Results of the Interlaboratory Computational Fluid Dynamics Study of the FDA Benchmark Blood Pump
Results of the Interlaboratory Computational Fluid Dynamics Study of the FDA Benchmark Blood Pump
复制标题
DOI：
10.1007/s10439-022-03105-w
复制
发表时间：
2022-11-18
影响因子：
3.8
通讯作者：
Craven, Brent A.
中科院分区：
文献类型：
--
作者：
Ponnaluri, Sailahari, V;Hariharan, Prasanna;Craven, Brent A.
计算流体动力学（CFD）被广泛用于模拟血液接触医疗器械。然而，要依靠模型来为高风险决策提供信息，模型的可信度应通过验证来证明。为了提供强大的数据集进行验证，FDA的研究人员和合作者开发了两种基准医疗设备流量模型：喷嘴和离心血泵。使用每个模型获得流场和溶血的实验测量值。同时，进行了独立的开放实验室间CFD研究，来自世界各地的参与者对测量结果不知情，提交了每个基准模型的CFD预测。在本研究中，我们报告了FDA基准血泵的实验室间CFD研究结果。我们分析了24个CFD提交的结果，使用各种不同的流动求解器，方法和建模参数。为了评估CFD预测的准确性，我们将结果与不同泵运行条件下的三个感兴趣的量（压头、速度场和溶血）的实验测量结果进行比较。我们还调查了参与者使用的不同CFD方法和建模选择的影响。我们的分析表明，虽然许多CFD提交准确预测了个别情况下的泵性能，但没有一个参与者能够准确预测所有条件下的所有感兴趣的量。几位参与者准确地预测了所有条件下的压头和除一两种情况外的所有情况下的速度场。提交溶血结果的八名参与者中只有一名在大多数条件下准确预测了绝对血浆游离血红蛋白水平，尽管大多数参与者成功预测了条件之间的相对溶血水平。总体而言，本研究强调需要在整个操作条件范围内以及针对所有感兴趣的量来验证旋转血泵的CFD建模，因为某些操作条件和区域（例如，泵扩散器）比其它的更难以准确预测。所有感兴趣的量都应进行验证，因为如图所示，尽管对流场的预测相对不准确，但仍有可能准确预测溶血。
Computational fluid dynamics (CFD) is widely used to simulate blood-contacting medical devices. To be relied upon to inform high-risk decision making, however, model credibility should be demonstrated through validation. To provide robust data sets for validation, researchers at the FDA and collaborators developed two benchmark medical device flow models: a nozzle and a centrifugal blood pump. Experimental measurements of the flow fields and hemolysis were acquired using each model. Concurrently, separate open interlaboratory CFD studies were performed in which participants from around the world, who were blinded to the measurements, submitted CFD predictions of each benchmark model. In this study, we report the results of the interlaboratory CFD study of the FDA benchmark blood pump. We analyze the results of 24 CFD submissions using a wide range of different flow solvers, methods, and modeling parameters. To assess the accuracy of the CFD predictions, we compare the results with experimental measurements of three quantities of interest (pressure head, velocity field, and hemolysis) at different pump operating conditions. We also investigate the influence of different CFD methods and modeling choices used by the participants. Our analyses reveal that, while a number of CFD submissions accurately predicted the pump performance for individual cases, no single participant was able to accurately predict all quantities of interest across all conditions. Several participants accurately predicted the pressure head at all conditions and the velocity field in all but one or two cases. Only one of the eight participants who submitted hemolysis results accurately predicted absolute plasma free hemoglobin levels at a majority of the conditions, though most participants were successful at predicting relative hemolysis levels between conditions. Overall, this study highlights the need to validate CFD modeling of rotary blood pumps across the entire range of operating conditions and for all quantities of interest, as some operating conditions and regions (e.g., the pump diffuser) are more challenging to accurately predict than others. All quantities of interest should be validated because, as shown here, it is possible to accurately predict hemolysis despite having relatively inaccurate predictions of the flow field.