Realistic spatial sampling for MEG beamformer images

Gareth R Barnes, Arjan Hillebrand, Ian P Fawcett, Krish D Singh

Research output: Contribution to journalArticleAcademicpeer-review

83 Citations (Scopus)


The spatial resolution achievable using magnetoencephalography (MEG) beamformer techniques is inhomogeneous across the brain and is related directly to the amplitude of the underlying electrical sources [Barnes and Hillebrand, Hum Brain Mapp 2003;18:1-12; Gross et al., Proc Natl Acad Sci USA 2001;98:694-699; Van Veen et al., IEEE Trans Biomed Eng 1997;44:867-860; Vrba and Robinson, Proc 12th Int Conf Biomagn 2001]. We set out to examine what an adequate level of spatial sampling of the brain volume is in a realistic situation, and what implications these inhomogeneities have for region-of-interest analysis. As a basis for these calculations, we used a simple retinotopic mapping experiment where stimuli were 17-Hz reversing gratings presented in either left or right visual hemifield. Beamformer weights were calculated based on the covariance of the MEG data in a 0-80 Hz bandwidth. We then estimated volumetric full-width half-maximum (FWHM) maps at a range of sampling levels. We show that approximately 10% of the 1 mm cubic voxels in the occipital volume have a FWHM smoothness of <5 mm, and 80% <10 mm in three subjects. This was despite relatively low mean signal-to-noise ratios (SNR) values of 1.5. We demonstrate how visualization of these FWHM maps can be used to avoid some of the pitfalls implicit in beamformer region-of-interest analysis.

Original languageEnglish
Pages (from-to)120-7
Number of pages8
JournalHuman brain mapping
Issue number2
Publication statusPublished - Oct 2004
Externally publishedYes


  • Brain Mapping
  • Brain/physiology
  • Humans
  • Magnetoencephalography
  • Photic Stimulation
  • Visual Perception/physiology

Cite this