Authors :
Presenting Author: Rupesh Chikara, PhD – The University of Texas at Arlington
Saeed Jahromi, M.Sc. – The University of Texas at Arlington; Mark Ottensmeyer, PhD – Department of Radiology – Harvard Medical School; Gianluca De Novi, PhD – Department of Radiology – Harvard Medical School; Eleonora Tamilia, PhD – Harvard Medical School; Phillip Pearl, MD – Department of Neurosurgery – Harvard Medical School; Steven Stufflebeam, MD – Harvard Medical School; Christos Papadelis, PhD – Cook Children’s Health Care System
Rationale:
Magnetic and electric source imaging (MSI/ESI) are source localization techniques based on magnetoencephalography (MEG) and high-density EEG recordings, which offer useful clinical information in the presurgical evaluation of drug resistant epilepsy (DRE) patients. Yet, they are rarely performed simultaneously despite the fact that they yield both confirmatory and complementary information and present several limitations when they are used individually. MSI has higher spatial resolution than ESI but is almost blind to epileptiform activity from deep brain structures or sources having radial orientation. Here, we assess the localization accuracy of combined ESI and MSI, namely electromagnetic source imaging (EMSI), in a 3D printed human head phantom and a cohort of DRE patients by comparing the accuracy of EMSI vs. individual modalities (ESI and MSI).
Methods:
We constructed a human head phantom that resembles the electromagnetic properties of human head. The phantom was built based on segmented surfaces taken from a three year old child with DRE (Figure 1a). Mixing chopped carbon fibers with liquid silicone at various ratios, molding the brain onto a base, and then creating the skull and soft tissue layers, we constructed a phantom that approximates the electrical properties of human head (Figure 1b). Dipolar sources were implanted at various locations (i.e., superficial, subcortical and deep areas) and activated with a signal generator using intracranial EEG (iEEG) recordings from a DRE patient. MEG and HD-EEG recordings were performed with this phantom (Figure 1c). The generated activity was localized with ESI, MSI and EMSI using equivalent current dipoles (ECDs) and dynamic statistical parametric mapping (dSPM). We assessed the localization accuracy as the distance of source imaging solutions from the actual source (Ds) identified through CT. We further analyzed simultaneous MEG and HD-EEG recordings from 23 children with DRE (12 females, mean age: 12.9±4.07) who underwent iEEG and surgery. The patients were dichotomized based on their surgical outcome 14 good outcome (Engel I) and nine poor outcome (Engel ≥ 2-4) patients. We localized spikes using the same methods as with the phantom. We calculated the distance from seizure onset zone (DSOZ) and resection (DRES).
Results:
By using the phantom, we generated signals that resemble human spikes in terms of amplitude, duration and morphology with both MEG and HD-EEG (Figure 1d). On the phantom, EMSI showed improved localization accuracy over the individual modalities versus both ECD [D
s=10.57±0.51 mm for EMSI vs.16.42±4.82 mm for MSI and 39.96±17.08 mm for ESI] and dSPM [D
s=6.50±1.00 mm for EMSI vs.11.09±2.61 mm for MSI and 27.74±10.02 mm for ESI] (averaged values across all sources) (Figure 1e). In good outcome patients, EMSI presented shorter D
SOZ (15.18±9.06 mm) and D
RES (8.56±6.24 mm) than ESI (D
SOZ: 25.04±16.20 mm, p< 0.009; D
RES: 18.88±17.30 mm, p< 0.03) and MSI (D
SOZ: 23.37±8.98 mm, p< 0.03; D
RES: 15.51±10.11 mm, p< 0.02) (Figure 2).