As, I said it doesn't make sense at all to overlay sensors on source MIP images. Not sure why you get the option in the pop-up menu at all, but we can look into that. If you are using MSP it has some symmetric source priors that might affect your results. You can try removing them by commenting out lines 608-610 of spm_eeg_invert or also try the other inversion types - IID and COH.

Best,

Vladimir

On Thu, Nov 27, 2014 at 4:29 PM, Thomas Cope <[log in to unmask]> wrote:

I guess, but it is still a concern that when you ask it to overlay channels they are overlaid with left-right reversal. It also provides the MNI space co-ordinates of the sensor it offers, and it gives me right sided co-ordinates for left sided sensors.

 

Essentially, I am looking at a language task and in sensor space the responses are strongly left lateralised as you might expect. In source space, the reconstruction is strikingly bilateral, with slightly stronger responses on the right – this concerns me.

 

 

From: Vladimir Litvak [mailto:[log in to unmask]]
Sent: 27 November 2014 16:09
To: Thomas Cope
Cc: [log in to unmask]
Subject: Re: [SPM] MEG source reconstructions - channel overlay problems

 

Dear Thomas,


What you are trying to do doesn't make much sense. ‘goto nearest suprathreshold channel’ makes sense for sensor-level images. For source images you are not supposed to look at things in channel space but in source space (e.g. go to the nearest suprathreshold voxel), that's the whole purpose of source reconstruction.

 

Best,

 

Vladimir

 

 

 

On Thu, Nov 27, 2014 at 3:41 PM, Thomas Cope <[log in to unmask]> wrote:

Dear (MEG) imagers,

 

I have come across a problem that I am sure reflects a simple misunderstanding on my behalf, but I can’t for the life of me figure it out so I was hoping someone on the list could tell me my error. Everything is done in SPM12, and the source reconstructions have been done using the GUI to do reconstructions based on combined Gradiometer/Magnetometer/EEG data.

 

Essentially, I made my source reconstructions and I wanted to look at channel power using the ‘goto nearest suprathreshold channel’ function. I noticed that I was being consistently given what I knew were left sided channels for right sided sources and vice versa. Overlaying the channels on the reconstruction (fig 1: http://s1165.photobucket.com/user/thomas_cope/media/fig1_Reconstruction_zps789fd86a.jpg.html ) shows that the whole EEG array is mirrored (arrows point out EEG18, which is on the left, and a set of right sided (positive x) MNI co-ordinates). The MEG arrays are similarly mirrored. Initially I thought that it must be a problem with radiological vs neurological convention of display, but the MNI co-ordinates should be robust to this. I then thought that the co-registration must be reversed but (fig 2: http://s1165.photobucket.com/user/thomas_cope/media/fig2_Co-registration_zps276a749e.jpg.html ) it seems that it is not. Finally I thought it might be an error in the rather opaque ‘group reconstruction’ routine, so I ran an individual reconstruction (fig 3: http://s1165.photobucket.com/user/thomas_cope/media/fig3_Individual_reconstruction_zps72cffb38.jpg.html ) and found the same result.

 

I am sure that I am making a simple beginner’s error, but I can’t figure it out. Can anyone help?

 

Best wishes, and thanks in advance,

 

Thomas