Multi-scale coupling of the vascular hemodynamics for an AI-assisted, standardized evaluation of neurological pathologies

神经血管疾病可导致受影响个体的严重限制和残疾，也是德国的主要死亡原因之一。特别是脑血管的患者特异性变化，例如，以所谓的颅内动脉瘤（永久性，球囊样血管凸起）或动静脉畸形（动脉和静脉血管连接异常，没有毛细血管床）的形式表现出来。尽管不断发展的成像模式使可靠的诊断成为可能，但个体风险评估非常复杂，受到众多影响变量的影响，并且由于缺乏模型，在临床实践中过于简化。因此，最佳的治疗决策是具有挑战性的。在本研究项目的背景下，需要通过多尺度建模来实现对神经血管病理的整体评估。首先，用一维模型描述了心血管血流动力学。随后，利用计算流体动力学将神经血管循环和静脉系统绘制为3D。通过这种高度个性化和人工智能辅助的方法，上述病理可以在形态学和血流动力学上精确描述，从而沿着时间尺度计算跟踪其生长和重塑过程。在时间尺度上，采用随时间变化的四维流量测量和层析成像数据以及纵向研究。在成功实现了“从主动脉到静脉”的建模后，该项目旨在通过框架内的可用性模块对开发的芯片模型进行标准化。同时，将进行高分辨率体外验证测量，以确保模型的合理性。最后，为了准备在临床环境中的应用，计划将开发转移到评分系统中。标准化和评分系统将利用基于人工智能（AI）的方法。这包括基于图像和网格的流模拟预处理和评估（重点是深度学习）以及提取参数的分类（重点是机器学习）。总之，评估神经血管病变的计划整体方法使患者特异性血流动力学与医学成像、基于人工智能的图像处理和评估以及模拟描述的高度跨学科结合成为可能。因此，将这些影响变量和条件转移到一个标准化的评估系统中，可以对患者的实际疾病状态进行精确和无风险的评估。

Neurovascular diseases can lead to severe limitations and disabilities in affected individuals and are also among the leading causes of death in Germany. Particularly patient-specific changes in the cerebral vessels are expressed, for example, in the form of so-called intracranial aneurysms (permanent, balloon-like vessel bulges) or arteriovenous malformations (abnormal connections of arterial and venous vessels without capillary bed). Although continuously evolving imaging modalities enable a reliable diagnosis, individual risk assessment is highly complex, subject to numerous influencing variables, and too simplified in clinical practice due to the lack of models. As a result, the optimal treatment decision is challenging.In the context of this research project, a holistic approach to evaluate neurovascular pathologies shall be realized by means of multi-scale modeling. First, the cardiovascular hemodynamics are described by means of a one-dimensional model. Subsequently, the neurovascular circulation and the venous system are mapped in 3D using computational fluid dynamics. Through this highly individualized and AI-assisted approach, the aforementioned pathologies can be precisely described morphologically and hemodynamically to computationally track their growth and remodeling processes along the time scale. For the time scale, time-dependent 4D flow measurements and tomographic image data are employed as well as longitudinal studies. After the successful realization of the modeling "from the aorta to the vein", the project aims to standardize the developed in-silico models via a usability module within the framework. In parallel, high-resolution in vitro validation measurements will be performed to ensure the plausibility of the models. Finally, the transfer of the developments into a scoring system is planned in order to prepare an application in the clinical environment. The standardization as well as the scoring system will exploit methods based on artificial intelligence (AI). This comprises the image- and mesh-based preprocessing and evaluation of flow simulation (with focus on deep learning) as well as the classification of extracted parameters (with focus on machine learning).In summary, the planned holistic approach to assess neurovascular pathologies enables a highly interdisciplinary combination of patient-specific hemodynamics with medical imaging, AI-based image processing and evaluation, and simulative description. Consequently, transferring these influencing variables and conditions into a standardized assessment system can enable a precise and risk-free assessment of the actual disease state for the patient.