Creating/Importing Forward field

From NUTMEGwiki

Compute within NUTMEG

The lead field is the forward model; i.e., a model of what the sensors would record given a source at a known location and orientation on the voxel grid. NUTMEG has several options for computing the forward field for MEG data. Use of 3rd order gradient correction as available from CTF data is allowed.

• A volume of interest (VOI) must be defined (in addition to the coregistration of this volume to the MEG sensors). See VOI section. Note that manual VOI selection in NUTMEG is no longer necessary if you have loaded a spatially normalized MRI. NUTMEG will automatically generate a whole-head VOI based on the MNI template brain during the lead field computation. If you prefer a different VOI or do not have a normalized MRI for some reason, you may still manualy create one as before from the "Manual VOI" menu. Also note that you may need to add the /nutmeg/templates directory to your Matlab path if you get an error message that Matlab is unable to load the MNIVoxel.mat file.

Select the sensors you want in the "View/Select MEG Channels" window (or skip to include all sensors).

• Single sphere

In nut_compute_lead_field, set lfcomp=2 for 2component in the theta/phi directions (since radial is zero). Or, lfcomp=3 for 3component xyz but rank 2 (though can cause problems in some places...careful). This lead field algorithm uses a single-shell spherical head model, and is based on Sarvas (1987).

• Multiple sphere centers

Uses a single sphere modeled for each sensor relative to local curvature of head; thus you get as many spheres as sensors, but for each sensor, it's as if the whole model is a single-sphere model (not mutli-layer concentric spheres like for EEG). CTF software computes the location of the multiple spheres, with the localSpheres command. We plan to eventually compute these local sphere centers within NUTMEG as well, not requiring CTF software. If multiple local spheres exist, NUTMEG will automatically compute this option.

• Fill in the desired voxel size (default 5 mm voxels are appropriate in most cases) in the text box next to the button Compute Lead Field, and finally click on the button itself. This will compute the lead field for the voxels within the current VOI, at the specified grid sampling, and for the selected sensors only. This will take a couple minutes. You may also call nut_compute_leadfield from command line.
• Select Save Session and name the file according to your favorite convention (this is the afore-mentioned session.mat), in order to save lead field with this session. Something short is recommended with a subject identifier and experiment identifier, e.g., SSDmotorRD2.mat.

• For TFBF purposes: create a stripped version of the session file for use with the qsub scripts later. On the Matlab command line, type nut_liposession sessionfile. The output will be named "sessionfileLp.mat". (Yes the name of this program is terrible, but it works on so many levels. Okay, just two levels.)

Compute outside NUTMEG and import in

Both of the below options happen automatically when you click 'Obtain Lead Field' on the main GUI:

• If you have loaded Neuromag MEG data:

NUTMEG now has easy interface to load in a previously computed lead field in either MNE software (MGH) or Brainstorm.

• If you have loaded BrainProducts EEG data:

NUTMEG will call Fieldtrip's code to compute a 3-sphere lead field.

• Other tie-ins are possible upon request! (We soon hope to allow for importing BEM).