Quantification of Phase-Amplitude Coupling in Neuronal Oscillations: Comparison of Phase-Locking Value, Mean Vector Length, Modulation Index, and Generalized-Linear-Modeling-Cross-Frequency-Coupling
Quantification of Phase-Amplitude Coupling in Neuronal Oscillations: Comparison of Phase-Locking Value, Mean Vector Length, Modulation Index, and Generalized-Linear-Modeling-Cross-Frequency-Coupling
复制标题
DOI：
10.3389/fnins.2019.00573
复制
发表时间：
2019-06-07
影响因子：
4.3
通讯作者：
Rasch, Bjorn
中科院分区：
文献类型：
--
作者：
Huelsemann, Mareike J.;Naumann, Ewald;Rasch, Bjorn
相位-振幅耦合是脑电和脑磁图研究认知过程的一个很有前途的结构。由于概念的新奇，在文献中使用各种措施来计算相幅耦合。在这里，性能的三个最广泛使用的相位-幅度耦合措施-锁相值（PLV），平均矢量长度（MVL），调制指数（MI）-和广义线性模型交叉频率耦合（GLM-CFC）方法的帮助下，仿真数据进行了彻底的比较。我们结合联合收割机的优势，以前的评论，并使用一个现实的数据模拟，检查主持人和提供推论统计比较的所有四个指标的相位-振幅耦合。我们的分析表明，所有四个指标成功区分耦合强度和耦合宽度时，单相耦合。虽然MVL是最敏感的调制耦合强度和宽度，只有MI和GLM-CFC可以检测到双相耦合。所有四个指标的耦合值的影响主持人，包括数据长度，信噪比，采样率接近奈奎斯特频率。MI是最强大的对这些主持人的混杂影响。根据我们的分析，我们建议MI的噪声和短的数据时期与未知形式的耦合。对于具有双相耦合和高信噪比的高质量和长数据历元，建议使用MVL。理想情况下，两个指数同时报告一个数据集。
Phase-amplitude coupling is a promising construct to study cognitive processes in electroencephalography (EEG) and magnetencephalography (MEG). Due to the novelty of the concept, various measures are used in the literature to calculate phase-amplitude coupling. Here, performance of the three most widely used phase-amplitude coupling measures - phase-locking value (PLV), mean vector length (MVL), and modulation index (MI) - and of the generalized linear modeling cross-frequency coupling (GLM-CFC) method is thoroughly compared with the help of simulated data. We combine advantages of previous reviews and use a realistic data simulation, examine moderators and provide inferential statistics for the comparison of all four indices of phase-amplitude coupling. Our analyses show that all four indices successfully differentiate coupling strength and coupling width when monophasic coupling is present. While the MVL was most sensitive to modulations in coupling strengths and width, only the MI and GLM-CFC can detect biphasic coupling. Coupling values of all four indices were influenced by moderators including data length, signal-to-noise-ratio, and sampling rate when approaching Nyquist frequencies. The MI was most robust against confounding influences of these moderators. Based on our analyses, we recommend the MI for noisy and short data epochs with unknown forms of coupling. For high quality and long data epochs with monophasic coupling and a high signal-to-noise ratio, the use of the MVL is recommended. Ideally, both indices are reported simultaneously for one data set.