Human Brain Mapping
Florence, Italy
June 11-15, 2006
Diane Whitmer (1), Il Keun Lee (1), Greg Worrell (2) & Scott Makeig (1) , 1. Institute for Neural Computation, University of California San Diego, La Jolla CA
2. The Mayo Clinics, Rochester, Minnesota
Independent co-modulation of spectral EEG power in subsets of independent brain sources
OBJECTIVE: Localization of epileptic foci is a critical aspect of treatment of patients with medically intractable epilepsy, and field potential recordings from surgically implanted intracranial EEG (iEEG) electrodes are used by neurologists in complex cases. However, because of volume conduction within the brain, signals recorded from clinical iEEG arrays or strips (with typical 1-cm spacing) may not be generated near to the electrodes, though activity near to the active or reference electrodes may typically dominate. Here we show how sufficient analysis of simultaneous intracranial iEEG and scalp EEG recordings may improve the signal-to-noise ratio, spatial resolution, and interpretability of iEEG data. We apply independent component analysis (ICA) to simultaneously recorded iEEG and EEG data to determine the relative contributions of proximal and distal brain sources in iEEG channel signals, and to explore whether separating proximal and distal contributions can improve the specificity and interpretability of iEEG recordings.
METHODS: Simultaneous scalp data and intracranial EEG (iEEG) data were recorded from clinical cortical-surface grids (70 channels) and strip (8 channels) in a patient undergoing pre-surgical continuous video-EEG monitoring for epilepsy. In a second epilepsy patient, concurrent bilateral temporal depth electrode probes (16 channels) and scalp EEG data (31 channels) were recorded during performance of a cued motor response task. In each patient, maximally independent components of the joint EEG and iEEG data were separated by infomax ICA (Makeig et al, 1996) applied to 20 minutes of continuously recorded data.
RESULTS: We used kurtosis of the independent component (IC) maps to quantify the degree to which the component projection was distributed across the intracranial channels or isolated to a single channel. This served as an indirect measure of proximity of the IC source region to the recording array. In the first patient, ICs varied widely in the diffusivity of their projections across the electrode arrays, as illustrated in a histogram (figure, top). Low kurtosis values correspond to spatially diffuse IC maps; high kurtosis values to focal maps (figure, middle left). Percent variances of electrode channel activities accounted for by each independent component were computed separately for the scalp electrodes and intracranial electrodes. Combined with kurtosis, these measures likely disambiguated larger, proximal sources from weak, distal ones. The figure also shows a similar histogram and two IC maps for the patient with temporal depth strips.