Building domain-agnostic frameworks for understanding complex dynamical systems — applied to the brain as the world's most complex test case.
Introduces functional inertia — a representation-agnostic constraint measuring how accumulated prior states resist ongoing reorganization. Three recurrent dynamical regimes (Locked / Stabilizing / Shifting) and a system-level inertial magnitude with opposite cognitive signatures across diagnostic contexts. Full mediation proves cumulative integration is essential.
Introduces DyCoM — a compact operator-level framework expressing dynamic connectivity estimators as compositions of four fundamental signal processing operations. Domain-agnostic by design: applies to any multivariate time series. Resolves previously conflicting neurobiological signatures through principled operator decomposition.
Kuramoto oscillator-informed graph neural network uncovering structure-function coupling in children with prenatal drug exposure. Links dynamic network topology to neurodevelopmental trajectories using physics-inspired graph learning.
Novel framework quantifying dynamic convergence and divergence across overlapping brain states in four psychiatric disorders simultaneously. Demonstrates shared and disorder-specific dynamic signatures. Published in Network Neuroscience, 10(1):93–117.
Koopman operator dynamics with Riemannian manifold alignment for predicting early substance use initiation from longitudinal connectome data. Bridges behavior-conditioned latent dynamics with brain connectivity trajectories using geometric deep learning.
Multi-scale adaptive graph attention network learning joint structural-functional brain representations for cognitive insight. Links anatomical connectivity to functional dynamics at multiple spatial scales via adaptive attention mechanisms.
Adapts the DyCoM co-modulation operator principle to structural neuroimaging — introducing individualized inter-component interaction measures in source-based morphometry. Demonstrates operator-level decomposition generalizes from functional to structural imaging.
Longitudinal study linking activity patterns to dynamic functional network states. Demonstrates brain-behavior associations evolve with training, with functional network regime occupancy providing a stable organizing framework across time.
Training-induced plasticity expressed through shifts in dynamic network state occupancy and divergence metrics. Associates regime engagement changes with behavioral and cognitive improvement in healthy young adults over training.
Characterizes functional connectivity alterations in major depressive disorder, identifying network-level disruptions and dynamic state changes as candidate biomarkers for clinical expression and treatment stratification.
Introduces Warp Quantification Analysis (WQA) — a comprehensive framework formalizing path-based signal alignment metrics. Unifies DTW distance, warp deviation variance (WDV), path length ratio, and alignment cost into a coherent, general-purpose analysis system applicable to any time-series alignment problem.
Proposes CVPS — a complex-valued phase synchrony framework using an adaptive wavelet that preserves both cosine and sine components, recovering true phase offsets and directional lead-lag structure. Reveals cortical drivers consistent with pharmacological receptor targets. PLV discards direction; CVPS keeps it.
Normalized DTW framework for capturing timescale-aligned amplitude balance across brain networks. Healthy brains maintain balanced signal distribution; schizophrenia shows aberrant recovery patterns in this balance that directly link to positive/negative symptom severity and cognitive performance.
Bootstrap Monte Carlo Singular Spectrum Analysis for robust state-space reconstruction of brain dynamics. Provides noise-resilient latent trajectory estimation with full uncertainty quantification — a principled approach directly applicable to any noisy multivariate dynamical system.
Communication-theoretic analysis proving phase synchronization and SWPC are complementary estimators, not competing ones. Phase methods capture weak high-frequency coupling; amplitude methods capture strong low-frequency dynamics. Method choice encodes a computational prior about signal structure.
Normalization of DTW distances substantially improves sensitivity for detecting functional network connectivity differences in schizophrenia. Establishes normalized DTW as a more clinically powerful connectivity metric than standard correlation approaches.
Early formulation of the functional inertia index through memory-retaining dynamics. Establishes adaptability as the inverse of inertia — a measurable, continuous axis for quantifying how readily any dynamical system reorganizes from its accumulated history.
Timescale normalization via DTW reveals systematic disruptions in dynamic signal-energy balance in schizophrenia. Establishes temporal normalization as a necessary preprocessing step before amplitude-based connectivity analysis in clinical populations.
Applies timescale-normalized DTW framework to reveal sex-specific differences in neural flexibility and metabolic demand dynamics in late childhood. Links dynamic amplitude balance to domain-specific cognitive support mechanisms.
Dynamic coupling across structural and functional MRI modalities reveals sex differences in developing brains from the ABCD dataset. Demonstrates operator-based co-modulation analysis generalizes seamlessly across imaging modalities.
Sex-specific coupling between dynamic functional connectivity and structural morphology in the ABCD pediatric dataset. Dynamic connectivity-morphology co-modulation differs systematically by sex across development.
Spatial dissimilarity trajectory analysis reveals rapid functional network refinement in early infancy, demonstrating temporal alignment methods capture neurodevelopmental dynamics across vastly different timescales — from milliseconds to months.
Sleep duration and aging interact to shape neural reactivation and episodic memory performance. Dynamic network state analysis reveals age-dependent changes in memory consolidation mechanisms across the adult lifespan.
Proposes Warp Elasticity — a novel metric using DTW alignment to capture temporal coupling timescales between network pairs. Captures nonlinear stretching and shrinking across brain regions. High replicability, noise robustness, and significant clinical discriminability across independent datasets. 16 citations.
Conference paper introducing warp elasticity for inter-network temporal coupling analysis. Validates warp states as clinically meaningful markers distinguishing healthy controls from schizophrenia with high replicability.
Foundational work establishing the amplitude-phase duality in functional connectivity estimation. Demonstrates the correlation between SWPC and phase synchrony is time-dependent — motivating the full communication-theoretic analysis that followed in Brain Connectivity 2025.