AI 中文总结
本综述系统梳理100余项研究，指出LVAD控制临床转化受限源于建模假设与临床现实不匹配，并提出自适应数据驱动方法结合植入传感限制是未来方向。
AI 中文摘要
心室辅助装置（VADs），特别是旋转式左心室辅助装置（LVADs），对于无法接受移植的晚期心力衰竭患者至关重要。尽管心血管建模和控制领域取得了进展，但所提出的LVAD控制方法的临床转化仍然有限。本综述考察了LVAD建模和控制的演变，并认为这一差距并非源于算法不够复杂，而是源于建模假设、传感限制和现实世界心血管变异性之间的不匹配。对100多篇同行评审研究的系统化综述涵盖了数学模型心血管模型、集总参数和降阶表示、经典和先进控制策略，以及新兴的数据驱动和机器学习方法。文献通过一个以问题为导向的框架进行综合，该框架将建模和控制选择与临床挑战联系起来，包括生理可观测性、参数可辨识性、患者变异性，以及预防心室抽吸和血栓形成等不良事件。分析表明，高保真模型和智能控制器在模拟中表现良好，但它们对不可测量状态、大量参数调整和密集传感的依赖限制了临床实施。更简单的控制方法在临床约束下通常提供更强的鲁棒性。自适应和数据驱动技术可能有助于弥合这一差距，但前提是它们要考虑植入式传感限制和可解释性要求。通过识别采用的结构性障碍，本综述综合了LVAD建模和控制范式，并概述了生理自适应、临床可实施且患者安全的LVAD系统的研究方向。
英文摘要
Ventricular Assist Devices (VADs), particularly rotary Left Ventricular Assist Devices (LVADs), are essential for patients with advanced heart failure who are ineligible for transplantation. Despite advances in cardiovascular modeling and control, clinical translation of proposed LVAD control methods remains limited. This review examines the evolution of LVAD modeling and control and argues that the gap results not from insufficiently sophisticated algorithms, but from mismatches between modeling assumptions, sensing limitations, and real-world cardiovascular variability. A systematized review of more than 100 peer-reviewed studies covers mathematical cardiovascular models, lumped-parameter and reduced-order representations, classical and advanced control strategies, and emerging data-driven and machine learning approaches. The literature is synthesized using a problem-driven framework linking modeling and control choices to clinical challenges, including physiological observability, parameter identifiability, patient variability, and prevention of adverse events such as ventricular suction and thrombosis. The analysis shows that high-fidelity models and intelligent controllers can perform well in simulation, but their dependence on unmeasurable states, extensive parameter tuning, and dense sensing limits clinical implementation. Simpler control approaches often provide greater robustness under clinical constraints. Adaptive and data-driven techniques may help bridge this gap, but only if they account for implantable sensing limitations and interpretability requirements. By identifying structural barriers to adoption, this review synthesizes LVAD modeling and control paradigms and outlines research directions for physiologically adaptive, clinically implementable, and patient-safe LVAD systems.
