Clinical utility of current-generation dipole modelling of scalp EEG

Abstract

Objective: To investigate the clinical utility of current-generation dipole modelling of scalp EEG in focal epilepsies seen commonly in clinical practice.

Methods: Scalp EEG recordings from 10 patients with focal epilepsy, five with Benign Focal Epilepsy of Childhood (BFEC) and five with Mesial Temporal Lobe Epilepsy (MTLE), were used for interictal spike dipole modelling using Scan 4.3 and CURRY 5.0. Optimum modelling parameters for EEG source localisation (ESL) were sought by the step-wise application of various volume conductor (forward) and dipole (inverse) models. Best-fit ESL solutions (highest explained forward-fit to measured data variance) were used to characterise best-fit forward and inverse models, regularisation effect, additional electrode effect, single-to-single spike and single-to-averaged spike variability, and intra- and inter-operator concordance. Inter-parameter relationships were examined. Computation times and interface problems were recorded.

Results: For both BFEC and MTLE, the best-fit forward model was the finite element method interpolated (FEMi) model, while the best-fit single dipole models were the rotating non-regularised and the moving regularised models. When combined, these forward-inverse models appeared to offer clinically meaningful ESL results when referenced to an averaged cortex overlay, best-fit dipoles localising to the central fissure region in BFEC and to the basolateral temporal region in MTLE. Single-to-single spike and single-to-averaged spike measures of concordance for dipole location and orientation were stronger for BFEC versus MTLE. The use of an additional pair of inferior temporal electrodes in MTLE directed best-fit dipoles towards the basomesial temporal region. Inverse correlations were noted between unexplained variance (RD) and dipole strength (Amp), RD and signal to noise ratio (SNR), and SNR and confidence ellipsoid (CE) volume. Intra- and inter-operator levels of agreement were relatively robust for dipole location and orientation. Technical problems were infrequent and modelling operations were performed within 5min.

Conclusions: The optimal forward-inverse single dipole modelling set-up for BFEC and MTLE interictal spike analysis is the FEMi model using the combination of rotating non-regularised and moving regularised dipoles. Dipole modelling of single spikes characterises best-fit dipole location and orientation more reliably in BFEC than in MTLE for which spike averaging is recommended.

Significance: The clinical utility of dipole modelling in two common forms of focal epilepsy strengthens the case for its place in the routine clinical work-up of patients with localisation-related epilepsy syndromes.