Condition-dependent disruption of low-frequency EEG–fMRI coupling reveals delayed hemodynamic timing and motor-network reorganization in chronic stroke

• Stroke delays and attenuates event-locked BOLD responses to motor stimulation. • EEG–fMRI reveals altered temporal alignment of neural oscillations and BOLD in stroke. • Lag polarity shifts may reflect delayed hemodynamics, but not causal reversal. • Hemispheric analysis reveals ipsilesional suppression and network reweighting in stroke. • EEG–fMRI coupling strength tracks motor impairment severity in stroke. Neurovascular coupling (NVC) is frequently disrupted after stroke, yet its temporal alignment and motor-network organization during sensorimotor engagement remain incompletely characterized. We tested whether chronic hemiparetic stroke alters the temporal alignment between EEG low-frequency oscillations (LFOs; 0.1–1.5 Hz) and the fMRI BOLD response during peripheral finger stimulation, and whether these alterations map onto motor-network anatomy and motor impairment. Simultaneous EEG–fMRI was acquired in 13 participants (7 chronic stroke; 6 healthy controls), during block-design peripheral transcutaneous index finger stimulation. EEG was decomposed using empirical mode decomposition to isolate physiologically grounded LFO components and Hilbert amplitude envelopes were incorporated into voxel-wise EEG-informed fMRI models. To evaluate robustness to hemodynamic assumptions, we implemented canonical HRF modeling, a derivative-augmented basis set, and subject-specific HRF estimation. Spatial inference was anchored to the Human Motor Area Template (HMAT). Stroke participants exhibited delayed and attenuated stimulation-evoked BOLD responses relative to controls. Cross-modal timing analyses revealed a distributional shift of peak EEG–BOLD alignment lags in stroke toward smaller and more frequently negative values, consistent with altered temporal alignment in the context of delayed hemodynamics, as opposed to causal inversion. The relative ordering of timing and coupling effects was preserved across HRF modeling strategies. HMAT-constrained analyses demonstrated reduced ipsilesional M1/PMd coupling during paretic stimulation with increased contralesional premotor and SMA recruitment. Importantly, ipsilesional motor-region coupling scaled with upper-extremity Fugl–Meyer scores. Together, these findings demonstrate convergent temporal and spatial reorganization of EEG–BOLD coupling after stroke and support LFO-informed EEG–fMRI as a physiologically interpretable framework for quantifying motor-network neurovascular timing.