Dear Haiteng,
I looked at the coarse mesh and I don't see that the solution is
radically different but it is slightly different which could be
expected. As you probably know MEG dipole patterns are very sensitive
to orientation of sources, more than to their location. If the mesh is
coarse then the activations can be localized to the nearest mesh node
with the suitable orientation which might be quite far from the true
location. One thing that is slightly problematic is that for some
combinations of mesh and head model there is a source close to the
center of the head, which is unlikely to be a true source. We'll look
at why this source appears.
What I suggest you to do if you are interested in differences between
conditions is to find a combination of mesh and head model for which
you have reasonable localization results and then use those to look at
experimental effects. What you should understand is that no matter how
hard we try EEG and MEG will not get close to fMRI in terms of their
spatial precision and reliability of localization. My opinion is that
the purpose of 3D source reconstruction is not to find precise true
locations of brain sources but to unmix experimental effects
observable at the sensor level and assign them in a very rough way to
some brain areas. If you look at it this way it's OK to choose out of
possible parameters those that give you the most reasonable
localizations. If however, precise source locations is all you are
interested in, you should do fMRI or dipole fitting to early sensory
ERPs or ERFs.
Perhaps just to be on the safe side, send me some pictures of what you
get for different settings and what you call 'different' because my
version of the code is different from yours.
Best,
Vladimir
2009/7/28 hiten <[log in to unmask]>:
> Dear Vladimir,
> Thanks for your response.I have another request.Would you please try the
> "coarse" mesh using 'Single shell' again? Comparing to the 'normal' mesh,the
> active brain regions are quite different. It is very stange and don't know
> why? Our concern is activation of brain regions map and I evalute the
> results by the activation map.I summarize the 3D reconstruction results ,eg
> time band 300-400ms, frequency band of interest 0 HZ.would you please spend
> some time on this issue again? Thanks very much!
> best ,
>
>
>
> --
> haiteng jiang
> Research Center for Learning Science,
> Southeast University
> Si Pai Lou 2 # , Nanjing, 210096, P.R.China
> Brain Imaging Lab
> Email: [log in to unmask]
>
> 在2009-07-28 00:39:16,"Vladimir Litvak" <[log in to unmask]> 写道:
>>Dear Haiteng,
>>
>>I looked at your data and there is nothing fundamentally wrong with
>>it. As a rule of thumb I'd say use 'Single shell' if you have
>>individual anatomy and 'Single sphere' otherwise. In your case both
>>options lead to solutions somewhere in the ventral visual pathway
>>which is what one would expect, however with single sphere the highest
>>activation is more posterior. I only tried the 'normal' mesh. If you
>>look at the log-evidence you can see that the highest evidence is
>>actually for single-sphere. This means that this is probably a better
>>option to use in your case. I don't think there is any bug here.
>>
>>Hope this helps.
>>
>>Vladimir
>>
>>2009/7/27 hiten <[log in to unmask]>:
>>> Dear Vladimir,
>>> Thanks for your time,They have been very helpful.Since the file is too big
>>> ,so I use another email([log in to unmask])to send you my data.I am very
>>> sorry for the resulting inconvenience. You can download the four files
>>> fowlling the two steps as the attachment described.
>>> The warnings don't arise when I choose 'single Sphere' or 'MEG local
>>> spere',and the results of 3D reconstruction using 'MEG local spere' seem
>>> to very be good. I am surprised that the results of 'single shell' are quite
>>> different between "normal" mesh and "coarse" mesh using standard imaging
>>> approach.The result of 'coarse' mesh make sense while "normal" mesh not.
>>> I summarize the results of inverse reconstruction as an image ,setting time
>>> band and specifying a frequency band of interest as zero.
>>> The test is incongruent color-words Stroop task by a a healthy female,.
>>> Every picture prensents 300 ms,ISI 3000ms,64 trials. I have pre-processed
>>> the MEG data as following,epoching pre-stimulus -200ms,post-stimulus 800
>>> ms;bandpass filter 5-30hz. Previous neuroimaging studies in healthy
>>> volunteers show that anterior cingulated cortex appears to be heavily
>>> involved int the task.
>>> It would be very grateful if you spend some time.Looking forward to your
>>> answer.Thanks again!
