You can use the WHU atlas implemented in FSLview. We did threshold the maps and transformed them into individual space then (see Antonenko et al 2013 NeuroImage). Best, Daria
Mit freundlichen Grüßen
Daria Antonenko
-------------------------------------------------------------------------
Dr. Daria Antonenko
Charité – Universitätsmedizin Berlin
NeuroCure Clinical Research Center NCRC
AG Kognitive Neurologie / Wiss. Mitarbeiterin
Charitéplatz 1, D-10117 Berlin
Tel +49 30 450 660 223
Interne Besuchsadresse: CCM, Sauerbruchweg 5, E2, R 024
www.neurocure.de
www.ncrc.de
Bitte schützen Sie unsere Umwelt. Drucken Sie diese E-Mail nur, wenn unbedingt notwendig. Work sustainably - Please consider environment before printing.
-----Ursprüngliche Nachricht-----
Von: FSL - FMRIB's Software Library [mailto:[log in to unmask]] Im Auftrag von FSL automatic digest system
Gesendet: Freitag, 30. Januar 2015 01:00
An: [log in to unmask]
Betreff: FSL Digest - 28 Jan 2015 to 29 Jan 2015 (#2015-32)
There are 27 messages totaling 6752 lines in this issue.
Topics of the day:
1. design matrix for TBSS: can I use rows filled with zeros? (2)
2. ERROR: empty mask image
3. applywarp --rel and --abs options
4. paired ttest
5. Repeated measures ANOVA 2x3 (3)
6. GLM gui w/ Covariates (2)
7. Use of randomise : -m <mask_image> (2)
8. filtered_func_data file in 1st and 2nd FEAT
9. remove nonzero voxels outside head of MNI152_T1_2mm
10. bedpostX - intensities of merged_th<i>samples and merged_ph<i>samples
11. Warning: Data demeaning selected, but at least one design column has
non-zero mean (5)
12. mean FA of superior longitudinal fasciculus
13. Melodic output (melodic_IC.nii.gz) for Dual_regression input : type of
file ?
14. Oxford_asl : single delay pCASL and PVC
15. large change in waytotal values
16. Problem using fix to train classifier
17. Install to /usr/local
18. Vertex Analysis: Minimum ROI Volume?
----------------------------------------------------------------------
Date: Thu, 29 Jan 2015 00:27:15 +0000
From: Fergus Kane <[log in to unmask]>
Subject: Re: design matrix for TBSS: can I use rows filled with zeros?
Dear Anderson,
Philip's question was conveniently very close to one I was about to ask. I just wanted to clarify that I understand the answer, especially the exchangeability bit:
(A little background, which I might later bring to another thread for people's thoughts. I have a multiple group design using twins, some of whom are patients, some of whom are unaffected relatives and others of whom are controls (and a mixture of identical and non identical twin pairs). However, due to my study being a twin study, and due to problems of non-independence, I have decided to run an initial analysis between just probands (affected individuals) and controls (one from each twin pair). This is to ascertain diagnosis specific differences. It's all a little dodgy, but I can't identify an ideal analysis method for my dataset.)
As a result, I've collated all my participants images into one file, but only want to include some of them for the initial two group analysis. There's over 100 people, but I'll just see if I understand here with 10 people, of whom I wish to exclude three from group 2.
So I think my Design Matrix should look like this:?
/NumWaves 3
/NumPoints 10
/PPheights 1 1 1
/Matrix
1 0 0
1 0 0
1 0 0
1 0 0
1 0 0
0 1 0
0 1 0
0 0 1
0 0 1
0 0 1
With a contrast file like this:
/ContrastName1 group A > group B
/ContrastName2 group B > group A
/ContrastName3 group A mean
/ContrastName4 group B mean
/NumWaves 3
/NumContrasts 4
/PPheights 1 1 1 1
/RequiredEffect 11.914 11.914 6.368 10.069
/Matrix
1.000000e+00 -1.000000e+00 0.000000e+00
-1.000000e+00 1.000000e+00 0.000000e+00
1.000000e+00 0.000000e+00 0.000000e+00
0.000000e+00 1.000000e+00 0.000000e+00
1 -1 0
-1 1 0
1 0 0
0 1 0
And finally a design.grp like this:
1
1
1
1
1
1
1
1
2
3
4
Does that make sense or am I missing something (FSL newbie here)?
Also, I generated that contrasts file with Glm, but I don't understand where the "/required effect" values come from.. could you help explain?
Thanks for your help!
------------------------------
Date: Thu, 29 Jan 2015 13:04:29 +0800
From: zhang mingxia <[log in to unmask]>
Subject: Re: ERROR: empty mask image
Hi Stephen,
Thanks again! I need to make sure I understand you correctly because I
tried as how I understond, but res4d.nii.gz also contained negative values.
First, did you mean to demean the nuisance variables and regress out the
"demeaned" nuisance variables? Second, does "demean" mean taking off the
mean value? If I understand you correctly, it seems this method can not
produce a "positive" res4d.nii.gz.
Mingxia
On Wed, Jan 21, 2015 at 5:12 PM, Stephen Smith <[log in to unmask]> wrote:
> Hi - you can regress out demeaned nuisance variables without having to
> demean the data....I *think* this would work ok via fsl_glm (you would need
> to pre-demean the confounds).
>
> Cheers.
>
>
>
> On 21 Jan 2015, at 09:08, zhang mingxia <[log in to unmask]>
> wrote:
>
> Hi, Stephen, thanks again. Acturally, I am analysizing the resting state
> data. I regressed out some nuisance covariates and used the res4d.nii.gz
> which contain negative value for subsequent analysis. Any suggestion? I had
> an idea of adding a constant value to the data which may not influence the
> result from subsequent correlation analysis.
>
> Mingxia
>
> On Tue, Jan 20, 2015 at 8:30 PM, Stephen Smith <[log in to unmask]>
> wrote:
>
>> Hi - well I would talk to whoever provided the data. Correctly
>> converted (magnitude) imaging data should not in general contain negative
>> values...
>> Cheers.
>>
>>
>>
>> On 20 Jan 2015, at 05:31, zhang mingxia <[log in to unmask]>
>> wrote:
>>
>> Stephen, thanks for your answer! Yes, I also think the easiest way is to
>> change the input data. But how can i convert the input data without
>> influence on the subsequent results?
>>
>> Mingxia
>>
>> On Fri, Jan 16, 2015 at 4:44 PM, Stephen Smith <[log in to unmask]>
>> wrote:
>>
>>> Hi - yes many analyses in general will expect the data to be >0 - I
>>> would recommend fixing that in the input data before running anything else
>>> - that's probably the best and easiest way forwards.
>>> Cheers.
>>>
>>>
>>> On 14 Jan 2015, at 03:27, zhang mingxia <[log in to unmask]>
>>> wrote:
>>>
>>> Dear FSL experts,
>>>
>>>
>>> I met an error in the first level analysis (stats+poststats). I knew it
>>> caused by unsuccessful mask making. The reason is the input 4D data is
>>> not >0, which was required by the fslmaths -bin for making mask(this
>>> commond "/opt/fmritools/fsl/bin/fslmaths filtered_func_data -Tmin -bin
>>> mask -odt char" can not been successfully executed).
>>>
>>> I can certainly make a mask and complete the whole analysis by a lot of
>>> commonds. However, because I always analysize data by .fsf files saved from
>>> GUI, which is easier and with less chance of error. I wonder whether I can
>>> resolve this isuue on the GUI. Or any simple way to resolve the problem is
>>> appreciated!
>>>
>>> Thanks a lot!
