Abstracts

RATIONALE: Precise non-invasive MEG dipole localization of interictal spikes is useful for surgery in medically intractable partial epilepsy. Precise localization improves estimates of: the center of the epileptic zone to optimize placement of intracranial electrodes and the extent of the zone for complete removal. Multiple factors affect localization including neuromagnetometer resolution, co-registration between MEG and MRI, volume conductor effects and noise. These effects were assessed for a whole cortex neuromagnetometer system in research and clinical use.
METHODS: The neuromagnetometer (CTF Systems, Vancouver, CAN) had 100 SQUIDs, 68 sensor and 32 reference channels to form 3rd derivative gradiometers for noise cancellation in a shielded chamber (BTi, San Diego, CA; Vacuumschmelze, Hamburg, GER). A single moving dipole model, the clinical standard, from each of two software packages (CTF and BrainStorm USC) was used to analyze the data. Two physical dipole phantoms were used. One phantom was a [dsquote]dry[dsquote] dipole simulator often used in calibration (Neuromag, Helsinki, FINN); the dipole locations were known precisely in relation to the phantom fiducials. The other phantom was a [dsquote]wet[dsquote] saline-filled sphere, in which the current dipole locations were measured in relation to the phantom fiducials which were co-registered to the MEG sensors by a 3D Polhemus digitizer. The sphere also was imaged in a Toshiba 0.5 T open MRI scanner. A sinusoidal current at 7 Hz was passed through each dipole for 10 seconds and the models applied to determine the best fit to the data. The difference between estimated location and measured location of each dipole was calculated. We tested 11 dipoles in the calibration phantom and 35 dipoles in the sphere phantom, from a depth of 5 mm to 50 mm.
RESULTS: Localization errors were : 1) calibration phantom dipole 0.97 mm (SD 0.40) and 2) and sphere phantom dipole 1.4 mm (SD 0.9). The sphere phantom errors were larger than the calibration phantom errors (Student[ssquote]s t-test, p[lt]0.02).
CONCLUSIONS: MEG dipole localization in physical phantoms was confirmed to be accurate for a whole cortex system with 68 sensor sites and 3rd derivative gradiometers in a shielded room, as theoretically predicted (Mosher J, Spencer M, Leahy R, Lewis, P. Electroenceph clin Neurophysiol.1993;86:303-320). The larger errors in the co-registered sphere, which more closely mimics the human head in MEG and MRI clinical studies, were remarkably small. This has implications for better estimates of the epileptic zone.
Support: National Institutes of Health grants NS20806, RR13276, MH53213, Huntington Medical Research Institutes, Huntington Hospital, Epilepsy and Brain Mapping Program, and Robert S. and Denise Zeilstra.