>>> best,
>>> haiteng
>>>
>>> 在2009-07-27 16:35:07,"Vladimir Litvak" <[log in to unmask]> 写道:
>>>>Dear Haiteng,
>>>>
>>>>Those warnings are normal and are not the reason why your results do
>>>>not make sense. I think if you choose 'Single sphere' model there will
>>>>be no warnings but the results should not be very different. Imaging
>>>>source reconstruction even with very sophisticated methods used in SPM
>>>>is not guaranteed to give you the right answer. You can try to improve
>>>>the results by limiting the time window to the part you are interested
>>>>in and/or filtering your data in a narrower band around the
>>>>frequencies you expect to contain your physiological signals. Try
>>>>playing with different options in the 'Custom' button. If you don't
>>>>manage to make any progress, send us an example of your data and some
>>>>details about the paradigm and we'll look at at it. How long is your
>>>>ERP? Is it 2 min? If so I wouldn't be surprised 3D may fail.
>>>>
>>>>Best,
>>>>
>>>>Vladimir
>>>>
>>>>
>>>>
>>>>2009/7/27 hiten <[log in to unmask]>:
>>>>> Dear SPMers:
>>>>> After pre-processing ,I load MEEG dataset which is averaged to do 3D source
>>>>> reconstruction .I use the subject's sMRI,choose 'normal' mesh,select our
>>>>> used nasion and preauricular points as fiducials,then go to the Forward
>>>>> computation using single shell model ,following warning arises:
>>>>> using headmodel specified in the configuration
>>>>> using gradiometers specified in the configuration
>>>>> computing surface normals
>>>>> Warning: Matrix is close to singular or badly scaled.
>>>>> Results may be inaccurate. RCOND = 4.227452e-017.
>>>>>> In fieldtrip\private\meg_ini>getcoeffs at 94
>>>>> In fieldtrip\private\meg_ini at 36
>>>>> In fieldtrip\private\prepare_vol_sens at 270
>>>>> In fieldtrip\private\prepare_headmodel at 223
>>>>> In fieldtrip\private\headmodelplot at 205
>>>>> In ft_headmodelplot at 11
>>>>> In spm_eeg_inv_checkforward at 70
>>>>> In spm_eeg_inv_forward_ui at 48
>>>>> In spm_eeg_inv_imag_api>Forward_Callback at 87
>>>>> In spm_eeg_inv_imag_api at 53
>>>>> undoing the G3BR balancing
>>>>> undoing the G3BR balancing
>>>>> Foward model complete - thank you
>>>>> Ignoring that ,next I choose "Standard" inversion,produciing below
>>>>> warnings:
>>>>> computing surface normals
>>>>> Warning: Matrix is close to singular or badly scaled.
>>>>> Results may be inaccurate. RCOND = 4.206769e-017.
>>>>>> In forwinv\private\meg_ini>getcoeffs at 94
>>>>> In forwinv\private\meg_ini at 36
>>>>> In forwinv\private\prepare_vol_sens at 270
>>>>> In forwinv_prepare_vol_sens at 11
>>>>> In spm_eeg_lgainmat at 91
>>>>> In spm_eeg_invert at 115
>>>>> In spm_eeg_invert_ui at 95
>>>>> In spm_eeg_inv_imag_api>Inverse_Callback at 94
>>>>> In spm_eeg_inv_imag_api at 53
>>>>> NB :According to my test paradigm, the inverse result can not make sense at
>>>>> all.
>>>>> Any suggestion? Thanks in advance!
>>>>>
>>>>>
>>>>> --
>>>>> haiteng jiang
>>>>> Research Center for Learning Science,
>>>>> Southeast University
>>>>> Si Pai Lou 2 # , Nanjing, 210096, P.R.China
>>>>> Brain Imaging Lab
>>>>> Email: jianghaiteng@126com
>>>>>
>>>>>
>>>>> ________________________________
>>>>> 网易YEAH.NET免费邮箱:您的终身免费邮箱
>>>
>>>
>>> ________________________________
>>> 网易YEAH.NET免费邮箱:您的终身免费邮箱
>
>
> ________________________________
> 网易YEAH.NET免费邮箱:您的终身免费邮箱
|