>>>
>>> Mingxia Zhang
>>>
>>>
>>>
>>>
>>>
>>>
>>> ---------------------------------------------------------------------------
>>> Stephen M. Smith, Professor of Biomedical Engineering
>>> Associate Director, Oxford University FMRIB Centre
>>>
>>> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>>> +44 (0) 1865 222726 (fax 222717)
>>> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>>>
>>> ---------------------------------------------------------------------------
>>>
>>> Stop the cultural destruction of Tibet <http://smithinks.net/>
>>>
>>>
>>>
>>>
>>>
>>
>>
>>
>> ---------------------------------------------------------------------------
>> Stephen M. Smith, Professor of Biomedical Engineering
>> Associate Director, Oxford University FMRIB Centre
>>
>> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>> +44 (0) 1865 222726 (fax 222717)
>> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>>
>> ---------------------------------------------------------------------------
>>
>> Stop the cultural destruction of Tibet <http://smithinks.net/>
>>
>>
>>
>>
>>
>
>
> ---------------------------------------------------------------------------
> Stephen M. Smith, Professor of Biomedical Engineering
> Associate Director, Oxford University FMRIB Centre
>
> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
> +44 (0) 1865 222726 (fax 222717)
> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
> ---------------------------------------------------------------------------
>
> Stop the cultural destruction of Tibet <http://smithinks.net>
>
>
>
>
>
------------------------------
Date: Thu, 29 Jan 2015 09:09:23 +0000
From: paul mccarthy <[log in to unmask]>
Subject: Re: applywarp --rel and --abs options
Hi Arnold,
It turns out that if your warp field was generated by fnirt, applwarp will
detect this, and will always treat the warp file as relative (i.e. it will
ignore the '--abs' flag), as fnirt always outputs relative rather than
absolute dispacement.
When fnirt saves the warp file, it sets the NIFTI header intent code to a
specific value (2006) - this is how applywarp is able to detect that the
file contains relative displacements.
If, for whatever reason, you want to interpret a relative displacement warp
file as if it contains absolute displacements, then you can simply clear
the NIFTI intent code (set it to 0) using fsledithd. Applywarp will then
honour your use of the '--abs' flag, and will subsequently produce garbage
output :)
Cheers,
Paul
On Wed, Jan 28, 2015 at 8:31 PM, Arnold Evia <[log in to unmask]> wrote:
> Thanks for the reply Paul.
>
> I have uploaded the files used in the following commands:
>
> applywarp --abs -v -i HemiQ3in.nii -r HemiQ3ref.nii -w test.nii.gz -o
> checkabs.nii
> applywarp --rel -v -i HemiQ3in.nii -r HemiQ3ref.nii -w test.nii.gz -o
> checkrel.nii
>
> Note that test.nii.gz is a relative warpfield.
>
> -Arnold
>
>
------------------------------
Date: Thu, 29 Jan 2015 09:33:48 +0000
From: "Anderson M. Winkler" <[log in to unmask]>
Subject: Re: paired ttest
Hi Mayte,
This design is fine on what concerns running dual_regression and randomise.
There may be something else going on there. Try perhaps with a small set of
subjects first, making sure you follow the steps in the manual, etc.
That said, I thought from earlier emails that each subject had been scanned
under both conditions (H and E), so it sounds like a paired t-test would be
needed.
All the best,
Anderson
2015-01-27 15:43 GMT+00:00 Mayte Parada, Ms. <[log in to unmask]>:
> Hi Anderson,
>
> After creating my design.mat and design.con files using FEAT I've run
> them through dual_regression and I don't see the full complement of files
> that I'm supposed to. I'm wondering if the design files are incorrect?
>
>
> *Design.con file looks like this. I have two conditions in my 1 scan
> (humour vs erotic) and two groups of subjects (men & women) so I've tried
> to indicate the contrasts here:*
> /ContrastName1 W>M_Humour
> /ContrastName2 M>W_Humour
> /ContrastName3 W>M_Erotic
> /ContrastName4 M>W_Erotic
> /ContrastName5 H_mean
> /ContrastName6 E_mean
> /NumWaves 4
> /NumContrasts 6
> /PPheights 1.000000e+00 1.000000e+00 1.000000e+00 1.000000e+00
> 5.000000e-01 5.000000e-01
> /RequiredEffect 3.758 3.758 3.758 3.758 1.879 1.879
>
> /Matrix
> 1.000000e+00 -1.000000e+00 0.000000e+00 0.000000e+00
> -1.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00
> 0.000000e+00 0.000000e+00 1.000000e+00 -1.000000e+00
> 0.000000e+00 0.000000e+00 -1.000000e+00 1.000000e+00
> 1.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00
> 0.000000e+00 0.000000e+00 1.000000e+00 1.000000e+00
>
>
>
> *Design.mat file looks like this. Here I try to indicate who are the
> women and who are the men by using the cope files that I got from the first
> level feat in the Humour mean (H_mean) and Erotic mean (E_mean) from above.*
> /NumWaves 4
> /NumPoints 20
> /PPheights 1.000000e+00 1.000000e+00 1.000000e+00 1.000000e+00
>
> /Matrix
> 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
> 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
> 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
> 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
> 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
> 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00
> 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00
> 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00
> 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00
> 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00
> 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00
> 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00
> 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00
> 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00
> 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00
> 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00
> 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00
> 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00
> 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00
> 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00
>
>
>
>
>
>
>
>
> --
> Mayte Parada Ph.D.
> ____________________________
> Laboratory for the Biopsychosocial Study of Sexuality
> McGill University
> 1205 Docteur Penfield
> Montreal QC
> Canada H3A 1B1
> [log in to unmask]
> (514) 398-5323
>
> On Jan 22, 2015, at 10:23 AM, Mayte Parada, Ms. <[log in to unmask]>
> wrote:
>
> Thanks Anderson,
>
> I completely understand now. Found the cope images in the stats folders
> and will input the ones that represent each video type. Then I will
> identify men and women into the EVs for the higher level analysis and
> contrast them.
>
> Thanks!
> --
> Mayte Parada Ph.D.
> ____________________________
> Laboratory for the Biopsychosocial Study of Sexuality
> McGill University
> 1205 Docteur Penfield
> Montreal QC
> Canada H3A 1B1
> [log in to unmask]
> (514) 398-5323
>
> On Jan 21, 2015, at 4:45 PM, Anderson M. Winkler <[log in to unmask]>
> wrote:
>
> Hi Mayte,
>
> You can do a 2nd level between erotic and humour, even if these two
> conditions are in the same run, as long as they were coded at the lower
> level with separate EVs, which would give one COPE for each of these two
> conditions. These can then be compared at the 2nd level with a paired
> t-test.
>
> Alternatively, compare these two conditions still within the 1st level,
> then in the 2nd level select these FEAT directories for a comparison men
> vs. women. Note that for the effect of sex, even if the groups happen to be
> matched, it's a simple 2-sample t-test that needs to be done, i.e., not a
> paired t-test.
>
> All the best,
>
> Anderson
>
>
> 2015-01-21 20:06 GMT+00:00 Mayte Parada, Ms. <[log in to unmask]>:
>
>> Thanks Anderson,
>>
>> Yes I can do a higher level analysis between men and women using their
>> FEAT directories. That's fairly straight forward.
>>
>> Its when I try to distinguish between men and women, and erotic movie
>> vs humour movie....since both movies are played in the same run. I don't have
>> separate FEAT outputs for the movie types to enter for each participant. I
>> don't know how to generate those if each participant has two
>> conditions/movies within the same run.
>>
>> --
>> Mayte Parada Ph.D.
>> ____________________________
>> Laboratory for the Biopsychosocial Study of Sexuality
>> McGill University
>> 1205 Docteur Penfield
>> Montreal QC
>> Canada H3A 1B1
>> [log in to unmask]
>> (514) 398-5323
>>
>> On Jan 21, 2015, at 10:36 AM, Anderson M. Winkler <
>> [log in to unmask]> wrote:
>>
>> Hi Mayte,
>>
>> Not sure if I follow, but it's fine to do a paired t-test on the outputs
>> from 1st level FEAT. You wouldn't enter the same information twice, though.
>> Perhaps the best choice would be if you selected the option "Inputs are 3D
>> cope images from FEAT directories".
>>
>> All the best,
>>
>> Anderson
>>
>>
>> 2015-01-20 23:23 GMT+00:00 Mayte Parada, Ms. <[log in to unmask]>:
>>
>>> Hi experts,
>>>
>>> I'm trying to run a paired ttest and I want to make sure I understand
>>> the wiki properly.
