Task-merging for finer separation of functional brain networks in working memory

Functional Delineation of Prefrontal Networks Underlying Working Memory in Schizophrenia: A Cross-data-set Examination

“…Multi-experiment Constrained PCA for fMRI A multivariate approach that allows comparisons of functional brain networks between tasks/data sets has been previously put forward (Sanford, Whitman, & Woodward, 2020;Lavigne & Woodward, 2018;Ribary et al, 2017;Lavigne, Metzak, & Woodward, 2015), which is referred to as multi-experiment constrained PCA for fMRI (fMRI-CPCA). Multi-experiment fMRI-CPCA combines the power of cross-data-set comparisons with the comprehensive framework of a whole-brain, data-driven analysis.…”

“…However, if a cognitive process is elicited by one task but not the other, the task not eliciting this process should show a flat HDR for that network. Previously, we demonstrated that distinct cognitive/behavioral processes that load onto a single network because of being temporally correlated may separate onto different networks when data are analyzed with a task involving different types of cognitive demands (Sanford et al, 2020). In this study, multi-experiment fMRI-CPCA was used to characterize task-related networks comprising pFC.…”

Altered functional brain dynamics in chromosome 22q11.2 deletion syndrome during facial affect processing

“…After completing the outlined preprocessing steps, we used constrained principal components analysis (CPCA) 23,24 to extract task-evoked spatial modes of brain activation at the group-level with subject-level temporal weights 13,[20][21][22] . Briefly, this approach involves using a finite impulse response (FIR) basis set 34 to extract task-related variance from a set of BOLD timeseries, applying principal component analysis (PCA) to extract orthogonal spatiotemporal modes from the taskrelated variance, and then a second regression step using the same FIR basis set to determine how the temporal scores of each PC fluctuate with stimulus presentation.…”

“…In order to perform constrained principal component analysis (CPCA) 13,20,22 , we began with (N * T ) × P matrix X, containing the BOLD time series for P = 214 cortical and subcortical parcels (see "Functional data processing using XCP software") over T = 204 image acquisitions, concatenated across N = 116 total subjects from both HC and 22q11.2DS cohorts, such that X was 23664 × 214. Next, we constructed an FIR basis set F that modeled the r = 6 image acquisitions following the v = 6 task events (unique combinations of correct, incorrect, and non-responses to threatening or non-threatening stimuli 36,37 ) for each of the N subjects as a binary indicator, plus an intercept term.…”

“…Here, we hypothesized that primary visual and motor processing would be preserved in individuals with 22q11.2DS, while frontolimbic interactions subserving bottom-up emotion-processing 6,7 would be disrupted in individuals with 22q11.2DS. We applied constrained principal component analysis (CPCA) 13,[20][21][22][23][24] to identify brain activation patterns evoked by images of faces, and quantified their time course of activation after emotion identification. Specifically, we used emotion identification task fMRI data 8,11,25 acquired from 58 individuals with 22q11.2DS identified through the 22q and You Center at the Children's Hospital of Philadelphia, examined as part of a prospective brain-behavior study of 22q11.2DS, and 58 age-, sex-, and race-matched healthy controls (HCs) from the Philadelphia Neurodevelopmental Cohort 26,27 .…”