Authors :
Presenting Author: Glykeria Sdoukopoulou, MSc – Cook Children's Health Care System
Saeed Jahromi, MS – Cook Children's Health Care System
Ludovica Corona, PhD – Jane and John Justin Institute for Mind Health, Neurosciences Center, Cook Children's Medical Center
Cynthia Keator, MD – CookChildren's Health care System
Linh Tran, MD – CookChildren's Health care System
Saleem Malik, MD – CookChildren's Health care System
M. Scott Perry, MD – Jane and John Justin Institute for Mind Health, Neurosciences Center, Cook Children's Medical Center
Dave Shahani, MD – CookChildren's Health care System
Christos Papadelis, PhD – Cook Children's Health Care System
Rationale: Interictal spikes are the most sensitive biomarkers of epilepsy. Yet, they cannot be seen in the electroencephalography (EEG) of ~10% of epilepsy patients. It is unclear whether their brain generate spikes or the underlying generators do not produce an electromagnetic field measurable out of the scalp. Prior studies of simultaneous EEG and magnetoencephalography (MEG) have shown that ~30% of the epileptiform activity is visible in only one modality. This may be due to the different EEG and MEG sensitivity profiles for varied source orientations. For example, MEG is mostly blind to activity from quasi-radial sources at the gyral crest, while EEG is sensitive to all source orientations. In this computational simulation study, we assess the sensitivity profiles of high-density EEG and MEG regarding the spike location and orientation and their effect on localization.
Methods:
We reconstructed cortical surfaces of ~25,000 tessellations using MRI from five children with drug resistant epilepsy. Using these surfaces as source space, we modeled the sources as dipoles and placed them at seven regions in both hemispheres, whose tessellation defined each source’s orientation. The angle between the normal vectors of the cortical dipolar tessellation and the nearest inner skull surface defined the source category (quasi-radial or quasi-tangential). We activated one source at a time using a spike waveform obtained from intracranial EEG of an epilepsy patient. We then extracted artifact-free background activity of five age-matched healthy controls (Fig. 1A). Epochs were obtained by superimposing the background activity to the simulated EEG and MEG recordings and then reviewed by an epileptologist for spike marking. We computed the signal-to-noise (SNR) ratio of each detected spike (Fig. 1B), percentage of detected spikes per orientation, and descriptive statistics of their SNR. We then performed electric, magnetic, and electromagnetic source imaging (ESI, MSI, and EMSI, respectively) on the simulated EEG, MEG, and combined modalities. Sources were localized with dipole scanning. Finally, we computed the localization error (LE) as the Euclidean distance between the reconstructed and real sources (Fig. 1C). Results: We found that MEG spike sensitive in terms of spike detectability, increases with SNR (
p<0.001,
Wilcoxon rank-sum test) for quasi-tangential sources (
Fig. 2B&C). Similarly, EEG detects more spikes with a higher SNR (
p<0.001,
Wilcoxon rank-sum test) for quasi-radial sources (
Fig. 2B&C). ESI was insensitive to different orientations; MSI decreased the LE for the quasi-tangential sources (
p< 0.001,
Wilcoxon rank-sum test). We did not find a LE improvement for EMSI (
Fig. 2D). Finally, we found increased regional SNR in MEG (
p<0.001,
Wilcoxon rank-sum test) for all source configurations (
Fig. 2E).
Conclusions: MEG is mainly sensitive to quasi-tangential sources located at the sulci walls, while EEG captures the quasi-radial sources at the gyri crown or sulci bottom. Both methods offer LE< 20 mm, yet MEG is more precise than EEG in localizing spikes generated by quasi-tangential sources. EMSI does not improve the LE for superficial cortical sources.
Funding: R01NS104116 and R01NS134944 by NINDS.