>>>
>>> I have one run/trial/scan (the terms are a bit confusing so I want to be
>>> clear) and there are two conditions within the trial (humour vs erotic). I
>>> already did a lower level feat for each subject in the experiment and
>>> gathered the individual feat outputs (10 in total). I want to compare
>>> activation in the humour condition with the erotic condition for those
>>> subjects. I am now using higher-level analysis to do a paired test based on
>>> the wiki. I've entered the lower level feat files twice to get 20 input
>>> files and made an EV for each of the participants and an extra EV for
>>> comparing the trials.
>>> For some reason this doesn't feel right.
>>>
>>> my EVs and contrasts look like the example provided in the wiki but I'm
>>> not 100% about the feat files i'm entering. I had 10 participants but I'm
>>> comparing them with themselves for a paired test so I have to enter the
>>> feat files 2x correct?
>>>
>>>
>>>
>>> --
>>> M
>>>
>>>
>>
>>
>
>
>
------------------------------
Date: Thu, 29 Jan 2015 09:44:42 +0000
From: "Anderson M. Winkler" <[log in to unmask]>
Subject: Re: Repeated measures ANOVA 2x3
Hi Pavel,
Please, see attached an example design. Note that each subject is allocated
on its own exchangeability block (EB).
- For contrasts C1-C3, use the option "-e design.grp" in randomise.
- For contrasts C4-C6, and for the F-test, use the option "-e design.grp"
and also the option "--permuteBlocks" in randomise.
You may need to split the contrast file (*.con) into two, one with just
C1-C3, another with just C4-C6, then make the necessary changes to the
*.fts file (F-contrast), and run randomise twice (in separate directories
to avoid overwriting results).
All the best,
Anderson
2015-01-28 14:41 GMT+00:00 Pavel Hok <[log in to unmask]>:
> Dear FSL Experts,
>
> I know that similar questions have been already answered many times, but I
> could not find any satisfactory conclusions on the forum.
>
> I was trying to run 2x3 repeated measures ANOVA (i.e. 3 groups, 2
> timepoints, all subjects have complete data) in FEAT. Since this design is
> not mentioned in the GLM manual page, I tried to combine two other designs:
> the factors effect approach (
> http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#ANOVA:_2-factors_2-_.26_3-levels_.282x3_between-subjects_ANOVA.29)
> and mixed effects ANOVA (
> http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#ANOVA:_2-groups.2C_2-levels_per_subject_.282-way_Mixed_Effect_ANOVA.29
> ).
>
> Unfortunately, the Feat complains about rank deficiency. Is there a way to
> model such a study design? My second guess would be to get the
> within-subject contrasts using a FE analysis and then run simple T-test. My
> original design matrix is attached.
>
> Thanks for any ideas!
>
> Kind regards
>
> Pavel
>
>
>
> --
>
> -------------------------------------------------------------------------------
> -- MUDr. Pavel Hok
>
> -------------------------------------------------------------------------------
> -- Laboratoř funkční magnetické rezonance
> -- Neurologická klinika
> -- Lékařská fakulta
> -- Univerzita Palackého v Olomouci
> -- Fakultní nemocnice Olomouc
>
> -------------------------------------------------------------------------------
> -- Laboratory of functional magnetic resonance imaging
> -- Department of Neurology
> -- Faculty of Medicine and Dentistry
> -- Palacky University Olomouc
> -- University Hospital Olomouc
> -- Czech Republic
>
> -------------------------------------------------------------------------------
> -- I.P. Pavlova 6, 775 20 Olomouc
> -- web: fmri.upol.cz
> -- tel.: +420 588 443 418
> -- e-mail: [log in to unmask]
>
> -------------------------------------------------------------------------------
>
------------------------------
Date: Thu, 29 Jan 2015 09:51:49 +0000
From: "Anderson M. Winkler" <[log in to unmask]>
Subject: Re: GLM gui w/ Covariates
Hi Jon,
This design isn't great. It isn't wrong, but I believe it's just not a very
good one.
The problem is the coding of age. I suggest replacing the age categories
for the actual subjects' ages. This should give gains in power -- both for
testing the effects of age itself, or to test the effect of the other
variables while having age as nuisance.
If you are worried that the relationship between age and diffusion
parameters could be non-linear, and wanted to have separate groups to
capture this potential effect, then you could still use the three age
groups, but then consider using separate EVs for each.
The remaining of the design is fine, as well as the call to randomise.
All the best,
Anderson
2015-01-28 21:55 GMT+00:00 Jon Siegel <[log in to unmask]>:
> Hi FSL Experts,
>
> I had a quick question about using the GLM gui and the randomise function
> with a set of our TBSS data. I read an email regarding different group
> sizes and I wasn't sure how to account for this, or what the correct answer
> is. Here's where I'm at, though. I have a 3 groups of subjects (40 young
> adults, 36 middle-aged adults, and 66 older adults) that I've taken through
> the TBSS processing stream. One major question of interest that I have is
> whether there are any tracts that correlate with memory performance (ie.
> CVLT, WMS, Associative memory) after accounting for age effects (and other
> covariates like: Years of education, Gender, APOE4 status, etc). I'm
> curious whether I can address this using the GLM gui and the randomise
> command. Here is what I've done so far and I'd like to double-check to see
> whether this is valid or whether I need to change the way I'm doing things
> (ie. change the permutation blocks so that each age group is done
> separately instead of 1 entire group?):
>
> AgeVectorC CVLTc EducationC Intercept
> -0.18 13.94 -0.5 1
> -0.18 6.94 -0.5 1
> 0.82 -6.06 -1.5 1
> 0.82 -2.06 1.5 1
> -1.18 8.94 -1.5 1
> -1.18 -8.06 5.5 1
>
> To clarify-- This was just 6 participants from the design matrix. Age
> Vector/Group is -1 for young, 0 for middle, and 1 for old. All variables
> were centered using the entire sample (all 142 participants) and then I
> created the design matrix using these centered values.
>
> I'm interested in just those voxels that are significantly correlated with
> CVLT performance after accounting for age and years of education and so
> I've created my contrasts file to be as follows:
>
> AgeVectorC CVLTc EducationC Intercept
> C1: 0 1 0 0
> C2: 0 -1 0 0
>
> C1 would then tell me which voxels show a significant positive correlation
> with FA after accounting for age and years of education and C2 would then
> tell me which voxels show a significant negative correlation with FA after
> accounting for age and years of education????
>
> Here is the randomise command that I've used as well:
>
> #randomise -i all_FA_skeletonised -o
> AgeGroupCVLT_EducationAgeControlled_FA -m mean_FA_skeleton_mask -d
> AgeGroupCVLT_EducationAgeControlled.mat -t
> AgeGroupCVLT_EducationAgeControlled.con -n 5000 --T2 &
>
> Thank you in advance for your help. I know there are other ways to do the
> analyses (ie. splitting up the groups into separate EVs and looking for
> group differences and such), but I'm curious if this design and contrast
> file would be okay for looking at correlations within the entire sample?
>
> Best,
> Jon
>
------------------------------
Date: Thu, 29 Jan 2015 09:53:44 +0000
From: "Anderson M. Winkler" <[log in to unmask]>
Subject: Re: Use of randomise : -m <mask_image>
Hi Matthieu,
The voxels inside the mask and which have constant values across subjects
(that is, zero variance) will be ignored. Otherwise, if a voxel is inside
the mask, the test will be performed, if not, not.
All the best,
Anderson
2015-01-28 16:23 GMT+00:00 Matthieu Vanhoutte <[log in to unmask]>:
> Dear FSL's experts,
>
> I would like to use randomise to compute statistics on ICA maps.
>
> In this way, I was wondering what does the mask_image input have to look
> like comparatively to the 4D_input_data file ? I couldn't manage to get
> this information on
> http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/randomise/UserGuide.
>
> Best regards,
>
> -------------------------------------
> Matthieu Vanhoutte, MSc
> Research Engineer - Department of Neuroradiology
> Regional University Hospital, Lille, France
>
------------------------------
Date: Thu, 29 Jan 2015 09:54:58 +0000
From: "Anderson M. Winkler" <[log in to unmask]>
Subject: Re: design matrix for TBSS: can I use rows filled with zeros?
Hi Fergus,
We developed a tool that can be used to do permutation tests with twin
data, including sibships that have both twins and non-twins. If you'd like
to give it a try, send me an email off-list.
