• View
• Edit
• History
• Print

Analysis with Maxfilter

SSS or SSSt should always be applied if you recorded with MaxShield ON!

 You can process your data using the GUI interface (type "maxfilter_gui" in the command line on your linux station) 
 or directly as command lines with options (see at the bottom of this page, some available automatized scripts).

Maxfilter Options:

 Expansion Origin
   (1) default (0 0 40) <-> device origin 
   (2) fit to isotrak   <-> head coordinates

   It is preferable to use (1)in most circumstances:
   - fitting to isotrak data can only remove 1 point which is out of the inside sphere/bin. 
     For instance, when recording MEG+EEG, the nose reference, if digitized will be taken out of the sphere. 
     HOWEVER, if you digitize other points on the face for instance, these will still be considered part of your Bin.
   - algorithms have been optimized for these values
   - if you do decide to use option (2), do make a sanity check!

 Multiple Expansions Orders: <=> # of degrees of freedom used for SSS/SSSt
   default values are Lin = 8 and Lout = 3
   Rules of thumb:
     - Lin >> Lout
     - it is not recommended to change the default values unless you plan a very specific analysis
     - using bigger values may cause instability 
     - for further information, see SSS, SSSt

  Bad channel setup
     - that's where it is useful to know in advance which channels are bad by taking notes during acquisition and pre-screening with Graph! 
     - you can combine MANUAL bad channels and AUTOMATIC bad channel detection when running SSS
     - BUT in SSSt processing (i.e. MaxST), automatic bad channel detection will automatically be turned OFF
     - nbr of raw tags to scan: default is 60 <=> 60 blocks of data screened at the beginning and end of the data file
                                        if 1  <=> all blocks will be screened
     - variance used to set the threshold (default value of 7) is computed on every tag being screened 

 MaxST parameters
     - it is recommended to use 8 seconds for the length of the raw data buffer
     - a subspace correlation of 0.98 is fine (see detailed literature)

 You can run MaxFilter for a given subject on all the runs at once, using the following programs.

This program could only be used if you selected continuous HPI option during your MEG recording.

If continuous HPI monitoring is not available in your data

Even if you didn't select continuous HPI option during your recording, you can transform the head position measured at the beginning of each MEG recording session in a common head position across all your runs. This means that you will correct initial head position variability between your different runs. But of course, you will assume that your subject did not move his/her head during the run. So, this approach is mainly acceptable if your run doesn't last more than few minutes (3 to 5 min). This operation is imperative if you want to do averaging across several runs or compare different conditions!!!

Here is a global view of the approach when you don't have continuous HPI monitoring:

To do this, use the following command line:

show_fiff –vt 222 /neurospin/unicog/protocols/MEG/your protocol/nom_sujet/filename.fif

This command line will make possible to inspect accurately the head position. The first three rows indicate the rotation matrix R and the last row the translation vector T in the coordinate transformation. In brackets, you have the head position with respect to the device coordinates. Note this position for each of your runs.

Then,

Finally, to apply the head position transformation, sss correction and bad channel detection, use this command line:

/neurospin/unicog/protocols/MEG/your protocol/nom_sujet/filename.fif –origin 0 0 40 –frame head –bad xxxx(if you want to specify bad channels) -autobad 60 –badlimit 7 –trans /unicog/protocols/MEG/your protocol/nom_sujet/filename.fif (corresponding to your common head position that you have chosen)

If you don't want to apply common head coordinates across all you runs (not recommended), use this Simpler program : Source Space Separation, Bad channel detection