Phase Synchronization Analysis for Reconstructing Physiologic Networks

人体器官可以看作是神经调节下的复杂系统。通过记录长的多变量时间序列来监测它们的行为。器官之间的各种耦合和反馈相互作用增加了这些信号的复杂性。非平稳、间歇和非线性的波动和振荡发生，这需要统计物理学的工具来进行全面的描述。时间序列分析方法面临的主要挑战是对动力学的描述和相互作用的量化，以改进生理模型和识别诊断相关参数。心脏和呼吸系统之间的相关相互作用通过相位同步来描述。非线性动力学的进步导致了几种相位同步分析方法的出现。由于没有进行系统的比较，我们将使用1400多名受试者的数据来比较这些方法，并系统地研究噪声和离群值的有害影响。目标是找出每种方法的优点和缺点。还将考虑与心血管低频振荡的相位同步。此外，我们还希望开发和建立新的自同步和延时同步方法来表征时间序列中的振荡，并量化两个时间序列之间的相互关系。该项目的目标是基于同步度量确定可用于预测心肌梗死后死亡率的参数以及帕金森氏病、阿尔茨海默病和抑郁症的早期诊断工具。在项目的第二部分，我们将研究成熟的自相关标度行为、最近引入的交叉调制特性和时间序列的新的自同步特性之间的关系。我们的目标是通过研究脑电波幅度、身体运动、心跳和呼吸的比例变化来澄清经常观察到但很难解释的比例定律的原因。除了这个一般的方法，我们还将重点介绍压力反射生理学，作为一个有限但仍然复杂的系统中与生理相关的相互作用的例子。主要目标是开发和应用新的方法来可靠地量化压力反射敏感性，并将压力反射相互作用与其他生理相互作用区分开来。该项目的第三部分涉及多器官相互作用网络的重建、表征和应用。基于我们最新的工作，我们将从多变量数据中的相互关联的波动和振荡中确定生理网络。我们计划监测和描述这些网络在不同状态和疾病期间的动态和演变。具体目标是确定睡眠阶段等生理状态之间的转换指标，并开发上述疾病的早期诊断工具。

Organs in the human body can be regarded as complex systems under neural regulation. Their behavior is monitored by recording long multivariate time series. The complexity in these signals is increased by various couplings and feedback interactions between the organs. Non-stationary, intermittent, and non-linear fluctuations and oscillations occur, which require tools from statistical physics for a full description. The characterization of the dynamics and quantification of the interactions to improve physiological models and to identify diagnostically relevant parameters are major challenges to the methods of time series analysis.A relevant interaction between the cardiac and respiratory systems is described by phase synchronization. Advances in nonlinear dynamics have led to several phase-synchronization analysis methods. Since no systematic comparison was done, we will compare these methods using data from more than 1400 subjects and study systematically detrimental effects of noise and outliers. The goal is to identify advantages and disadvantages for each method. Phase synchronization with cardiovascular low-frequency oscillations will also be considered. In addition, we want to develop and establish novel auto-synchronization and time-delayed synchronization approaches to characterize oscillations in time series and to quantify the inter-relations between two time series. The goal is to identify parameters based on synchronization measures that can be used as predictors of mortality after myocardial infarction and early diagnostic tools for Parkinson's disease, Alzheimer's disease, and depression.In the second part of the project we will study relations between well-established auto-correlation scaling behavior, recently introduced cross-modulation properties, and novel auto-synchronization properties of time series. The goal is to clarify the causes of frequently observed, but hardly explained scaling laws by investigating transitions in the scaling of brain-wave amplitudes, body motion, heartbeat, and respiration. Besides this general approach, we will focus on baroreflex physiology as an example of a physiologically relevant interaction in a limited but still complex system. The main goals are developing and applying novel methods for a reliable quantification of baroreflex sensitivity and distinguishing baroreflex interactions from other physiologic interactions.The third part of the project deals with the reconstruction, characterization, and application of multi-organ interaction networks. Based on our very recent works, we will determine physiologic networks from cross-correlated fluctuations and oscillations in multivariate data. We plan to monitor and characterize the dynamics and evolution of these networks during different states and diseases. Particular goals are identifying indicators for transitions between physiologic states like sleep stages and developing early diagnostic tools for the diseases mentioned above.