>To vary dipole moments you don't need a fitting routine. That's what
>you get for free when you do that subspace projection. Try the
>attached code. It seems to work sensibly. It'll ask you for dipole
>labels but they are not actually used so you can put anything there.
Dear Vladimir,
I tried the code and got the following output:
ans.inv{1} =
mesh: [1x1 struct]
comment: 'my comment'
date: [2x11 char]
datareg: [1x1 struct]
forward: [1x1 struct]
I am not sure, where should I look for the dipole moments?
I also have a couple of related questions. Sorry if they reflect my limited level of understanding.
1. The D.inv{1}.inverse.jmni{1} includes values that describe both orientations and amplitudes of the dipoles. However, when asked to provide moment priors should one provide amplitude-normalized [to the vector length] values?
2. I have noticed that when fitting the [symmetrical] dipole pair I end up with exactly symmetrical locations but not exactly left-to-right-symmetrical orientations. Why is it?
On the other hand, with the symmetrical dipole pair the program asks for only one-side orientation priors.
Elena
________________________________________
From: Vladimir Litvak [[log in to unmask]]
Sent: Tuesday, September 04, 2012 5:57 PM
To: Elena Orekhova
Cc: [log in to unmask]
Subject: Re: [SPM] dipole fit question
To vary dipole moments you don't need a fitting routine. That's what
you get for free when you do that subspace projection. Try the
attached code. It seems to work sensibly. It'll ask you for dipole
labels but they are not actually used so you can put anything there.
Best,
Vladimir
On Tue, Sep 4, 2012 at 4:04 PM, Elena Orekhova
<[log in to unmask]> wrote:
> Dear Vladimir,
> Thank you for your explanations.
>
> We have quite noisy (VERP) data. However, the grand average is satisfactory model by a pair
> of symmetrical dipoles (RV<0.05). Our idea was to fix locations and moments, but vary amplitudes of the
> left and right dipoles to obtain best approximation of the individual data, and then compare
> dipole amplitudes between clinical groups.
>
> Would it work in this case to give the spm_eeg_inv_vbecd_gui location and moment priors with very low prior location and moment variance?
>
> Elena
>
>
>
> ________________________________________
> From: Vladimir Litvak [[log in to unmask]]
> Sent: Monday, September 03, 2012 5:59 PM
> To: Elena Orekhova
> Cc: [log in to unmask]
> Subject: Re: [SPM] dipole fit question
>
> Dear Elena,
>
> It sounds like you are not actually fitting dipoles but just want to
> calculate the RV per subject for a dipole solution that you already
> have. There is no simple solution to that in SPM but it should be
> quite simple to do it in a few lines of Matlab code if you have the
> dipole lead fields and the data. Basically you need to look at the
> difference between the original data and its projection to the
> subspace of the dipole lead fields. If you want a GUI this is
> available in BESA where you can load a pre-computed dipole solution
> with different data and get RV and other measures.
>
> Vladimir
>
> On Mon, Aug 27, 2012 at 11:38 AM, Elena Orekhova
> <[log in to unmask]> wrote:
>> Dear M/EEG SPM community,
>>
>>
>> I would like to fit dipoles obtained from the group average to the
>> individual EEG data.
>>
>> The dipoles should have fixed position and orientation (no location or
>> moment variance).
>>
>> I would also like to obtain goodness of fit and residual variance for the
>> fitted model.
>>
>> What is the best way to do it?
>>
>>
>>
>> Elena
>>
>>
>>
|