REDUCED ORDER MODELLING AND BIOCHEMICAL PATIENT-SPECIFICITY IN A COMPUTATIONAL MODEL OF CEREBRAL ANEURYSM THROMBOSIS: TOWARDS CLINICAL APPLICABILITY

The files contain data on thrombin and fibrin concentration, velocity and clot size for haemophilliac, healthy and thrombotic patients. 
Computational fluid dynamics (CFD) models of cerebral aneurysm thrombosis are patient-specific insofar as geometry and haemodynamics are concerned. The biochemical reactions that result in clotting require considerable resources to fit all parameters on a per patient or population basis. Furthermore, translation of these CFD models to clinical contexts is limited by model complexity and computational cost. In this study, we present a model that couples results from a calibrated automated thrombogram (CAT), an in vitro test that is used to determine clotting function on a per patient basis, with CFD. The CAT data was fitted into population-specific biochemical profiles of haemophiliac, healthy and thrombotic patients, and applied to the aneurysmal wall of a 2D idealized geometry. There was faster clot growth in the thrombotic case, followed by the normal and haemophiliac cases, respectively. Complex vorticial structures formed as the different clots evolved. The patterns in clot growth, distribution of thrombin and fibrin, and velocity profile showed that there is a strong link between haemodynamics and biochemistry. The model was verified by comparing it to experimental results, and it successfully captured the qualitative features of in vitro clotting.
Polynomial and logistic regression machine learning algorithms were used to develop a reduced order model from CFD results. This model is relatively simple but would have far greater utility in a clinical context as it does not require solution of numerical methods or specialized CFD training.

该数据集包含血凝酶和纤维蛋白浓度、速度以及凝血块大小等数据，针对血友病、健康以及血栓形成患者。就几何形状和血流动力学而言，脑动脉瘤血栓形成的计算流体动力学（CFD）模型具有个体特异性。导致凝血的生化反应需要大量资源以适应每个患者或人群的参数。此外，由于模型复杂性和计算成本的限制，将这些CFD模型转化为临床情境的应用受到限制。在本研究中，我们提出了一种模型，该模型将校准自动血栓图（CAT）的结果与CFD相结合。CAT数据被拟合到血友病、健康和血栓形成患者的特定生化图谱中，并应用于二维理想化几何形状的动脉瘤壁上。在血栓形成病例中，凝血块生长速度较快，随后依次为正常和血友病病例。随着不同凝血块的发展，形成了复杂的涡流结构。凝血块生长模式、凝血酶和纤维蛋白的分布以及速度剖面显示，血流动力学与生化反应之间存在紧密的联系。该模型通过与其实验结果进行比较得到验证，并成功捕捉了体外凝血的定性特征。利用多项式和对数逻辑回归机器学习算法，从CFD结果中开发了一种降阶模型。该模型相对简单，但在临床情境中具有更大的实用性，因为它不需要解决数值方法或专门的CFD培训。