Group ICA of resting-state data: a comparison

Abstract

Our results show that both approaches provide an adequate way of group ICA obtaining a comparable number of RSNs differing mainly in calculation times.

Fully exploratory network ICA (FENICA) on resting-state fMRI data

“…It is, however, still under discussion whether ICA should be performed at a group-level or specific single-subject components preselected for group inferences (Damoiseaux et al, 2006;DeLuca et al, 2006;Calhoun et al, 2008;Li et al, 2007;Schöpf et al, 2010). Group ICA approaches in which single-subject data is concatenated in time have been introduced by Jafri et al (2008) and Damoiseaux et al (2006).…”

Model-free fMRI group analysis using FENICA

“…A number of group ICA (gICA) approaches have been developed to extend ICA from the analysis of a single data set to the analysis of data from a group of subjects (Calhoun et al, 2009). The most widely adopted method is to concatenate singlesubject data in time prior to decomposition (Calhoun et al, 2001; (for a comparison of toolboxes using temporal concatenation ICA see Schöpf et al (2010b)). The computer memory requirements associated with processing all data sets simultaneously generally necessitate data be downsampled.…”

A Hitchhiker's Guide to Functional Magnetic Resonance Imaging

“…The selection of components of interest is not trivial (in the absence of an a priori hypothesis) and is usually performed by visual inspection or correlation with a predefined RSN template. While straightforward to implement in single-subject analyses, group ICA analyses are more complex and require choosing between several different workflows and algorithm definitions (Beckmann and Smith, 2004; Calhoun et al, 2009; Schöpf et al, 2010; Du et al, 2016). ICA methods also have been used extensively in rs-fMRI studies (Beckmann et al, 2005; Soares et al, 2016), task-based fMRI (Calhoun et al, 2008), and for artifact removal (Perlbarg et al, 2007; Feis et al, 2015; Pruim et al, 2015).…”