Paving the way for cross-site pooling of magnetoencephalography (MEG) data

Abstract Pooling of magnetoencephalography (MEG) data across laboratories is non-trivial because of differences in hardware, software, and environmental noise levels. To investigate these issues, we conducted a study with the same five subjects at three sites with different MEG arrays: Elekta-Neuromag Vectorview (Boston), VSM MedTech Omega 275 (Albuquerque), and 4D Neuroimaging Magnes 3600 WH (Minneapolis). Subjects were run in a simple somatosensory paradigm. Phantom data were also taken at each site. We developed software to allow comparisons of test/retest reliability across subjects, machines, and analysis methods. Preliminary analyses showed excellent test/retest results within subjects across instruments. Analysis of phantom data showed a localization error of less than 2 mm across multiple software packages. Our results show that (1) instruments from different manufacturers yield similar results for somatosensory data, and that (2) multiple software packages produce very accurate results for simple source configurations.