Integrated prediction of cardiovascular events by automated coronary plaque and pericoronary adipose tissue quantification from CT Angiography

项目摘要 冠状动脉疾病仍然是全球死亡的主要原因， 患心肌梗塞（心脏病发作）没有先兆症状。患者研究 冠状动脉疾病传统上仅关注冠状动脉狭窄的严重程度 动脉粥样硬化斑块，而不是冠状动脉斑块的不良特征， 容易破裂并导致心肌梗死。冠状动脉CT血管造影（CTA）是一种无创性的 允许评估冠状动脉狭窄和斑块特征的测试。目前，CTA 目视解释狭窄。CTA狭窄严重程度和斑块特征的定量测量是 不是当前临床常规的一部分。 我们建议开发新的图像处理算法，用于完全自动化的，鲁棒的量化。 冠状动脉斑块特征。我们还建议自动量化脂肪的特征， 冠状动脉周围组织（冠状动脉周围脂肪组织，PCAT），已被证明 区分易破裂、高风险的冠状动脉斑块和稳定斑块。我们建议使用机器 学习有效地将联合收割机狭窄、斑块和PCAT特征与患者临床数据沿着组合的方法， 转化为新的综合风险评分，用于预测未来的不良心血管事件。我们将对此进行评估 真实世界、前瞻性、具有里程碑意义的SCOT-HEART试验中的风险评分（包括所有2073例患者， 试验的CTA组），在大型多中心患者登记研究中增加外部确认，并提供CTA 扫描、临床数据和心血管事件（致命性和非致命性心肌梗死和 总计7844例患者的心血管死亡）。我们提出三个具体目标： 1)完善、扩展和自动化测量整个冠状动脉的冠状动脉斑块和管腔 树，并在连续CTA中标准化斑块变化的测量; 2)评估自动定量斑块特征和PCAT特征对以下患者的预后价值： 在前瞻性SCOT-HEART试验和多中心CTA登记研究中预测未来MACE; 3)开发和评估一种新的自动化患者风险评分，并进行全面的外部验证 患者临床数据、CTA测量的定量斑块特征和PCAT特征，使用机器 学习-在前瞻性SCOT-HEART试验和多中心CTA中预测未来MACE事件 登记处。 拟议的工作将使自动化，多方面和可重复的分析斑块，狭窄和 来自CTA的PCAT，结合反映不良心血管事件可能性的客观风险评分。 这项工作将提供一个新颖的，个性化的，现实世界的范例，客观和准确地识别 常规CTA成像中存在未来心血管事件风险的个体患者。

PROJECT SUMMARY Coronary artery disease remains the leading cause of death worldwide, and more than half of the individuals suffering myocardial infarction (heart attacks) have no premonitory symptoms. Studies of patients with coronary artery disease have traditionally focused only on the severity of narrowing (stenosis) of the coronary arteries by atherosclerotic plaques, rather than the adverse features of coronary plaques which are predisposed to rupture and precipitate myocardial infarction. Coronary CT Angiography (CTA) is a noninvasive test that allows assessment of both coronary stenosis and plaque characteristics. Currently, however, CTA is interpreted visually for stenosis. Quantitative measurements of CTA stenosis severity and plaque features are not part of current clinical routine. We propose to develop novel image processing algorithms for fully automated, robust quantification of coronary plaque features from CTA. We also propose to automatically quantify the characteristics of adipose tissue around the coronary arteries (pericoronary adipose tissue, PCAT), which have been shown to differentiate rupture-prone, high-risk coronary plaques from stable ones. We propose to apply machine learning methods to efficiently combine stenosis, plaque and PCAT features, along with patient clinical data, into a new integrated risk score for the prediction of future adverse cardiovascular events. We will evaluate this risk score in the real-world, prospective, landmark SCOT-HEART trial (including all 2073 patients in the CTA arm of the trial), with added external validation in large multicenter patient registries, with available CTA scans, clinical data, and followup for cardiovascular events (fatal and non-fatal myocardial infarction and cardiovascular death in a grand total of 7844 patients). We propose three specific aims: 1) To refine, expand and automate measurements of coronary plaque and lumen for the entire coronary artery tree, and to standardize measurement of plaque changes in serial CTA; 2) To evaluate the prognostic value of automatically-quantified plaque features and PCAT characteristics for the prediction of future MACE in the prospective SCOT-HEART trial and multicenter CTA registries; 3) To develop and evaluate with full external validation a new automated patient risk score—combining patient clinical data, CTA-measured quantitative plaque features and PCAT characteristics, using machine learning—for the prediction of future MACE events in the prospective SCOT-HEART trial and multicenter CTA registries. The proposed work will enable automated, multi-faceted and reproducible analysis of plaque, stenosis and PCAT from CTA, combined with objective risk scores reflecting likelihood of adverse cardiovascular events. This work will provide a novel, personalized, real-world paradigm that objectively and accurately identifies individual patients at risk of future cardiovascular events, from routine CTA imaging.