Spatiotemporal localization of significant activation in MEG using permutation tests.

Area of Science:

• Neuroimaging
• Biophysics
• Statistical analysis

Background:

• Magnetoencephalography (MEG) provides high temporal resolution for studying brain activity.
• Cortical surface mapping is crucial for localizing neural sources from MEG data.
• Multiple testing is a significant challenge in neuroimaging analysis due to the large number of statistical tests performed.

Purpose of the Study:

• To develop and validate a statistical method for detecting activation in MEG-derived cortical current density maps.
• To address the multiple testing problem in analyzing high-density cortical surface data.
• To establish reliable thresholds for statistical significance in event-related MEG studies.

Main Methods:

• Application of non-parametric permutation tests to cortically-constrained MEG data.
• Computation of current density maps using inverse imaging methods.
• Random permutation of pre- and post-stimulus data to control the familywise error rate.
• Thresholding based on the distribution of maximum intensity to account for spatial and temporal correlations.

Main Results:

• The developed permutation test method effectively controls the familywise error rate.
• Thresholds derived from maximum intensity distribution implicitly handle correlations in cortical maps.
• Successful demonstration on both simulated and experimental somatosensory evoked response data.

Conclusions:

• Non-parametric permutation tests offer a robust approach for detecting neural activation from MEG data.
• The method provides a statistically sound way to manage multiple comparisons in cortical source imaging.
• This technique enhances the reliability of findings in event-related MEG studies.