Dear Amir,
On Thu, Jan 14, 2010 at 10:35 AM, Amir H Javadi <[log in to unmask]> wrote:
> Dear Vladimir :-)
> Thanks. It was my mistake submitting TF data to 3D SR routine. The factor
> contributing to this is that the routine doesn't return any error message
> mentioning wrong data entry.
I can't check for all possible mistakes a user can make. What I do is
once a user actually makes a mistake I can check for I add the check.
It's now done for your case.
> Now I have a couple of question on running 3D SR repeatedly a the same data.
> This routine, group conversion, performs a couple of more calculations for
> the first time, such as "checking for prevous inversions", "forward model".
> In addition to these, after creating the image files, it also writes the
> result of function "spm_eeg_inv_results" to the relevant data files. Also it
> generates some files beginning with "SPMgainmatrix_". I need to run the 3D
> SR for the same data files repeatedly for different time intervals and
> different frequency bands. Should I consider anything? Such as keeping a
> copy of the main data files and run the procedure on the original files all
> the time and delete "SPMgainmatrix_*" files? or the saved data doesn't have
> anything to do with the time intervals and frequency bands and those are
> just subject dependent? Concerning frequency bands, I should say that I want
> to pre-filter data and then send the filtered data to group conversion to
> extract source of different frequency activities.
SPMgainmatrix_*" files indeed only depend on the electrode locations
so there is no reason to remove them when you change time or frequency
interval. What you can do is set up the head model in your unfiltered
file and replicate it (by pressing the 'new' button) according to the
number of time intervals you need. Then save the dataset. It will be
your master dataset. Now create filtered datasets (you can do
spm_eeg_copy to rename the master dataset first if you want to keep
them all in the same directory) and run spm_eeg_invert from a script
with inputs D and val where D is a cell array of D objects for all the
input (filtered) files and val is an index for the time interval. If
everything works correctly group inversion will run automatically once
you specify the parameters.
See also the header of spm_eeg_invert.
Best,
Vladimir
>
> 2010/1/14 Vladimir Litvak <[log in to unmask]>
>>
>> Dear Amir,
>>
>> 3D source reconstruction in SPM is presently only implemented for time
>> domain data. So even if you want to reconstruct induced activity you
>> should provide time-domain data (i.e. epoched dataset) and
>> time-frequency analysis will be done as part of creating the images.
>> There is no way to use an SPM TF dataset as an input to source
>> reconstruction. In principle it is possible to do source
>> reconstruction in frequency domain, indeed using both power and phase.
>> However, still you wouldn't be able to do baseline correction of power
>> before source reconstruction because it is a non-linear step that
>> invalidates the assumptions of source inversion algorithms.
>>
>> Best,
>>
>> Vladimir
>>
>> On Thu, Jan 14, 2010 at 8:18 AM, Amir H Javadi <[log in to unmask]>
>> wrote:
>> > Dear Vladimir :-)
>> > Thanks for your prompt reply. Is it impossible cause we might have
>> > negative
>> > data in our Time-Frequency data, in the sense of 3D SR? Or there is
>> > something else that I'm missing?
>> > Now, one more question. How important is phase data in 3D SR from TF
>> > data?
>> > What if we correct the negative values by applying a phase inversion?
>> >
>> > Have a good day :-)
>> > Amir
>> >
>> >
>> >
>> >
>> > 2010/1/14 Vladimir Litvak <[log in to unmask]>
>> >>
>> >> Dear Amir,
>> >>
>> >> Since the first option is mathematically impossible, you have no
>> >> choice but to use the second one.
>> >>
>> >> Vladimir
>> >>
>> >> On Thu, Jan 14, 2010 at 8:06 AM, Amir H Javadi <[log in to unmask]>
>> >> wrote:
>> >> > Hi :-)
>> >> > I wonder which approach is more accurate in terms of extracting
>> >> > active
>> >> > brain
>> >> > areas due to a specific task; Baseline correcting the time-frequency
>> >> > data
>> >> > first and then 3D Source Reconstruction OR 3D Source Reconstruction
>> >> > of
>> >> > both
>> >> > Data and Baseline and then subtracting the two images?
>> >> > Thanks :-)
>> >> >
>> >> > Have a good time
>> >> > Amir
>> >> >
>> >> >
>> >
>> >
>
>
|