Dear Andy and other contributors,

you don't have an effect in the ventricles at the second level, if I  
understand correctly. You have this effect in 2 participants out of 27.

A possible explanation of your finding is that it is just noise with a  
specific spatial covariance, due to the different physical properties  
of the ventricles (as others have explained). If the signal in the  
ventricle voxels covaries, then you are bound to find this covariation  
in uncorrected maps -- i.e., you are bound to find voxels over the  
threshold at uncorrected levels, and these voxels follow the spatial  
covariance of this noise (the spatial covariance of t maps is the  
spatial correlation of the residuals). One way to check this is  
inspecting the RESELS maps, which should show low values in the  
ventricles. Another way to check this is to see if you have subjects  
with deactivations in the ventricles.

Best wishes,
Roberto Viviani
University of Ulm, Germany
University of Innsbruck, Austria

Quoting Andy Yeung <[log in to unmask]>:

> Dear all,
>
> Supplementing the above conversation, I've performed everything from Step 1
> again, but this time session 3 files are not involved.
> Result is the same, lateral ventricle activation...
>
> Best,
> Andy
>
> On Thu, Nov 20, 2014 at 4:06 PM, Andy Yeung <[log in to unmask]> wrote:
>
>> Dear all,
>>
>> My TR was 2000 ms.
>> So far I have 27 subjects, and I found 2 of them activating nearly only
>> lateral ventricles, and 2 of them activating nearly nothing (all checked at
>> voxel level uncorrected p < 0.05).
>> All 4 of them had spiky head movements at 1 or 2 of their sessions (3
>> sessions in total).
>> Other subjects are fine.
>>
>> I would still like to ask, regarding discarding the 3rd session, do you
>> recommend I:
>> 1) remove 3rd session files and re-do the whole pre-processing again;
>> 2) exclude them at specify 1st level;
>> 3) simply ignore 3rd session at defining contrast; or
>> 4) use stronger (?) software like ArtRepair besides SPM slice timing +
>> realignment?
>> Following this, how should I re-weight my contrast. As I give a [1 1 1]
>> for other subjects, if this subject only has 2 sessions, should the
>> contrast be [1.5 1.5]?
>>
>> Sincerely thanks,
>> Andy
>>
>> On Thu, Nov 20, 2014 at 10:43 AM, Hasan, Khader M <
>> [log in to unmask]> wrote:
>>
>>> Since CSF in the LV is expected to have a T1 on the order of 5000 ms, I
>>> would truncate the first runs with within 5*T1 or ~ 25,000 msec which may
>>> translate to 10 runs or so assuming TR ~ 2.5 ms. Thus depending on the TR
>>> "spin echo has longer TR", the 6 dummy scans may help attain full steady
>>> state for non-CSF, but for CSF I would still avoid the leading 5-10 scans
>>> from each run. Note also that activation of CSF can happen if oxygen
>>> dissolved in CSF is altered, thus my suggestion is based on the assumption
>>> that you wish to exclude CSF effects. In the analysis you may also modulate
>>> all results with a parenchyma or gray matter only mask. KH
>>>
>>>
>>>
>>>
>>> Khader M Hasan, PhD
>>> Associate Professor of Radiology
>>> MSE 168, Tel 713 500 7690 (FAX 713 500 7684)
>>> University of Texas Health Science Center at Houston
>>> Medical School
>>> Diagnostic and Interventional Imaging
>>> Magnetic Resonance Imaging Research Division
>>> Diffusion Tensor Imaging Lab,  Tel 713 500 7683
>>> http://www.uth.tmc.edu/radiology/faculty/khader-m-hasan/index.html<
>>> http://www.uth.tmc.edu/radiology/faculty/hasan.html>
>>> ________________________________
>>> From: SPM (Statistical Parametric Mapping) [[log in to unmask]] on
>>> behalf of Andy Yeung [[log in to unmask]]
>>> Sent: Wednesday, November 19, 2014 8:07 PM
>>> To: [log in to unmask]
>>> Subject: Re: [SPM] lateral ventricle activation
>>>
>>> Dear all,
>>>
>>> I'm using Philips 3T and has 6 dummy scans before actual experiment. And
>>> I discarded first 3 functional scans for each session.
>>> Do you feel this is enough or I need to discard more?
>>>
>>> Regarding discarding the 3rd session, do you recommend I:
>>> 1) remove 3rd session files and re-do the whole pre-processing again;
>>> 2) exclude them at specify 1st level; or
>>> 3) simply ignore 3rd session at defining contrast?
>>> Following this, how should I re-weight my contrast. As I give a [1 1 1]
>>> for other subjects, if this subject only has 2 sessions, should the
>>> contrast be [1.5 1.5]?
>>>
>>> Best,
>>> Andy
>>>
>>> On Wed, Nov 19, 2014 at 10:23 PM, Helmut Nebl <
>>> [log in to unmask]<mailto:
>>> [log in to unmask]>> wrote:
>>> As far as I can see there are three runs, and the third one seems to
>>> suffer from large motion. One option would be to just analyze the first two
>>> sessions. Do you still encounter this ventricle activation then?
>>>
>>> While discaring a session is imperfect, as the estimates are based on a
>>> lower number of sessions/trials for that subject, it is certainly better
>>> than taking into account a session in which problems are obvious (and jumps
>>> of several mm are definitely probelematic).
>>>
>>> Alternatively, you could also try to control for those volumes which are
>>> affacted by large motion. The easiest way is to add a dummy regressor for
>>> each of the bad volumes and probably also dummy regressors for each of the
>>> preceding and following volumes as artefacts might be present there as
>>> well. However, if there is lots of motion you will lose lots of degrees of
>>> freedom due to the many dummy regressors. Also make sure that motion is not
>>> associated with a particular condition. If the predicted time course of the
>>> dummy regressors fall together with onsets of a particular condition then
>>> the corresponding regressor can't be estimated properly of course.
>>>
>>> Best,
>>>
>>> Helmut
>>>
>>>
>>
>