CC to SPM JISCMail.
OK. So I'll try to use only 'current' and 1 iteration even if I have 1
epoch. Than I'll run stats (connectivity, rPDC) on such time-series and
see how it compares with 'standard' MSP.
Thanks again for the feedback,
PR
On 09.09.2015 18:05, Barnes, Gareth wrote:
> Hi Paolo
> Just a quick reply:
> The iteration is ineffective in 'current' as there is just a single set of priors which it uses all the time.
> I'm afraid that doing it by hand- using the .M- sounds like the only option at the moment.
> Best
> Gareth
>
>
>
> -----Original Message-----
> From: Paolo Ranzi [mailto:[log in to unmask]]
> Sent: 09 September 2015 16:36
> To: Barnes, Gareth <[log in to unmask]>
> Subject: Re: problems with MSP iterative 'Classic'
>
> Thanks for the kind reply.
>
> I thought the opposite: 'classic' more data-driven (not rescaling), whereas 'current' it corrects itself by the trials or conditions present. But remember I have only one epoch.
>
> 'current' seems to work producing meaningful time-series. It also produce nice .nii images. But the iteration seems to be ineffective.
> There is a Warning message saying more or less that the number of iterations will not be considered for such option.
>
> I've also forgot to mention that 'classic' works perfectly when generating a .nii image (1 .nii image each 5 secs). Instead when I'm trying to extract the source by "Source extraction" module, it fails.
> Instead with 'current', the "Source extraction" module works perfectly.
>
> Further, I'm using matlabbatch. Therefore trying to extract sources manually from D.inv{1}}.inverse.M is not that easy. I'd rather prefer to use the already implemented "Source extraction" module ... :-)
>
> Lastly, what does it mean 'Merge source estimates for multiple inversions'? No exhaustive info are found in SPM batch.
>
> Thanks for your help,
> PR
>
> On 09.09.2015 16:16, Barnes, Gareth wrote:
>> HI Paolo
>> The main difference between the 'current' and 'classic' implementations is that 'current' is more general. So if that works I would stick with it.
>> You can extract time series by using the M matrix in
>> D.inv{1}}.inverse.M M links the data Y to the sources J. You can apply
>> M to any epoch of Y.
>> Does that help or am I missing the point ?
>> Best
>> Gareth
>>
>>
>>
>> -----Original Message-----
>> From: Paolo Ranzi [mailto:[log in to unmask]]
>> Sent: 09 September 2015 15:06
>> To: Barnes, Gareth <[log in to unmask]>
>> Subject: problems with MSP iterative 'Classic'
>>
>> Dear Gareth Barnes,
>> I'm trying to use MSP + 'Source inversion, iterative'. Regarding the 'Inversion function' the option 'Current' is able to produce 13 time-series of 13 EEG sources. Unfortunately the option 'Classic' fails to extract the sources' time-series. Even setting 'iterations' to '1' it fails.
>> I didn't want to use 'Current' since I've single-subject 5 secs epochs.
>> Each epoch is source reconstructed separately and then they get merged by custom-made EEGlab script.
>> Not sure if the 'Merge source estimates for multiple inversions' should also be used in combination with 'Classic' and before extracting the time-series.
>>
>> I've checked the manual and internet for template pipeline (e.g. present in DAiSS), but I couldn't find any. I'm not using priors ('Priors image'
>> left empty). I'm using SPM12 v. 6475.
>>
>> Bests,
>> Paolo Ranzi
>>
|