Model order effects on independent vector analysis applied to complex-valued fMRI data

Model order effects on ICA of resting-state complex-valued fMRI data: Application to schizophrenia

“…We used the cluster quality index in ICASSO, which is software for investigating reliability of ICA estimates by clustering and visualization (Himberg et al, 2003(Himberg et al, , 2004, to evaluate consistency of a spatial component estimated from different runs of ICA. More specifically, we used the mean cluster quality index ̅ averaged first over selected components and then over all subjects for evaluation (Kuang et al, 2017a). The definitions of ̅ and the standard deviation ̅ averaged across 82 subjects are provided in Appendix B.…”

“…Motivated by model order effects seen in ICA of magnitude-only fMRI data, we were interested in how model order would affect ICA of complex-valued fMRI data. In a preliminary study, we investigated model order A C C E P T E D M A N U S C R I P T effects on independent vector analysis using 16 task-related complex-valued fMRI data sets (Kuang et al, 2017a). Results indicated that complex-valued ICA analysis also detected component splitting at higher model orders but differed from magnitude-only analysis in that an intact component and its subcomponents existed simultaneously.…”

Spatial source phase: A new feature for identifying spatial differences based on complex‐valued resting‐state fMRI data

“…In addition, Adrian et al () provided detailed and comprehensive evaluation of the benefits of using both magnitude and phase information in fMRI. Multiple publications have provided evidences that phase data plays a complementary role to magnitude data in extracting contiguous and meaningful brain activations (Kuang et al, ; Kuang, Lin, Gong, Cong, & Calhoun, ; Rodriguez, Calhoun, & Adalı, ; Rodriguez, Correa, Eichele, Calhoun, & Adalı, ; Yu et al, ). Moreover, the incorporation of fMRI phase data better preserves the integrity of larger networks (Kuang et al, ; Kuang et al, ), and the intact group default mode network (DMN) at higher model orders shows significant difference between patients with schizophrenia (SZs) and healthy controls (HCs; Kuang et al, ).…”

Shift-Invariant Canonical Polyadic Decomposition of Complex-Valued Multi-Subject fMRI Data With a Phase Sparsity Constraint

“…For each of shared SM estimate, i.e., s n , phase de-ambiguity is first performed using (17) and (18) to obtain θ(s v,n ), phase de-noising is then performed based on θ(s v,n ). Specifically, a binary mask…”

“…Each shared TC estimate b n is also phase corrected. Replacing z n in (17) with b n to estimate its rotation angle θ n , we obtain the phase corrected TC as follows:…”