Hi! Everyone! I am a beginner with MEG source localization.
I want to know why the Flash sequence is better to resolve the inner/outer skull segmentation? I remember seeing it on MNE or Freesurfer tutorials once, but I can’t find it.
I’m really confused about this problem, any suggestions? Thanks a lot!
Best
This section talks about the different flash sequences used, but doesn’t say why certain sequences are better / worse for different tissue contrasts: Algorithms and other implementation details — MNE 1.4.2 documentation
maybe @larsoner or @agramfort or @jstout211 or @jhouck know?
I do not, but if someone does we should add it to the docs 
@TTingW - this is much more important for EEG localization to have the 3 layer BEM. The skull is like 20-40X the resistance of the other tissues, so the thickness of this layer has an outsized effect on the forward model. For MEG it is sufficient to have a single layer BEM of the inner skull as all of the tissues have a similar effect on the magnetic fields. (Hand wavy explanation ->) The inner skull bem is necessary for MEG because you can get a buildup of charge or secondary current at the boundary of the inner skull. This needs to be incorporated into the model or else the depth of the source localization will be off.
If you do want get a 3-layer model without the Flash seq, you can add the -T2 flag during your freesurfer processing. This requires a T2 scan, which is much more common than the suggested multi angle FLASH sequence for BEM. I don’t know how the results compare to each other - but visually the T1+T2 freesurfer produces good tissue boundaries.
But back to your original question - why does the flash sequence resolve the skull better. It just has to do with different sensitivities to the scan parameters (water vs fat content of the tissue).
–Jeff
What @jstout2112 said 
This is also addressed pretty well in the Fischl et al. 2004 footcite on the page, e.g.:
Finally, it is clear that multispectral information (e.g., different flip angles or different TRs) provides additional information that is not present in a single T1-weighted image. This is due to the highly correlated nature of T1 and proton density (their correlation coefficient in brain is greater than 0.5) and the fact that they influence contrast in opposite directions. For example, cortical gray matter has a longer T1 and a higher proton density than the underlying white matter. The longer T1 tends to darken the gray matter, while the higher PD brightens it, resulting of course in the reversal of contrast seen in T1-weighted versus PD-weighted images. Thus, it is clear separating the effects of different tissue properties can significantly enhance class separation.
For MEG, the benefits of the PD sequence probably come more from the difference in proton density between CSF and everything else.