Test-Retest Reliability of Synchrony and Metastability in Resting State fMRI

Abstract

In recent years, interest has been growing in dynamic characteristic of brain signals from resting-state functional magnetic resonance imaging (rs-fMRI). Synchrony and metastability, as neurodynamic indexes, are considered as one of methods for analyzing dynamic characteristics. Although much research has studied the analysis of neurodynamic indices, few have investigated its reliability. In this paper, the datasets from the Human Connectome Project have been used to explore the test-retest reliabilities of synchrony and metastability from multiple angles through intra-class correlation (ICC). The results showed that both of these indexes had fair test-retest reliability, but they are strongly affected by the field strength, the spatial resolution, and scanning interval, less affected by the temporal resolution. Denoising processing can help improve their ICC values. In addition, the reliability of neurodynamic indexes was affected by the node definition strategy, but these effects were not apparent. In particular, by comparing the test-retest reliability of different resting-state networks, we found that synchrony of different networks was basically stable, but the metastability varied considerably. Among these, DMN and LIM had a relatively higher test-retest reliability of metastability than other networks. This paper provides a methodological reference for exploring the brain dynamic neural activity by using synchrony and metastability in fMRI signals.

Keywords: metastability; resting state fMRI; resting-state network; synchrony; test–retest reliability.

Figure 1

Overview of the workflow. Green boxes represent different factors. The orange box and blue box represent ICC index calculation and neurodynamic indexes calculation, respectively. ∆T1 can be used to evaluate short-term reliability and ∆T2 evaluates long-term reliability. See the methods for details.

Figure 2

Reliabilities of synchrony (a) and metastability (b) in different magnetic flux strength strategies, evaluated by the intraclass correlation coefficient (ICC).

Figure 3

Reliabilities of synchrony (a) and metastability (b) in different temporal resolutions, evaluated by the intraclass correlation coefficient (ICC). Blue bars represent short-term reliability and pink bars represent long-term reliability.

Figure 4

Reliabilities of synchrony (a) and metastability (b) in different spatial resolutions, evaluated by the intraclass correlation coefficient (ICC). Blue bars represent short-term reliability and pink bars represent long-term reliability.

Figure 5

Reliabilities of synchrony (a) and metastability (b) in denoising strategies for the three brain atlases, evaluated by the intraclass correlation coefficient (ICC). Blue bars represent short-term reliability and pink bars represent long-term reliability.

Figure 6

Reliabilities of synchrony (a) and metastability (b) in different node definition strategies, evaluated by the intraclass correlation coefficient (ICC). Blue bars represent short-term reliability and pink bars represent long-term reliability.

Figure 7

Reliabilities of synchrony (a) and metastability (b) in different resting-state networks, evaluated by the intraclass correlation coefficient (ICC). Blue lines represent short-term reliability and pink lines represent long-term reliability.

Figure 8

Reliabilities of interaction matrix of synchrony (a) and metastability (b), evaluated by the intraclass correlation coefficient (ICC). Color bar: 0–0.6.