All the best,
Anderson
2015-01-29 0:27 GMT+00:00 Fergus Kane <[log in to unmask]>:
> Dear Anderson,
>
> Philip's question was conveniently very close to one I was about to ask.
> I just wanted to clarify that I understand the answer, especially the
> exchangeability bit:
>
> (A little background, which I might later bring to another thread for
> people's thoughts. I have a multiple group design using twins, some of
> whom are patients, some of whom are unaffected relatives and others of whom
> are controls (and a mixture of identical and non identical twin pairs).
> However, due to my study being a twin study, and due to problems of
> non-independence, I have decided to run an initial analysis between just
> probands (affected individuals) and controls (one from each twin pair).
> This is to ascertain diagnosis specific differences. It's all a little
> dodgy, but I can't identify an ideal analysis method for my dataset.)
>
> As a result, I've collated all my participants images into one file, but
> only want to include some of them for the initial two group analysis.
> There's over 100 people, but I'll just see if I understand here with 10
> people, of whom I wish to exclude three from group 2.
>
> So I think my Design Matrix should look like this:?
>
> /NumWaves 3
> /NumPoints 10
> /PPheights 1 1 1
>
> /Matrix
> 1 0 0
> 1 0 0
> 1 0 0
> 1 0 0
> 1 0 0
> 0 1 0
> 0 1 0
> 0 0 1
> 0 0 1
> 0 0 1
>
> With a contrast file like this:
>
> /ContrastName1 group A > group B
> /ContrastName2 group B > group A
> /ContrastName3 group A mean
> /ContrastName4 group B mean
> /NumWaves 3
> /NumContrasts 4
> /PPheights 1 1 1 1
> /RequiredEffect 11.914 11.914 6.368 10.069
>
> /Matrix
> 1.000000e+00 -1.000000e+00 0.000000e+00
> -1.000000e+00 1.000000e+00 0.000000e+00
> 1.000000e+00 0.000000e+00 0.000000e+00
> 0.000000e+00 1.000000e+00 0.000000e+00
>
> 1 -1 0
> -1 1 0
> 1 0 0
> 0 1 0
>
> And finally a design.grp like this:
>
> 1
> 1
> 1
> 1
> 1
> 1
> 1
> 1
> 2
> 3
> 4
>
> Does that make sense or am I missing something (FSL newbie here)?
>
> Also, I generated that contrasts file with Glm, but I don't understand
> where the "/required effect" values come from.. could you help explain?
>
> Thanks for your help!
>
------------------------------
Date: Thu, 29 Jan 2015 10:06:20 +0000
From: "Anderson M. Winkler" <[log in to unmask]>
Subject: Re: filtered_func_data file in 1st and 2nd FEAT
Hi Jie-Yu,
Please, see below:
2015-01-26 16:39 GMT+00:00 Jie Yu <[log in to unmask]>:
> Dear Expert
>
> I have 2 groups of participants each with fMRI data.
>
> I have 2 questions regarding the filtered_func_data file under 1st and 2nd
> level FEAT.
>
> (1)
>
> First, I performed 1st level FEAT for all of them.
>
> In the 1st level FEAT directory for each participant, there are many COPE
> directories, and under each COPE directory is a filtered_func_data file.
>
> *I think this filtered_func_data represents the time series of each
> participant?
>
Yes, and you can confirm this by opening the images and seeing how they
look like, and also using "fslinfo" that tells how many timepoints the file
has.
>
> (2)
>
> Then I used 2nd level whole-brain FEAT to find out if there is any
> significant group difference.
>
> However, there is not any significant group difference.
>
Maybe if there is nothing significant, there simply doesn't exist an
effect. Did you inspect the unthresholded maps as well?
>
> Then I am thinking of a ROI analysis instead of whole brain analysis. So I
> created a mask.
>
> and performed: fslmeants -m Mask -i
> group.feat/cope.feat/filtered_func_data -o Txt
>
> The values in this Txt file could be separated into 2 groups and followed
> by T-test to see if there is any group difference.
>
I think it doesn't sound a good idea to keep torturing the data until they
confess having committed the sin of significance :-)
If you really have a very good, strong, well justified hypothesis about an
ROI, you can either use the masking options in FEAT (see the tab "Post
stats"), or, as you did, extract the values for each subject with fslmeants
and run a separate test.
>
> *I think this time the filtered_func_data under group level feat's COPE
> directory represents the mean contrast estimate of each participant?
>
Yes, precisely.
All the best,
Anderson
------------------------------
Date: Thu, 29 Jan 2015 10:45:23 +0000
From: H van Steenbergen <[log in to unmask]>
Subject: Re: remove nonzero voxels outside head of MNI152_T1_2mm
Dear Paul,
Thanks a lot for your help. In the end I decided to manually draw a mask after doing some trick with fast and fslmaths.
Since more FSL-ers may like to use this mask I have uploaded these files on a server. Please find the links in this document: http://www.leidenuniv.nl/personal/steenbergenhvan1/docs/solutions/standard_brain_bg0/illustration_MNI_standard_brain_with_black_background.pdf
As illustrated in this document, the edited MNI standard brain allows you to make the background color of screenshots made in fslview transparant in MS Office while at the same time not affecting other “black” regions inside the head/brain.
Best, Henk
------------------------------
Date: Thu, 29 Jan 2015 10:48:52 +0000
From: Saad Jbabdi <[log in to unmask]>
Subject: Re: bedpostX - intensities of merged_th<i>samples and merged_ph<i>samples
Hi
Yes that is normal. The angles are not range-constrained, and only contribute to the model through their sine/cosine.
Cheers
Saad
On 28 Jan 2015, at 16:22, Laura pelizzari <[log in to unmask]<mailto:[log in to unmask]>> wrote:
Dear FSL experts,
I'm using BedpostX with all default options on data acquired with the standard Siemens 30 gradient directions.
I have a question about the intensities of merged_th<i>samples and merged_ph<i>samples that I obtain:
for example I obtain volumes with these ranges:
fslstats merged_ph1samples -R --> [-1164889030656.000000 1181305012224.000000]
fslstats merged_ph2samples -R --> [-1839787802624.000000 1774984364032.000000]
fslstats merged_th1samples -R --> [-653992984576.000000 553442082816.000000]
fslstats merged_th2samples -R --> [-558781890560.000000 684687949824.000000]
Why are the values so big? I would expect values in the range [0, 360] or [-360, 360] as, if I correctly understood, they should represent angles (samples from the distribution of an angle).
Are these values correct because of the periodic nature of the angles or is there somethig wrong in my preprocessing?
Thanks in advance.
Regards,
Laura
------------------------------------------------------------------
Saad Jbabdi, PhD
MRC Career Development Fellow &
University Research Lecturer
FMRIB Centre, University of Oxford,
John Radcliffe Hospital, Oxford, OX3 9DU, UK.
tel (+44)1865-222466 (fax 717)
www.ndcn.ox.ac.uk/team/researchers/saad-jbabdi<http://www.ndcn.ox.ac.uk/team/researchers/saad-jbabdi>
------------------------------
Date: Thu, 29 Jan 2015 12:32:24 +0100
From: Matthieu Vanhoutte <[log in to unmask]>
Subject: Re: Use of randomise : -m <mask_image>
Hi Anderson,
Thank you for helping, I better understand the type of this mask.
However, I wonder how can I compute such a mask and what does it look like
to ?
Best regards,
Matthieu Vanhoutte, MSc
Research Engineer - Neuroradiology Department
Regional University Hospital, Lille, France
Le 29 janv. 2015 11:21, "Anderson M. Winkler" <[log in to unmask]> a
écrit :
> Hi Matthieu,
>
> The voxels inside the mask and which have constant values across subjects
> (that is, zero variance) will be ignored. Otherwise, if a voxel is inside
> the mask, the test will be performed, if not, not.
>
> All the best,
>
> Anderson
>
>
>
> 2015-01-28 16:23 GMT+00:00 Matthieu Vanhoutte <[log in to unmask]
> >:
>
>> Dear FSL's experts,
>>
>> I would like to use randomise to compute statistics on ICA maps.
>>
>> In this way, I was wondering what does the mask_image input have to look
>> like comparatively to the 4D_input_data file ? I couldn't manage to get
>> this information on
>> http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/randomise/UserGuide.
