|
Abstract:
|
A significant focus of research on resting-state fMRI (rs-fMRI) data has recently begun to shift from the evaluation of static functional connectivity over a scanning session, to the analysis of dynamic functional connectivity. Previous work proposed the Dynamic Conditional Correlation (DCC) model as an efficient approach for estimating dynamic correlations between rs-fMRI time series from different brain regions. While DCC was shown to be less susceptible to noise-induced temporal variability in correlations than other commonly used approaches, the method becomes biased and computationally expensive as the number of dimensions (i.e. brain regions) increases. In this work we propose a moment-based estimator for DCC (denoted MDCC), along with a fast estimation algorithm, which achieves both more accurate estimation and higher efficiency for high-dimensional time series. We investigate the properties of the newly proposed estimator in various simulation settings, and compare its performance with a recently developed composite likelihood method. The application of MDCC to simulated and real rs-fMRI data, illustrates its efficacy.
|
