[Mne_analysis] questions

[dgw]

Hi Jeff,

We are getting closer. Please make sure to send your emails to the list 
so that everyone can learn (and help)



Dan G. Wakeman wrote:
> 
> 
> Hi Matti,
> 
> I followed your instructions as follows:
> 
> Changed time scales to 4-253 ms to avoid flat segments at begin and end 
> of EEG
> Had already done 1-100 Hz bandpass filtering
> Marked VREF as bad
> Recalculated everything except electrode alignment
Does this mean you recalculated the noise covariance matrix and inverse 
solution?
> Loaded average and surfaces back into mne_analyze
> 
> When viewing inverse results on the inflated surface, they do not look 
> any different than before, that is:
> a. Source activations do not generally match scalp EEG
> b. For dSPM and sLORETA, most source activity is confined to R hemi
> 
> Questions:
> 1. It looks like I can examine the noise-cov matrix with Matlab – what 
> should I see, good or bad?
There are some examples (particularly on the webpage www.martinos.org/mne )
> 2. Alex G suggested I use ico5, not ico3. Other than higher spatial 
> resolution, that should not be cuasing the problems I see (?)
Well, it is theoretically possible. ico 3 is very bad for any 
orientation constraints (which i believe you indicated you were using in 
an earlier e-mail). It is ok, to user for loose/no orientation 
constraint, although very misleading. I would never recommend using 
anything less than ico 5, and even with ico 5 I recommend using the 
loose variable orientation constraints.
> 3. Alex also suggested I use a diagonal noise-cov matrix – how would I 
> go about that? If I knew what to put in it, I suppose I could create and 
> save using Matlab toolbox
there is the --diagnoise option in mne_do_inverse_operator
> 4. Is there any way to confirm that channel 257 has indeed been 
> eliminated in the averaging and covariance calculation?
Well, did you check the output when re-running the commands. Did it talk 
about the bad channels detected? It is also possible to check with the 
MATLAB toolbox.
> 5. To that end, I have not been able to view the average EEG – the 
> window is blank. Someone suggested I create a new layout, but why can't 
> the software see the layouts I already created for the raw EEG?
It should be able to see them, if you put them in the right directory as 
described in the manual. You should be able to select it using Adjust -> 
Select trace layout...

Are you using a noise covariance matrix?
What exact command did you use to generate it?
How does the time period you used to generate your noise covariance 
relate to the time period you are trying to invert?
Did you make sure that --projon was used when generating your noise 
covariance matrix?

> 6. I do not understand how to set the eegflat value - what should by 
> entered?
> 
> Thanks,
> -Jeff
> 
> From: Matti Hamalainen <msh at nmr.mgh.harvard.edu 
> <mailto:msh at nmr.mgh.harvard.edu>>
> To: Jeff Eriksen <eriksenj at ohsu.edu <mailto:eriksenj at ohsu.edu>>
> Cc: Dan G Wakeman <dwakeman at nmr.mgh.harvard.edu 
> <mailto:dwakeman at nmr.mgh.harvard.edu>>
> Subject: Re: noise covariance calculation
> 
> 
> On Jun 1, 2012, at 3:49 AM, Jeff Eriksen wrote:
> 
>> Matti,
>>
>> Here is my difficulty. I think I need to do something like this:
>>
>> mne_process_raw
>> --raw A6_ec1_fen_ds250_raw.fif
>> --projon
>> --savecovtag -cov
>> –cov A1_ec1_fen_ds250.cov
>>
>> But I do not have the last .cov file. The only example I have to go on 
>> it the audvis.cov supplied with the sample subject, but I cannot 
>> figure out how I would modify it to work with my continuous raw files.
>>
>> I believe I did manage to calculate the noise-cov for my first 
>> subject, but the .cov file I have for that subject is not a text file, 
>> and I have no idea now where it came from.
>>
>> Also, in the above, do I really need the —projon switch? I am not 
>> clear if that is needed to enforce the average reference for EEG data.
> 
> 
> Jeff,
> 
> It looks like you figured out how to compute the covariance. Here are a 
> few points:
> 
> 1. Change the time scale in both in your ave and cov files to go from, 
> e.g., 4 to 253 seconds because there are zeros in the beginning and end 
> of your raw data file.
> 2. I propose you use a highpass filter, e.g., at 0.5 Hz to get rid of 
> the dc offsets.
> 2. The last channel (VREF) in your data file is flat and it messes up 
> the estimates.  Mark it bad:
> 
> * Create a file called, e.g., VREF.bad which contains one line: VREF
> * mne_mark_bad_channels --bad VREF.bad A6_ec1_fen_ds250_raw.fif 
> * This produces the output 
> 
> mne_mark_bad_channels version 1.1 compiled at Jun  4 2012 05:32:20
> 
> Read bad channel list from: VREF.bad
> 
> 1 bad channels from VREF.bad:
>         <VREF> 
> 
> A6_ec1_fen_ds250_raw.fif ... [done]
> 
> * You can also mark bad channels interactively, see Manual sections 
> 11.4.1 and 4.9.
> * Recreate both the 'average' and the 'noise covariance'
> * Recreate the inverse operator
> 
> 3. You can specify either --projon or --projoff swich (after marking 
> VREF bad). The source estimates are always calculated with the average 
> reference on. 
> 
> After eliminating VREF and recalculating the noise covariance and the 
> inverse operator, you source estimates should be more reasonable.
> 
> - Matti
> 
> 
> 
> 
> ---------
> 
> Matti Hamalainen, Ph.D.
> Athinoula A. Martinos Center for Biomedical Imaging
> Massachusetts General Hospital
> 
> msh at nmr.mgh.harvard.edu <mailto:msh at nmr.mgh.harvard.edu>
> mhamalainen at partners.org <mailto:mhamalainen at partners.org>
> 
> 
> 
> 
> 
> 
> The information in this e-mail is intended only for the person to whom it is
> addressed. If you believe this e-mail was sent to you in error and the 
> e-mail
> contains patient information, please contact the Partners Compliance 
> HelpLine at
> http://www.partners.org/complianceline . If the e-mail was sent to you 
> in error
> but does not contain patient information, please contact the sender and 
> properly
> dispose of the e-mail.
>