>>
>> Best regards,
>>
>> -------------------------------------
>> Matthieu Vanhoutte, MSc
>> Research Engineer - Department of Neuroradiology
>> Regional University Hospital, Lille, France
>>
>
>
------------------------------
Date: Thu, 29 Jan 2015 11:55:28 +0000
From: Matthew Webster <[log in to unmask]>
Subject: Re: Warning: Data demeaning selected, but at least one design column has non-zero mean
Hi Rosalia,
If the global mean isn't represented in your design ( e.g. with a column of 1's ) then it is probably safest to demean all your EVs. If the global mean is present in your design, then you can safely remove the -D option from your randomise call...
Kind Regards
Matthew
> Hi Mathew,
>
> I did not demean any EV in my design. I am testing for correlations....is there any problem with this?
>
> Kind regards,
>
> Rosalia.
>
> 2015-01-28 14:41 GMT+01:00 Matthew Webster <[log in to unmask]>:
> Hello,
> Have you demeaned all of the columns in design.mat. If so, then you can safely ignore the message. If you haven't, then although randomise will have demeaned the columns for you, you may want to consider exactly what your design is testing.
>
> Kind regards
> Matthew
>
>
>> Dear FSL experts,
>>
>> Anyone of you could explain me this message:
>>
>> randomise -i all_FA_skeletonised -o tbss_FA_D -m mean_FA_skeleton_mask -d design.mat -t design.con -n 5000 --T2 -V -D
>> randomise options: -i all_FA_skeletonised -o tbss_FA_D -m mean_FA_skeleton_mask -d design.mat -t design.con -n 5000 --T2 -V -D
>> Loading Data:
>> Data loaded
>> Warning: Data demeaning selected, but at least one design column has non-zero mean - therefore invoking automatic demeaning of design matrix
>> 2.3865e+50 permutations required for exhaustive test of t-test 1
>>
>> Kind regards,
>>
>> Rosalia
>
>
------------------------------
Date: Thu, 29 Jan 2015 12:02:02 +0000
From: "Anderson M. Winkler" <[log in to unmask]>
Subject: Re: Warning: Data demeaning selected, but at least one design column has non-zero mean
Hi Rosalia,
Could you show your design matrix and contrasts?
Thanks
Anderson
2015-01-28 15:14 GMT+00:00 Rosalia Dacosta Aguayo <[log in to unmask]>:
> Hi Mathew,
>
> I did not demean any EV in my design. I am testing for correlations....is
> there any problem with this?
>
> Kind regards,
>
> Rosalia.
>
> 2015-01-28 14:41 GMT+01:00 Matthew Webster <[log in to unmask]>
> :
>
>> Hello,
>> Have you demeaned all of the columns in design.mat. If so, then
>> you can safely ignore the message. If you haven't, then although randomise
>> will have demeaned the columns for you, you may want to consider exactly
>> what your design is testing.
>>
>> Kind regards
>> Matthew
>>
>>
>> Dear FSL experts,
>>
>> Anyone of you could explain me this message:
>>
>> randomise -i all_FA_skeletonised -o tbss_FA_D -m mean_FA_skeleton_mask -d
>> design.mat -t design.con -n 5000 --T2 -V -D
>> randomise options: -i all_FA_skeletonised -o tbss_FA_D -m
>> mean_FA_skeleton_mask -d design.mat -t design.con -n 5000 --T2 -V -D
>> Loading Data:
>> Data loaded
>> *Warning: Data demeaning selected, but at least one design column has
>> non-zero mean* - therefore invoking automatic demeaning of design matrix
>> 2.3865e+50 permutations required for exhaustive test of t-test 1
>>
>> Kind regards,
>>
>> Rosalia
>>
>>
>>
>
------------------------------
Date: Thu, 29 Jan 2015 20:42:45 +0800
From: Rujing Zha <[log in to unmask]>
Subject: mean FA of superior longitudinal fasciculus
Hi all
I completed the FA calculation of whole brain, then I want to ROI analysis.
I want to test the mean FA difference of superior longitudinal fasciculus between two groups.
How to do that ?
Thanks
Best regards.
r
------------------------------
Date: Thu, 29 Jan 2015 13:46:32 +0100
From: Matthieu Vanhoutte <[log in to unmask]>
Subject: Re: Melodic output (melodic_IC.nii.gz) for Dual_regression input : type of file ?
Hi,
Thank you for this quick answer !
Best regards,
Matthieu
Le 28/01/2015 18:11, Stephen Smith a écrit :
> Hi
>
>> On 28 Jan 2015, at 16:16, Matthieu Vanhoutte
>> <[log in to unmask] <mailto:[log in to unmask]>> wrote:
>>
>> Dear FSL's experts,
>>
>> I got on a separate software the ICA-group components from one
>> RS-fMRI study.
>>
>> I would like to use from this point the Dual_regression FSL command
>> to compute the ICA subject-specific maps.
>>
>> In this way, what kind of file does "the melodic_IC.nii.gz" look like
>> ? Is this file a 4D file of spatial t-maps ?
>
> yes - or in our case more often "z" maps.
>
>> And if it is the case are these spatial t-maps thresholded and how ?
>
> we generally use unthresholded maps for that.
>
> Cheers.
>
>
>
>>
>> Best regards,
>>
>> -------------------------------------
>> Matthieu Vanhoutte, MSc
>> Research Engineer - Department of Neuroradiology
>> Regional University Hospital, Lille, France
>
>
> ---------------------------------------------------------------------------
> Stephen M. Smith, Professor of Biomedical Engineering
> Associate Director, Oxford University FMRIB Centre
>
> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
> +44 (0) 1865 222726 (fax 222717)
> [log in to unmask] <mailto:[log in to unmask]>
> http://www.fmrib.ox.ac.uk/~steve <http://www.fmrib.ox.ac.uk/%7Esteve>
> ---------------------------------------------------------------------------
>
> Stop the cultural destruction of Tibet <http://smithinks.net>
>
>
>
>
------------------------------
Date: Thu, 29 Jan 2015 13:48:08 +0100
From: Matthieu Vanhoutte <[log in to unmask]>
Subject: Re: Oxford_asl : single delay pCASL and PVC
Many thanks for helping !
I will try as soon as possible and give back my obtained results.
Best regards,
Matthieu
Le 27/01/2015 17:11, Michael Chappell a écrit :
> I think that the problem in this case is related to the mask used fro
> processing and oxford_asl was unable to create a sensible brain mask.
> It would help if you provided the control image (that is used for
> calibration and registration here) after brain extraction. My attempt
> using the default BET parameters achieved a sensible brain extraction
> for the control image and using this oxford_asl with PVC produced
> results.
>
> Michael
>
>
> On 27 Jan 2015, at 11:54, Matthieu Vanhoutte
> <[log in to unmask] <mailto:[log in to unmask]>> wrote:
>
>> Hello Michael,
>>
>> Thank you for helping.
>>
>> I hope you will find quickly the solution and I am waiting
>> impatiently for the newer version !
>>
>> Best regards,
>>
>> -------------------------------------
>> Matthieu Vanhoutte, MSc
>> Research Engineer - Department of Neuroradiology
>> Regional University Hospital, Lille, France
>> Le 27/01/2015 11:16, Michael Chappell a écrit :
>>> It looks like the main (non partial volume corrected version) of the
>>> CBF quantification has worked and something has gone wrong at the
>>> final PVC step. I have seen a similar problem before and I suspect
>>> it will have something to do with the mask or partial volume
>>> estimates created internally. I will try to have a look soon and get
>>> back to you - it may be that this has been fixed by some
>>> improvements to oxford_asl that have not been released yet (in which
>>> case I might be able to give you a newer version temporarily).
>>>
>>> Michael
>>>
>>> On 23 Jan 2015, at 16:07, Matthieu Vanhoutte
>>> <[log in to unmask] <mailto:[log in to unmask]>>
>>> wrote:
>>>
>>>> Hello experts,
>>>>
>>>> I remembered to use "bash" before "oxford_asl" command to prevent
>>>> from some bugs.
