Fine-scale spatial dynamics within functional brain networks manifest as highspatial-frequency variations that conventional independent component analysis(ICA) methods fail to capture. These subtle changes may carry criticalinformation about transient connectivity and disordered brain function. Wedeveloped NeuroMark-DyFICA (dynamic frequency-informed ICA), a novel frameworkto enhance detection of spatiotemporal variability in fMRI data. It integratesthree stages: dynamic NeuroMark ICA across sliding windows to estimatetime-varying, spatially constrained networks; high-pass spatial filtering toemphasize fine-scale spatial features; and group-level ICA to extract refineddynamic components with subject-specific mixing weights. Unlike prior NeuroMarkapplications or conventional dynamic ICA, NeuroMark-DyFICA establishes areproducible latent space of high-frequency dynamics, uniquely capturingtransient, fine-scale reconfigurations of network topography. Validation using acontrolled 2D simulation demonstrated reliable detection of subtle spatialshifts mimicking pathology, which conventional ICA failed to recover. Applyingto resting-state fMRI from schizophrenia patients and healthy controls, multiplenetworks were estimated. We highlight six representative systems (thalamus,auditory, visual/fusiform, middle frontal, default mode, cerebellum). Resultsrevealed two complementary abnormalities in schizophrenia: an imbalance betweeninactive and hyper-engaged states and altered convergence among dynamic states.NeuroMark-DyFICA reveals fine-grained spatiotemporal disruptions in brainnetworks, offering mechanistic insights and potential biomarkers for psychiatricdisorders.Brain network dynamics are most often studied as changes in the expression orinteraction of large-scale, canonical patterns over time. While this perspectivehas been highly productive, it implicitly emphasizes low spatial-frequencyvariations and may overlook more focal, transient reconfigurations that occurwithin otherwise stable networks. Neurophysiological and clinical evidenceincreasingly suggests that cognition- and disease-related alterations canmanifest as subtle, localized shifts in network boundaries, internal structure,or spatial coherence—features that reside in the high spatial-frequencydomain and are difficult to capture with conventional dynamic approaches. Inthis study, we introduce NeuroMark-DyFICA (dynamic frequency-informedindependent component analysis), a framework designed to capture dynamic,fine-scale spatial variations in functional brain networks while preservingbiologically meaningful network structure. The method integrates template-guidedindependent component analysis with spatial-frequency filtering and group-leveldecomposition, enabling isolation of transient, high-frequency spatial changeswithin canonical networks in a fully automated and reproducible manner. Usingcontrolled simulations, we demonstrate that NeuroMark-DyFICA is sensitive tosmall spatial shifts that conventional static or low-frequency–focusedmethods often overlook. When applied to resting-state fMRI data from individualswith schizophrenia and healthy controls, the framework reveals abnormal dynamicspatial patterns across auditory, visual, frontal, default-mode, and cerebellarnetworks, including altered prevalence and coordination of dynamic spatialstates. By explicitly targeting fine-scale spatial dynamics, NeuroMark-DyFICAprovides a new lens for studying how brain networks reconfigure over time andhow these processes are disrupted in neuropsychiatric disorders. This approachoffers a principled pathway toward more sensitive biomarkers and advancesprecision neuroscience by bridging large-scale network organization withlocalized, clinically relevant variability.
Behzadfar et al. (Tue,) studied this question.