Authors :
Presenting Author: Kamran Salayev, MD, PhD – Boston Children's Hospital
Banu Ahtam, PhD – Boston Children's Hospital
Eleonora Tamilia, PhD – Boston Children’s Hospital, Harvard Medical School, Boston, MA, USA
Alexander Rotenberg, MD, PhD – Boston Children's Hospital, Harvard Medical School, Boston, MA, USA
Ellen Grant, MD – Boston Childrens Hospital
Rationale:
Magnetoencephalography (MEG) is a valuable tool for localizing the epileptogenic zone (EZ), offering a unique combination of high temporal and spatial resolution. Epileptic activity typically appears in MEG as spike waves, defined as sharp transients that last up to 70 milliseconds. Spikes provide valuable information for estimating the location of the EZ, with the Equivalent Current Dipole (ECD) model being the most widely used method. The ECD model assumes that brain activity can be approximated by a single dipole, with the dipole’s center representing the presumed source. Epileptic activity can propagate from the EZ to surrounding regions within milliseconds. Thus, the time point selected for ECD computation within the spike waveform can influence the estimated source location. Identifying the optimal time point for ECD modeling is critical to improving the accuracy of MEG-based localization. Some MEG spikes show a smaller, early peak preceding the main peak, which may also localize to the EZ. This study investigates how ECD localization varies depending on the timing of MEG spike waveform in children with refractory epilepsy under four years of age.Methods:
We analyzed recordings acquired with BabyMEG system during presurgical evaluation. MEG data were visually inspected for epileptic spikes. For selected spikes, Global Field Power (GFP) was computed across all sensors. We selected these time points along the GFP waveform: 1) the early smaller GFP peak preceding the main peak; 2) the first dipole on the upslope; 3) the main GFP peak at the maximum of the GFP curve (Figure 1). Only spikes with a well-defined dipole pattern at all three points were analyzed. We calculated and compared ECD coordinates across three time points using paired t-tests. Spatial clustering of ECD sources were evaluated by calculating centroids and pairwise Euclidean distances.
Results:
We analyzed a total of 100 spikes from four patients and selected 33 that met the inclusion criteria for further analysis. The distance between ECD locations at three time points was greater than zero (p < .0001). Paired t-tests confirmed that the difference between ECD locations was significant (p < .05). The spatial compactness of ECD clusters varied considerably across patients (Table 1).