>>>>
>>>> So it went far away but still got lot of errors :
>>>>
>>>> OXFORD_ASL - running
>>>> Creating output directory
>>>> Saving results in natve (ASL aquisition) space to pvc2/native_space
>>>> Saving results in structural space to pvc2/struct_space
>>>> Pre-processing
>>>> Creating mask
>>>> Performing registration
>>>> ASL_REG
>>>> Input file is: control
>>>> Running FLIRT
>>>> Tidying up
>>>> ASL_REG - Done.
>>>> Segmenting the structural image
>>>> Calculating M0a - calling ASL_CALIB
>>>> ASL_CALIB
>>>> Tissue reference is: csf
>>>> Image Exception : #22 :: ERROR: Could not open image
>>>> /tmp/fsl_7D0Jbn_asl_calib/seg_bias
>>>> terminate called after throwing an instance of
>>>> 'RBD_COMMON::BaseException'
>>>> Aborted (core dumped)
>>>> Cannot open volume /tmp/fsl_7D0Jbn_asl_calib/biasfield for reading!
>>>> Image Exception : #22 :: ERROR: Could not open image
>>>> /tmp/fsl_7D0Jbn_asl_calib/sens
>>>> terminate called after throwing an instance of
>>>> 'RBD_COMMON::BaseException'
>>>> Aborted (core dumped)
>>>> M0:785.90172355273613219438
>>>> ASL_calib - DONE.
>>>> Number of inversion times: 1
>>>> Number of timepoints in data: 30
>>>> Number of repeats in data: 30
>>>> 3.175 3.175 3.175 3.175 3.175 3.175 3.175 3.175 3.175 3.175
>>>> 3.175 3.175 3.175 3.175 3.175 3.175 3.175 3.175 3.175 3.175
>>>> 3.175 3.175 3.175 3.175 3.175 3.175 3.175 3.175 3.175 3.175
>>>> Setting up BASIL
>>>> Calling BASIL on data
>>>> Creating output directory: /tmp/fsl_eIyKnR_ox_asl/basil
>>>> Model for analysis is: aslpvc
>>>> Using bolus length: 1.650
>>>> STEP 1: VB (2 param: CBF and ^t)
>>>> ------------------
>>>> Welcome to FABBER v2.0
>>>> Logfile started: /tmp/fsl_eIyKnR_ox_asl/basil/step1/logfile
>>>> Start time: Fri Jan 23 16:52:01 2015
>>>> FABBER release v2.0
>>>> Forward Model version:
>>>> $Id: fwdmodel_asl_pvc.cc,v 1.6 2013/09/03 15:08:04
>>>> chappell Exp $
>>>> Loading mask data from '/tmp/fsl_eIyKnR_ox_asl/mask'
>>>> Loading data from 'diffdata'
>>>> here
>>>> . . . . . . . . . . Saving!
>>>> FABBER is all done.
>>>> Duration: 90 seconds.
>>>> Logfile was: /tmp/fsl_eIyKnR_ox_asl/basil/step1/logfile
>>>> STEP 2: VB (4 param: CBF and ^t, fwm, ^twm) - init with STEP 1
>>>> FABBER: MVNtool
>>>> Reading MVNs from /tmp/fsl_eIyKnR_ox_asl/basil/temp
>>>> Reading MVNs from /tmp/fsl_eIyKnR_ox_asl/basil/temp
>>>> FABBER: MVNtool
>>>> Reading MVNs from /tmp/fsl_eIyKnR_ox_asl/basil/temp
>>>> Reading MVNs from /tmp/fsl_eIyKnR_ox_asl/basil/temp
>>>> Spatial prior specification: MNMNII
>>>> STEP 2: Spatial VB (4 param: CBF and ^t, fwm, ^twm) - init with
>>>> STEP 1
>>>> FABBER: MVNtool
>>>> Reading MVNs from /tmp/fsl_eIyKnR_ox_asl/basil/temp
>>>> Reading MVNs from /tmp/fsl_eIyKnR_ox_asl/basil/temp
>>>> FABBER: MVNtool
>>>> Reading MVNs from /tmp/fsl_eIyKnR_ox_asl/basil/temp
>>>> Reading MVNs from /tmp/fsl_eIyKnR_ox_asl/basil/temp
>>>> FABBER: MVNtool
>>>> Reading MVNs from /tmp/fsl_eIyKnR_ox_asl/basil/temp
>>>> Reading MVNs from /tmp/fsl_eIyKnR_ox_asl/basil/temp
>>>> ------------------
>>>> Welcome to FABBER v2.0
>>>> FABBER: MVNtool
>>>> Reading MVNs from /tmp/fsl_eIyKnR_ox_asl/basil/temp
>>>> Reading MVNs from /tmp/fsl_eIyKnR_ox_asl/basil/temp
>>>> ------------------
>>>> Welcome to FABBER v2.0
>>>> Logfile started: /tmp/fsl_eIyKnR_ox_asl/basil/step2/logfile
>>>> Start time: Fri Jan 23 16:53:39 2015
>>>> FABBER release v2.0
>>>> Forward Model version:
>>>> $Id: fwdmodel_asl_pvc.cc,v 1.6 2013/09/03 15:08:04
>>>> chappell Exp $
>>>> Expanded, --param-spatial-priors=MNMNII
>>>> Loading mask data from '/tmp/fsl_eIyKnR_ox_asl/mask'
>>>> Loading data from 'diffdata'
>>>> fabber: inference_spatialvb.cc:1982: void
>>>> SpatialVariationalBayes::CalcNeighbours(const
>>>> NEWIMAGE::volume<float>&): Assertion `checkNofN == 1' failed.
>>>> Aborted (core dumped)
>>>> End.
>>>> Cannot open volume
>>>> /tmp/fsl_eIyKnR_ox_asl/basil/step2/mean_ftiss for reading!
>>>> Cannot open volume
>>>> /tmp/fsl_eIyKnR_ox_asl/basil/step2/mean_delttiss for reading!
>>>> Cannot open volume /tmp/fsl_eIyKnR_ox_asl/basil/step2/mean_fwm
>>>> for reading!
>>>> Cannot open volume
>>>> /tmp/fsl_eIyKnR_ox_asl/basil/step2/mean_deltwm for reading!
>>>> Image Exception : #22 :: ERROR: Could not open image
>>>> /tmp/fsl_eIyKnR_ox_asl///ftiss
>>>> terminate called after throwing an instance of
>>>> 'RBD_COMMON::BaseException'
>>>> /home/global//fsl50/bin/oxford_asl: line 173: 8854
>>>> Aborted (core dumped) flirt -in
>>>> $tempdir/$subdir/$param -applyxfm -init $tempdir/asl2struct.mat
>>>> -ref $struc -out $outdir/struct_space/$subdir/$parname
>>>> Image Exception : #22 :: ERROR: Could not open image
>>>> /tmp/fsl_eIyKnR_ox_asl///delttiss
>>>> terminate called after throwing an instance of
>>>> 'RBD_COMMON::BaseException'
>>>> /home/global//fsl50/bin/oxford_asl: line 173: 8903
>>>> Aborted (core dumped) flirt -in
>>>> $tempdir/$subdir/$param -applyxfm -init $tempdir/asl2struct.mat
>>>> -ref $struc -out $outdir/struct_space/$subdir/$parname
>>>> Image Exception : #22 :: ERROR: Could not open image
>>>> /tmp/fsl_eIyKnR_ox_asl///ftisswm
>>>> terminate called after throwing an instance of
>>>> 'RBD_COMMON::BaseException'
>>>> /home/global//fsl50/bin/oxford_asl: line 173: 8952
>>>> Aborted (core dumped) flirt -in
>>>> $tempdir/$subdir/$param -applyxfm -init $tempdir/asl2struct.mat
>>>> -ref $struc -out $outdir/struct_space/$subdir/$parname
>>>> Image Exception : #22 :: ERROR: Could not open image
>>>> /tmp/fsl_eIyKnR_ox_asl///deltwm
>>>> terminate called after throwing an instance of
>>>> 'RBD_COMMON::BaseException'
>>>> /home/global//fsl50/bin/oxford_asl: line 173: 9001
>>>> Aborted (core dumped) flirt -in
>>>> $tempdir/$subdir/$param -applyxfm -init $tempdir/asl2struct.mat
>>>> -ref $struc -out $outdir/struct_space/$subdir/$parname
>>>> Cannot open volume pvc2/native_space///perfusion for reading!
