Dear Kambiz,

On Fri, Oct 23, 2009 at 12:44 AM, Kambiz <[log in to unmask]> wrote:
> Hello Vladimir,
> The headshape.mat file is not in the version of the multimodal_meg.tar.gz
> archive I downloaded from the fil ftp server. As a matter of fact I did
> notice an EOF error message using tar xvzf on that particular archive.
> Nevertheless what is contained in the tarball I downloaded seems to work
> fine, apart from the absent headshape file.

Yes, there is a problem with the FIL server we are aware of and
working to fix. I'm surprised you managed to get anything out of that
file. The file downloaded from the link I sent you yesterday should be
OK.

>
> I admit I am inching along with the tutorial steps, but in the manual file
> preparation is a preprocessing step and if I understand the recommended
> steps correctly that is when coregisteration occurs? Because by now I've
> managed to figure out how to make a headshape.mat file containing only the
> fiducial information, and successfully coregistered a converted MEG dataset
> such that the sensors are not  rotated with respect to the head in the
> reviewing tool. I assume there is nothing wrong with this approach, correct?
> Or is it the case where we are talking about different procedures during
> preparation and 3D source reconstruction, because, I am not really
> interested in inverse modeling my data in SPM at the moment. All I want to
> do with the software is sensor space analysis (and I have to admit the
> documentation on that is rather slim).
>

In the case of MEG, for sensor-level analysis you don't need to do
anything. Sensible 2D positions for the sensors will be generated
automatically at conversion. The reason for that is that the MEG
sensors are represented in head coordinates (coordinate system defined
relative to the head). So they can be projected quite well without
precise coregistration and this is done by default.

You only need coregistration for source analysis and DCM. Even then
you don't need the 'Prepare' step because it's only to combine
externally measured headshape with a MEG dataset. What you can do is
go directly to coregistration in the 3D window (with the tutorial
dataset as well) and then you'll just have the 3 fiducials without
headshape points and this is good enough if you can pick the
corresponding locations in the structural scan precisely. What you did
with generating headshape from fiducials is not necessary and doesn't
make any difference. You've  just replaced the existing fiducials with
the same thing. Just loading a headshape doesn't do coregistration.
You need to either go to the 'Coregister' menu in the 'Prepare' tool
or to '3D source reconstruction' window.

But as I said the whole previous paragraph is not relevant for you. If
after epoching and computing ERFs you see that topographical ('scalp')
plots in the reviewing tool and the corresponding scalp maps look
reasonable to you, you can go directly to 'Convert2images'.  Ignore
the fact that the 'Sensors' plot in the reviewing tool looks rotated.
It's just to remind people who really need to do coregistration that
they haven't done  it.

Best,

Vladimir

> Thanks
> Kambiz
>
> On Thu, Oct 22, 2009 at 2:54 PM, Vladimir Litvak <[log in to unmask]>
> wrote:
>>
>> Dear Kambiz,
>>
>> On Thu, Oct 22, 2009 at 6:58 PM, Kambiz Tavabi <[log in to unmask]> wrote:
>> > Dear SPMers,
>> > Could somebody kindly point me to where I can dl/see the structure of
>> > headshape.mat file used during preparation of converted meg file.
>>
>> This file comes with the tutorial dataset and within the archive file
>> multimodal_meg.tar.gz that you can download from
>> ftp://ftp.mrc-cbu.cam.ac.uk/personal/rik.henson/
>>
>> > Question: After converting my CTF meg files it appears the sensors are
>> > rotated 90 degrees around Z (up-down) axis. I have no digitized data for
>> > my
>> > data, and if i understand in SPM I can create headshape.mat and enter
>> > the
>> > subjects fiducial coordinates into shape.fid.pnt?
>>
>> That's not necessary if you haven't digitized the headshape outside the
>> MEG.
>>
>> > I am further guessing that
>> > this preparatory step is what aligns the meg coordinate system with
>> > spm/mni
>> > system, which is a necessary condition for creating spm images?
>>
>> You need to perform the coregistration step in 3D source
>> reconstruction interface (see the manual). You can do it using the
>> fiducials that have been recorded by the MEG and are in your dataset
>> after conversion.
>>
>> > Finally, why is this realignment/coregisteration step not occuring
>> > automatically for meg data that already contains the 3 necessary
>> > fiducials
>> > (lt/rt periauriculars and nasion)?
>>
>> Because not everybody uses these particular fiducials and even these
>> are names of the fiducials saved in the CTF dataset, some people just
>> place coils at completely different locations. SPM has no way of
>> knowing that. This is why you need to supply this information
>> explicitly. Before you coregister, the sensors will be rotated with
>> respect to the head in the reviewing tool. In any case the right way
>> to check the coregistartion is in the 3D source localization
>> interface. I suggest that you go through all the steps of the MEG part
>> of the multimodal tutorial and then things will become clearer.
>>
>> Best,
>>
>> Vladimir
>>
>> >
>> > thanks in advance,
>> > Kambiz
>> >
>
>
>
> --
> Kambiz
>