State-specific individualized functional networks form a predictive signature of brain state

Abstract

There is extensive evidence that human brain functional organization is dynamic, varying within a subject as the brain switches between tasks demands. This functional organization also varies across subjects, even when they are all engaged in similar tasks. Currently, we lack a comprehensive model that unifies the two dimensions of variation (brain state and subject). Using fMRI data obtained across multiple task-evoked and rest conditions (which we operationally define as brain states) and across multiple subjects, we develop a state-and subject-specific functional network parcellation (the assignment of nodes to networks). Our parcellation approach provides a measure of how node-to-network assignment (NNA) changes across states and across subjects. We demonstrate that the brain's functional networks are not spatially fixed, but reconfigure with brain state. This reconfiguration is robust and reliable to such an extent that it can be used to predict brain state with accuracies up to 97%.Recent studies have made significant progress in characterizing state-evoked changes in functional connectivity elicited by task performance (20-28). However, none of these works explicitly examined the possibility that networks spatially reconfigure: (i) The majority of these studies explicitly assumed that networks remain spatially unchanged across tasks (21,22,29). These analyses were restricted to investigating differences in connectivity between networks, and not whether networks spatially reconfigure across tasks. (ii) Another line of research takes a more abstract perspective by computing global network measures (such as modularity and participation coefficient) and comparing these measures across different states (27,28,(30)(31)(32). Such approaches address the modular reconfiguration of the brain as a whole by defining a new set of networks for each state, but they do not quantify how or whether the same networks change across states. Such studies also typically define a small number of networks (~3-7) whose correspondence to the putative resting-state networks are unclear. These approaches cannot answer how specific networks reorganize as a function of brain state. (iii) Finally, many studies do not directly examine cross-subject variations in network reorganization (20), or treat crosssubject and cross-session changes as similar notions for defining reorganization (33). However, functional organization varies significantly across subjects, yielding large individual differences in network definitions (34,35). There is a clear need to account for both cross-subject and crossstate variability in functional network organization when characterizing brain dynamics.Here, we use a novel approach to dynamically map the brain's functional networks across taskevoked and resting states and across individuals. We operationalize the brain's functional subunits as a set of 268 pre-defined nodes (6). We then apply our recently developed exemplarbased parcellation method (36) to assign a set of pre-defined nodes to individualized, stat...

Unraveling reproducible dynamic states of individual brain functional parcellation

“…In the literature of brain parcellation, the only approach that had consistent findings with our approach was published by Salehi and colleagues [6,40] . We both suggested an approach that contradicts the notion of a fixed functional parcellation of the brain.…”

“…Even though dynamic functional connectivity is well studied, to the best of our knowledge, the only parcellation approach that considered dynamic changes of parcels was suggested by Salehi and colleagues. These authors demonstrated that the brain functional parcellations are not spatially fixed, but reconfigure with task conditions [6,40] . These reconfigurations were used to reliably predict different task conditions.…”

Functional Connectivity of Successful Picture-Naming: Age-Specific Organization and the Effect of Engaging in Stimulating Activities

“…Publicly available ROIs enhance comparability with other groups and minimize the bias of manually selecting components from our small sample. On the other hand, previous work demonstrated that sample-and task-specific parcellation was more functionally relevant (Dickie et al, 2017;Geerligs et al, 2017;Salehi et al, 2017Salehi et al, , 2018. To achieve a balance between sensitivity and specificity, a three-step methodology was used, combining a sample analysis of the task main effect along with spatial and time mapping with CONN parcellation.…”

Inside information: Systematic within-node functional connectivity changes observed across tasks or groups

“…ICA can provide multiple spatially overlapping components and these can be allowed to vary with condition. These ICA components are often few in number (30–70) and may not be spatially contiguous and share more in common with network analyses rather than node analyses ( Salehi et al, 2018 ; Smith et al, 2011 ). In the gradient work ( Margulies et al, 2016 ; Vázquez-Rodríguez et al, 2019 ) principal axes have been investigated and in some cases these have been shown to vary with task ( Tian et al, 2020 ).…”