>>>> Cannot open volume pvc2/native_space///perfusion_wm for reading!
>>>> Cannot open volume pvc2/native_space///arrival for reading!
>>>> Cannot open volume pvc2/native_space///deltwm for reading!
>>>> Cannot open volume pvc2/native_space///perfusion for reading!
>>>> Cannot open volume pvc2/struct_space///perfusion for reading!
>>>> Cannot open volume pvc2/native_space///perfusion_calib for reading!
>>>> Cannot open volume pvc2/native_space///perfusion_wm for reading!
>>>> Cannot open volume pvc2/struct_space///perfusion_wm for reading!
>>>> Cannot open volume pvc2/native_space///perfusion_wm_calib for
>>>> reading!
>>>> Output is pvc2/
>>>> OXFORD_ASL - done.
>>>>
>>>>
>>>> ==> End of this command, the output directory "native_space" is
>>>> incomplete (no perfusion and perfusion_calib) and "struct_space"
>>>> *is empty*.
>>>>
>>>> You can find attached to this link the files I have used with the
>>>> oxford_asl command :
>>>> http://we.tl/DXkTsFIQwX
>>>>
>>>> Could you please take a look at these and explain me why this
>>>> problem happened ?
>>>>
>>>> Many thanks in advance !
>>>>
>>>> Best regards,
>>>>
>>>> -------------------------------------
>>>> Matthieu Vanhoutte, MSc
>>>> Research Engineer - Department of Neuroradiology
>>>> Regional University Hospital, Lille, France
>>>>
>>>> 2015-01-23 16:43 GMT+01:00 Matthieu Vanhoutte
>>>> <[log in to unmask] <mailto:[log in to unmask]>>:
>>>>
>>>> Hello experts,
>>>>
>>>> I update this post and still can't generate output datas from
>>>> *oxford_asl* (with single delay pCASL data & PVC)**:
>>>>
>>>> /oxford_asl -i diffdata -o pvc2 --tis 3.175 --bolus 1.650
>>>> --casl -c control -s T1Proc_BrainSegmentation0N4 --tr 4.05 --te
>>>> 14 --regfrom control --pvcorr/
>>>>
>>>> This is the error message :
>>>>
>>>> /OXFORD_ASL - running
>>>> Creating output directory
>>>> Saving results in natve (ASL aquisition) space to pvc2/native_space
>>>> Saving results in structural space to pvc2/struct_space
>>>> Pre-processing
>>>> Creating mask
>>>> Performing registration
>>>> ASL_REG
>>>> Input file is: control
>>>> Running FLIRT
>>>> Tidying up
>>>> ASL_REG - Done.
>>>> Segmenting the structural image
>>>> Calculating M0a - calling ASL_CALIB
>>>> ASL_CALIB
>>>> Tissue reference is: csf
>>>> *Image Exception : #22 :: ERROR: Could not open image
>>>> /tmp/fsl_18mRY5_asl_calib/seg_bias
>>>> terminate called after throwing an instance of
>>>> 'RBD_COMMON::BaseException'
>>>> Aborted (core dumped)
>>>> Cannot open volume /tmp/fsl_18mRY5_asl_calib/biasfield for reading!
>>>> Image Exception : #22 :: ERROR: Could not open image
>>>> /tmp/fsl_18mRY5_asl_calib/sens
>>>> terminate called after throwing an instance of
>>>> 'RBD_COMMON::BaseException'
>>>> Aborted (core dumped)*
>>>> M0:785.90172355273613219438
>>>> ASL_calib - DONE.
>>>> */home/global//fsl50/bin/oxford_asl: 1060:
>>>> /home/global//fsl50/bin/oxford_asl: alltis[0]=3.175: not found*
>>>> Number of inversion times: 1
>>>> Number of timepoints in data: 30
>>>> Number of repeats in data: 30
>>>> */home/global//fsl50/bin/oxford_asl: 1074:
>>>> /home/global//fsl50/bin/oxford_asl: Syntax error: Bad for loop
>>>> variable*/
>>>>
>>>>
>>>> ==> End of this command, the output directories "native_space"
>>>> and "struct_space" are *empty*...
>>>>
>>>> You can find attached to this link the files I have used with
>>>> the oxford_asl command :
>>>> http://we.tl/DXkTsFIQwX
>>>>
>>>> Could you please take a look at these and explain me why this
>>>> problem happened ?
>>>>
>>>> Many thanks in advance !
>>>>
>>>> Best regards,
>>>>
>>>>
>>>> -------------------------------------
>>>> Matthieu Vanhoutte, MSc
>>>> Research Engineer - Department of Neuroradiology
>>>> Regional University Hospital, Lille, France
>>>>
>>>> 2014-12-03 20:32 GMT+01:00 Michael Chappell
>>>> <[log in to unmask]
>>>> <mailto:[log in to unmask]>>:
>>>>
>>>> That this error is internal to the software and not to do
>>>> with your data. However, it would not prevent the software
>>>> from giving you perfusion images and will not have
>>>> adversely affected these images.
>>>>
>>>> Michael
>>>>
>>>> On 2 Dec 2014, at 17:16, Matthieu Vanhoutte
>>>> <[log in to unmask]
>>>> <mailto:[log in to unmask]>> wrote:
>>>>
>>>>> So what should I do ?
>>>>> My data aren't good ? Or is this purely a software problem
>>>>> I can correct ? (to obtain perfusion_calib and pvcorr CBF
>>>>> file)
>>>>>
>>>>> Best,
>>>>>
>>>>> Matthieu
>>>>>
>>>>> -------------------------------------
>>>>> Matthieu Vanhoutte, MSc
>>>>> Research Engineer - Department of Neuroradiology
>>>>> Regional University Hospital, Lille, France
>>>>>
>>>>> 2014-12-02 18:04 GMT+01:00 Michael Chappell
>>>>> <[log in to unmask]
>>>>> <mailto:[log in to unmask]>>:
>>>>>
>>>>> This appears to be a tiny bug due to a missing ‘&’ in
>>>>> asl_calib. However, my testing suggests that it will
>>>>> not have affected anything that comes after that point.
>>>>>
>>>>> Michael
>>>>>
>>>>>
>>>>> On 26 Nov 2014, at 17:26, Matthieu Vanhoutte
>>>>> <[log in to unmask]
>>>>> <mailto:[log in to unmask]>> wrote:
>>>>>
>>>>>> Hello Michael,
>>>>>>
>>>>>> I have tried to run partial volume correction with
>>>>>> this command :
>>>>>> /bash oxford_asl -i diffdata -o pvc --tis 3.175
>>>>>> --bolus 1.650 --casl -c control -s
>>>>>> T1Proc_BrainExtractionBrain --tr 4.05 --te 14
>>>>>> --regfrom control --pvcorr/
>>>>>>
>>>>>> But I've got this error message on ASL_CALIB step :
>>>>>>
>>>>>> /OXFORD_ASL - running
>>>>>> Creating output directory
>>>>>> Saving results in natve (ASL aquisition) space to
>>>>>> pvc/native_space
>>>>>> Saving results in structural space to pvc/struct_space
>>>>>> Pre-processing
>>>>>> Creating mask
>>>>>> Performing registration
>>>>>> ASL_REG
>>>>>> Input file is: control
>>>>>> Running FLIRT
>>>>>> Tidying up
>>>>>> ASL_REG - Done.
>>>>>> Segmenting the structural image
>>>>>> Calculating M0a - calling ASL_CALIB
>>>>>> ASL_CALIB
>>>>>> Tissue reference is: csf
>>>>>> Image Exception : #22 :: ERROR: Could not open image
>>>>>> /tmp/fsl_3GAiyq_asl_calib/seg_bias
>>>>>> terminate called after throwing an instance of
>>>>>> 'RBD_COMMON::BaseException'
>>>>>> Aborted (core dumped)
>>>>>> Cannot open volume
>>>>>> /tmp/fsl_3GAiyq_asl_calib/biasfield for reading!
