Implanted medical devices or other strong sources of interference are not barriers to magnetoencephalographic recordings in epilepsy patients.

OBJECTIVE

Localization accuracy in magnetoencephalography (MEG) recordings is highly dependent on signal to noise ratio, which is difficult to control.

METHODS

We have post-processed our data in order to reduce noise to a level permitting adequate source localization with equivalent current dipole methods. In 30 consecutive epilepsy patients, MEG was recorded using a whole-head MEG system consisting of 204 planar gradiometer and 102 magnetometers, with simultaneous EEG. Data were reviewed to identify interictal spikes. The initial analysis was done after employing a spatiotemporal signal space separation (tSSS) method. A total of 18 dipole clusters in 15 patients were reanalyzed without tSSS, to compare the number, goodness of fit, and locations of acceptable dipoles before and after processing.

RESULTS

In 8 of 18 clusters, although acceptable dipole clusters were captured before processing, there was a clear improvement of all parameters with tSSS. In another 5 clusters, all from patients with vagus nerve stimulators, there were few or no acceptable dipoles before processing, but sufficient dipole clusters were obtained with tSSS.

CONCLUSION

In contrast to volunteer research subjects, clinical patients cannot be expected to cooperate as fully, and their MEG data are likely to include more interference. This study demonstrates that processing the MEG data with a method to eliminate artifact arising from outside the brain significantly improves the data.

SIGNIFICANCE

In some cases, this improvement can mean the difference between satisfactory dipole fits vs no possible localization.