Abstracts

Despite the advances in functional neuroimaging, the temporal resolution of electroencephalography (EEG) is still unmatched. EEG source modelling based on dipolar inverse solutions can improve spatial resolution. Recently updated [italic]Scan-Curry [/italic]software for [italic]Windows[/italic] offers greater ease-of-use and clinical utility in epilepsy investigation., Patients: Five children with BFEC and four children, one adult with MTLE due to Hippocampal Sclerosis (HS).
EEG and Dipole Modelling: 10-20 scalp recordings at 256Hz sampling (19 channel BFEC, 21 channel MTLE).[italic]Scan4.3[/italic] used to epoch interictal spikes (-200ms pre-spike, 500ms post-spike); uploaded to [italic]Curry5.0[/italic] for filtering, noise estimation, PCA/ICA analysis, and dipole modelling across spike onset-to-peak interval.
Volume conductor models: 1, 2, 3, 4 shell; and 3 realistic (boundary, finite element).
Dipole models: moving, rotating, regional, fixed coherent (+/- regularisation), fixed MUSIC.
Statistics: Best fit solutions by forward fit to measured data variance. Factor analysis on latency, fit strength (amplitude), confidence ellipsoid volume (CE), residual deviation (RD), and signal to noise ratio (SNR).
Ease-of-use of the software based on modelling computation times, database problem logbook., For BFEC and MTLE, FEM ranked highest on cumulative z score (amplitude, CE, RD), observed position, orientation, and intersubject consistency referenced to the MNI. Realistic models (BEM, FEM) gave more basal z+ axis dipole positions versus Shell models in MTLE; and gave xyz axis dipole positions that effaced the cortical MNI surface versus more scattered Shell model positions.
BFEC (FEM): rotating (non regularised) and moving (regularised) models ranked highest. Dipole models converged along the line of the central sulcus while orientations varied between subjects (tangential to radial).
MTLE (FEM): rotating and regional (non regularised) and moving (regularised) dipole models gave the best z scores. Versus the BFEC group, dipole models showed less convergence for position and more variability across parameters. Dipoles displayed horizontal-oblique orientations projecting to either mesial cortex or temporal tip.
Factor analysis showed a close correlation between CE and SNR.
Dipole modelling operations were performed within 5-10 minutes. Computation times ranged from milliseconds to within 2 seconds., EEG dipole source imaging with Scan4.3-Curry5.0 updated for Windows can be performed easily.
The confidence ellipsoid (CE) volume is a novel parameter which adds to the interpretation of the goodness of dipole fit and correlates closely to the SNR.
For BFEC and MTLE, the best head model to use with the MNI brain is the FEM.
The best dipole modelling approach for both conditions is the combined use of the non-regularised rotating dipole and the regularised moving dipole., (Supported by Australian NHMRC Grant. Software provided by Compumedics.)