Neuro-Current Response Functions

Neuro-Current Response Functions estimate spatially distributed linear filters that map cortical current responses from magnetoencephalography (MEG) and electroencephalography (EEG) data to localize and characterize neural responses to continuous auditory stimuli such as speech.

Key Features:

• Linear Filter Representation: Spatially distributed linear filters (NCRFs) predict cortical currents underlying MEG/EEG signals during continuous speech and extend temporal response functions (TRFs) to source space.
• Bayesian Framework: Estimation is embedded in a Bayesian framework that incorporates priors on structural properties and unifies TRF and source estimation.
• Generalization to non‑MR data: Supports free‑orientation dipoles for M/EEG recordings without individual structural MR scans and implements a regularizing scheme to reduce dependence on precise coordinate co‑registration.
• Champ‑Lasso Algorithm: Employs the Champ‑Lasso algorithm, a fast optimization‑based estimator for efficient reconstruction of response functions and dipole orientations.
• Direct Estimation: Directly estimates NCRFs from neuroimaging data, bypassing traditional two‑stage TRF then source localization approaches.
• Simulation and Experimental Validation: Demonstrated on simulated and experimentally recorded MEG data with reported improvements in spatial resolution and response function reconstruction compared to existing methods.

Scientific Applications:

• High Spatiotemporal Resolution: Delineates distinct cortical distributions of neural processes with high spatiotemporal precision during continuous stimuli such as speech.
• Behavioral Correlation: Links specific components of TRFs to behavioral attributes such as attention to relate neural response components to cognitive functions.

Methodology:

Direct estimation of NCRFs from neuroimaging data within a Bayesian framework using priors on structural properties, modeling with free‑orientation dipoles and a regularizing scheme to mitigate MR co‑registration dependence, parameter estimation via the Champ‑Lasso optimization algorithm, and validation using simulated and experimentally recorded MEG data.