibrosis by CMR are expected to dramatically improve risk stratification in MVP and establish future criteria for primary prevention ICD trials.

项目摘要 二尖瓣脱垂（MVP）是一种常见的瓣膜病，影响全球超过1.7亿人。每年，0.4- 1.9%的MVP患者会发生心脏骤停（SCA）或心脏性猝死（SCD），7%的MVP患者会发生心脏骤停（SCA）或心脏性猝死（SCD）。 年轻人的SCD由MVP引起。然而，这种破坏性结果的预测并不容易 可用，缺乏MVP一级预防植入式心律转复除颤器（ICD）的适应症。 重度二尖瓣返流仅解释MVP中50%的SCA病例。SCD/SCA风险也与 心脏瓣膜病患者的双叶表型伴轻度MR、二尖瓣环分离（MAD）和左心室局灶性纤维化 磁共振（CMR）-晚期钆增强（LGE）图像。这种成像参数（包括 LGE）尚未进行前瞻性评价。此外，它们并不总是在SCA幸存者中发现， 我们的研究小组和其他人基于以下观点提出弥漫性纤维化是一种替代性的代谢底物： CMR/T1标测、应变超声心动图和尸检数据。总的来说，确定一个独特的 成像表型，并且不确定哪些MVP患者应该接受CMR。无论 心律失常表型，复杂性心室异位（ComVE -定义为频繁的多态性PVC，二联律 或非持续性室性心动过速）在80-100%的MVP病例中检测到。ComVE， 通常与CMR的左心室纤维化有关，与较高的全因死亡率和SCA率有关 (20%，如果没有ComVE，则为12%，p < 0.05）。我们的核心假设是 ComVE的MVP患者，由于LGE或异常T1标测的患病率较高， 代表理想的CMR候选者，无论是否涉及瓣叶或MAD， 一个大型超声心动图数据库中的自动“监视”工具。此外，我们假设纤维化 是SCD/SCA的最强预测因子，这是一项前所未有的多中心纵向评估临床 MVP的CMR参数。具体而言，我们的目标是1）评估CMR作为筛查的作用 在MVP中使用ComVE的纤维化工具，除了LGE之外，ComVE还在MVP样本中合并T1标测; 2)开发一种基于回声的机器学习算法来检测ComVE的MVP，测试其与 心肌纤维化对CMR和纵向SCD/SCA风险的影响;以及3）建立新的前瞻性SCD/SCA风险 MVP中的预测模型。更好地选择CMR候选项目并制定SCD/SCA风险预测 包括CMR纤维化的工具预计将显著改善MVP的风险分层， ICD一级预防试验的未来标准。