Neuromark dFNC Patterns: A fully automated pipeline to estimate subject-specific states from rs-fMRI data via constrained ICA of dFNC in +100k Subjects

Abstract

Resting-state functional MRI (rs-fMRI) provides valuable insights into brain function during rest, but faces challenges in clinical applications due to individual differences in functional connectivity. While Independent Component Analysis (ICA) is commonly used, it struggles to balance individual variations with inter-subject information. To address this, constrained ICA (cICA) approaches have been developed using templates from multiple datasets to improve accuracy and comparability. In this study, we collected rs-fMRI data from 100,517 individuals across diverse datasets. Data were preprocessed through a standard fMRI pipeline. Our method first used replicable fMRI component templates as priors in constrained ICA (the NeuroMark pipeline), then estimated dynamic functional network connectivity (dFNC). Through clustering analysis, we generated replicable dFNC states, which were then used as priors in constrained ICA to automatically estimate subject-specific states from new subjects.This approach provides a robust framework for analyzing individual rs-fMRI data while maintaining consistency across large datasets, potentially advancing clinical applications of rs-fMRI.