Enabling reliable cardiovascular simulations via uncertainty quantification

 描述（由申请人提供）：心血管血流动力学和生理学的高级模拟现在被纳入临床决策、手术计划和FDA批准过程。模拟有可能影响病人改变生活的决定。因此，这些进步伴随着患者和治疗他们的临床医生越来越多的责任，以证明模拟产生可靠和安全的结果。模拟社区继续推动患者特定多尺度模型的常规临床使用，而不提供一个 意味着统计量化其预测的可靠性。发展转型 目前缺乏的评估不确定性的技术将减轻患者的风险，并最终实现个性化医疗模拟的安全和常规使用。患者特定的心血管（CV）模拟需要结合不确定的假设和来自临床和成像数据的输入。这个问题目前被掩盖了，要求最终用户接受确定性仿真预测作为“真理”，没有相关的置信区间。这导致了临床社区对模拟可信度的合理怀疑，并且是临床使用和最终FDA批准的障碍。我们建议通过创建一套高效的自动化不确定性量化（UQ）工具来评估和提高患者特定模拟预测的可靠性，以满足这一未满足的需求。我们将通过应用于冠状动脉疾病（CAD）的多尺度模拟来建立我们的UQ框架。冠状动脉建模是UQ方法的理想测试平台和挑战，具有来自图像分割、材料特性和复杂生理学的多参数不确定性。为了实现我们的目标，我们提出了三个具体目标：1）一个综合的多模态成像研究，将增加模型的保真度，并使不确定性评估，2）创建自动参数估计工具的同化临床数据到心血管模拟，和3）开发一个有效的计算框架，以量化CAD模拟的不确定性。拟议的工作意义重大，因为我们将（1）提高CV模拟社区报告输出统计数据的标准，（2）建立临床护理和其他研究人员采用模拟的标准，以及（3）通过开源SimVascular项目提供一套新的工具。这是创新的，因为（1）UQ是执行建立模拟输出的置信区间和（2）无数的不确定性通常未讨论的CV模拟社区严格量化。我们的多学科团队由具有患者特异性建模、UQ数学方法、高性能计算和医学成像专业知识的研究人员组成。我们在联合出版、临床翻译和资助合作方面有着良好的记录。我们的翻译目标是为心血管模拟社区提供有效的UQ工具，提高模拟可靠性的标准，并最终提高临床采用率。

 DESCRIPTION (provided by applicant): Advanced simulations of cardiovascular hemodynamics and physiology are now being incorporated into clinical decision-making, surgical planning, and the FDA approval process. Simulations have potential to influence life- altering decisions for patients. As a result, these advancements come with an ever-increasing responsibility to the patients and the clinicians who treat them to prove that simulations produce reliable and safe results. It is dangerous and irresponsible for the simulation community to continue to push for routine clinical use of patient-specific multiscale models without providing a means to statistically quantify the reliability of their predictions. Development of transformative technology to assess uncertainty, which is currently lacking, will mitigate patient risk and ultimately enable safe and routine use of simulations for personalized medicine. Patient specific cardiovascular (CV) simulations require a combination of uncertain assumptions and inputs from clinical and imaging data. This issue currently gets swept under the rug, asking end-users to accept deterministic simulation predictions as "truth" with no associated confidence intervals. This leads to justified skepticism in the clinical community regarding the trustworthiness of simulations, and is a roadblock to clinical use and eventual FDA approval. We propose to address this unmet need by creating a suite of efficient and automated uncertainty quantification (UQ) tools to assess and improve the reliability of patient-specific simulation predictions. We wil establish our UQ framework through application to multiscale simulations of coronary artery disease (CAD). Coronary modeling is an ideal test-bed and challenge for UQ methodologies, with multi-parameter uncertainty arising from image segmentation, material properties, and complex physiology. To accomplish our objectives, we propose three specific aims: 1) An integrative multi- modality imaging study that will increase model fidelity and enable uncertainty assessment, 2) Creation of automated parameter-estimation tools for assimilation of clinical data into cardiovascular simulations, and 3) Development of an efficient computational framework to quantify uncertainties in simulations of CAD. The proposed work is significant because we will (1) raise the bar for the CV simulation community to report output statistics, (2) establish standards for adoption of simulations in clinical care and by other researchers, and (3) provide a novel suite of tools through the open-source SimVascular project. It is innovative because (1) UQ is performed to establish confidence intervals on simulation outputs and (2) the myriad uncertainties typically un- discussed in the CV simulation community are rigorously quantified. Our multi-disciplinary team consists of investigators with expertise in patient-specifi modeling, mathematical methods for UQ, high-performance computing, and medical imaging. We have a strong track record of joint publication, clinical translation, and funded collaborations Our translational goal is to provide the cardiovascular simulation community with efficient tools for UQ, raising the bar for simulation reliability and ultimately increasing clinical adoption.