The Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Melbourne, VIC, Australia.
The Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Melbourne, VIC, Australia.
Neuroimage. 2018 Nov 1;181:85-94. doi: 10.1016/j.neuroimage.2018.06.020. Epub 2018 Jun 15.
Correlation-based sliding window analysis (CSWA) is the most commonly used method to estimate time-resolved functional MRI (fMRI) connectivity. However, instantaneous phase synchrony analysis (IPSA) is gaining popularity mainly because it offers single time-point resolution of time-resolved fMRI connectivity. We aim to provide a systematic comparison between these two approaches, on temporal, topological and anatomical levels. For this purpose, we used resting-state fMRI data from two separate cohorts with different temporal resolutions (45 healthy subjects from Human Connectome Project fMRI data with repetition time of 0.72 s and 25 healthy subjects from a separate validation fMRI dataset with a repetition time of 3 s). For time-resolved functional connectivity analysis, we calculated tapered CSWA over a wide range of different window lengths that were compared to IPSA. We found a strong association in connectivity dynamics between IPSA and CSWA when considering the absolute values of CSWA. The association between CSWA and IPSA was stronger for a window length of ∼20 s (shorter than filtered fMRI wavelength) than ∼100 s (longer than filtered fMRI wavelength), irrespective of the sampling rate of the underlying fMRI data. Narrow-band filtering of fMRI data (0.03-0.07 Hz) yielded a stronger relationship between IPSA and CSWA than wider-band (0.01-0.1 Hz). On a topological level, time-averaged IPSA and CSWA nodes were non-linearly correlated for both short (∼20 s) and long (∼100 s) windows, mainly because nodes with strong negative correlations (CSWA) displayed high phase synchrony (IPSA). IPSA and CSWA were anatomically similar in the default mode network, sensory cortex, insula and cerebellum. Our results suggest that IPSA and CSWA provide comparable characterizations of time-resolved fMRI connectivity for appropriately chosen window lengths. Although IPSA requires narrow-band fMRI filtering, it does not mandate a (semi-)arbitrary choice of window length and window overlap. A code for calculating IPSA is provided.
基于相关的滑动窗口分析（CSWA）是最常用于估计时分辨功能磁共振成像（fMRI）连接的方法。然而，瞬时相位同步分析（IPSA）越来越受欢迎，主要是因为它提供了时分辨 fMRI 连接的单时间点分辨率。我们旨在从时间、拓扑和解剖学水平上对这两种方法进行系统比较。为此，我们使用了来自两个具有不同时间分辨率的独立队列的静息态 fMRI 数据（来自人类连接组计划 fMRI 数据的 45 名健康受试者，重复时间为 0.72s，以及来自另一个验证 fMRI 数据集的 25 名健康受试者，重复时间为 3s）。对于时分辨功能连接分析，我们计算了广泛的不同窗口长度的锥形 CSWA，并将其与 IPSA 进行了比较。当考虑 CSWA 的绝对值时，我们发现 IPSA 和 CSWA 之间的连接动力学存在很强的关联。无论基础 fMRI 数据的采样率如何，CSWA 与 IPSA 的关联在窗口长度约为 20s（短于滤波 fMRI 波长）时比约 100s（长于滤波 fMRI 波长）时更强。fMRI 数据的窄带滤波（0.03-0.07Hz）产生了 IPSA 和 CSWA 之间更强的关系，而宽带滤波（0.01-0.1Hz）则不然。在拓扑水平上，对于短（约 20s）和长（约 100s）窗口，时间平均 IPSA 和 CSWA 节点是非线性相关的，主要是因为具有强负相关（CSWA）的节点显示出高相位同步（IPSA）。在默认模式网络、感觉皮层、脑岛和小脑中，IPSA 和 CSWA 在解剖学上相似。我们的结果表明，对于适当选择的窗口长度，IPSA 和 CSWA 提供了可比的时分辨 fMRI 连接特征。尽管 IPSA 需要窄带 fMRI 滤波，但它不需要（半）任意选择窗口长度和窗口重叠。提供了计算 IPSA 的代码。