>>>>>> Image Exception : #22 :: ERROR: Could not open image
>>>>>> /tmp/fsl_3GAiyq_asl_calib/sens
>>>>>> terminate called after throwing an instance of
>>>>>> 'RBD_COMMON::BaseException'
>>>>>> Aborted (core dumped)
>>>>>> M0:785.72033545478675808106
>>>>>> ASL_calib - DONE./
>>>>>>
>>>>>> Where could it come from ?
>>>>>>
>>>>>> Best,
>>>>>>
>>>>>> -------------------------------------
>>>>>> Matthieu Vanhoutte, MSc
>>>>>> Research Engineer - Department of Neuroradiology
>>>>>> Regional University Hospital, Lille, France
>>>>>>
>>>>>> 2014-11-26 15:09 GMT+01:00 Matthieu Vanhoutte
>>>>>> <[log in to unmask]
>>>>>> <mailto:[log in to unmask]>>:
>>>>>>
>>>>>> Hello Michael,
>>>>>>
>>>>>> Thank you very much for your quick and detailed
>>>>>> answer.
>>>>>>
>>>>>> Concerning the partial volume correction, how to
>>>>>> estimate PV and apply PV correction : just by
>>>>>> typing --pvcorr in the oxford_asl command ?
>>>>>>
>>>>>> What if I supply my own PV estimates : how should
>>>>>> I use this with command line ?
>>>>>>
>>>>>> Thanks again !
>>>>>>
>>>>>> Best,
>>>>>>
>>>>>> -------------------------------------
>>>>>> Matthieu Vanhoutte, MSc
>>>>>> Research Engineer - Department of Neuroradiology
>>>>>> Regional University Hospital, Lille, France
>>>>>>
>>>>>> 2014-11-26 14:16 GMT+01:00 Michael Chappell
>>>>>> <[log in to unmask]
>>>>>> <mailto:[log in to unmask]>>:
>>>>>>
>>>>>> CBF is computed using the ‘standard’ model
>>>>>> for ASL, this is very nicely described in the
>>>>>> Buxton 1998 paper. It is marginally different
>>>>>> from the assumptions made in the recent white
>>>>>> paper - although if you set the T1 of tissue
>>>>>> equal to that of blood (=1.65 s) the you will
>>>>>> get a model that is very close to those
>>>>>> recommendations. The algorithm used to
>>>>>> process the data according to the ASL model
>>>>>> is the Chappell 2009 paper. By default
>>>>>> oxford_asl uses the CSF in the calibration
>>>>>> image (in your case control) to estimate the
>>>>>> M0 value needed for absolute quantification,
>>>>>> this also differs from the simple procedure
>>>>>> in the white paper.
>>>>>>
>>>>>> Partial volume correction is possible using
>>>>>> the --pvcorr command - by default oxford_asl
>>>>>> will use FAST to segment the structural image
>>>>>> and create the partial volume estimates You
>>>>>> can supply your own PV estimates (these need
>>>>>> to be at the same resolution as your ASL
>>>>>> data) if you want. This will try to create a
>>>>>> perfusion image of grey matter perfusion
>>>>>> (along with a separate white matter
>>>>>> estimate). This uses the method in the
>>>>>> Chappell 2011 paper.In principle this method
>>>>>> was originally designed for multi-delay data,
>>>>>> experience suggests it will operate on single
>>>>>> delay data like you are using (but do enter
>>>>>> your data without having taken the mean over
>>>>>> the individual repeats), but the separation
>>>>>> will not be as effective.
>>>>>>
>>>>>> Michael
>>>>>>
>>>>>> References:
>>>>>> Buxton 1998: Buxton, R. /et al./A general
>>>>>> kinetic model for quantitative perfusion
>>>>>> imaging with arterial spin labeling. /Magn
>>>>>> Reson Med/*40,*383–396 (1998).
>>>>>> White Paper: Alsop, D. C. /et al./Recommended
>>>>>> implementation of arterial spin-labeled
>>>>>> perfusion MRI for clinical applications: A
>>>>>> consensus of the ISMRM perfusion study group
>>>>>> and the European consortium for ASL in
>>>>>> dementia. /Magn Reson Med/(2014).
>>>>>> doi:10.1002/mrm.25197
>>>>>> Chappell 2009: Chappell, M. A., Groves, A.,
>>>>>> Whitcher, B. & Woolrich, M. Variational
>>>>>> Bayesian Inference for a Nonlinear Forward
>>>>>> Model. /IEEE Transactions on Signal
>>>>>> Processing/*57,*223–236 (2009).
>>>>>> Chappell 2011: Chappell, M. A. /et
>>>>>> al./Partial volume correction of multiple
>>>>>> inversion time arterial spin labeling MRI
>>>>>> data. /Magn Reson Med/*65,*1173–1183 (2011).
>>>>>>
>>>>>> On 25 Nov 2014, at 13:41, Matthieu Vanhoutte
>>>>>> <[log in to unmask]
>>>>>> <mailto:[log in to unmask]>> wrote:
>>>>>>
>>>>>>> Dear FSL's experts,
>>>>>>>
>>>>>>> 1) From my own pCASL data, I would like to
>>>>>>> compute the CBF map and according my
>>>>>>> parameters sequence I used this command :
>>>>>>> /bash oxford_asl -i diffdata -o asl --tis
>>>>>>> 3.175 --bolus 1.650 --casl -c control -s
>>>>>>> ExtractionBrain --tr 4.05 --te 14/
>>>>>>>
>>>>>>> Could you tell me the underlying used
>>>>>>> mathematical equations and the publications
>>>>>>> whose these equations came from ?
>>>>>>>
>>>>>>> 2) I saw since FSL 5.0.6 Partial Volume
>>>>>>> Correction is possible but I didn't found
>>>>>>> how to apply this ?
>>>>>>>
>>>>>>> Thank you in advance !
>>>>>>>
>>>>>>> Best regards,
>>>>>>>
>>>>>>> -------------------------------------
>>>>>>> Matthieu Vanhoutte, MSc
>>>>>>> Research Engineer - Department of
>>>>>>> Neuroradiology
>>>>>>> Regional University Hospital, Lille, France
>>>>>>
>>>>>> ---
>>>>>> Michael Chappell MEng DPhil
>>>>>> W: http://www.ibme.ox.ac.uk/qubic
>>>>>> T: +44 1865 617657
>>>>>> <tel:%2B44%201865%20617657>
>>>>>> Associate Professor, Institute of Biomedical
>>>>>> Engineering, University of Oxford.
>>>>>> http://www.ibme.ox.ac.uk
>>>>>> <http://www.ibme.ox.ac.uk/>
>>>>>> Director of Training, EPSRC-MRC CDT in
>>>>>> Biomedical Imaging
>>>>>> http://www.onbi.ox.ac.uk
>>>>>> <http://www.onbi.ox.ac.uk/>
>>>>>> Governing Body Fellow, Wolfson College, Oxford.
>>>>>> http://www.wolfson.ox.ac.uk
>>>>>> <http://www.wolfson.ox.ac.uk/>
>>>>>> Research Fellow, FMRIB Centre
>>>>>> http://www.fmrib.ox.ac.uk
>>>>>> <http://www.fmrib.ox.ac.uk/>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>> ---
>>>>> Michael Chappell MEng DPhil
>>>>> W: http://www.ibme.ox.ac.uk/qubic
>>>>> T: +44 1865 617657 <tel:%2B44%201865%20617657>
>>>>> Associate Professor, Institute of Biomedical
>>>>> Engineering, University of Oxford.
>>>>> http://www.ibme.ox.ac.uk <http://www.ibme.ox.ac.uk/>
>>>>> Director of Training, EPSRC-MRC CDT in Biomedical Imaging
>>>>> http://www.onbi.ox.ac.uk <http://www.onbi.ox.ac.uk/>
>>>>> Governing Body Fellow, Wolfson College, Oxford.
>>>>> http://www.wolfson.ox.ac.uk
>>>>> <http://www.wolfson.ox.ac.uk/>
>>>>> Research Fellow, FMRIB Centre
>>>>> http://www.fmrib.ox.ac.uk <http://www.fmrib.ox.ac.uk/>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>
>>> ---
>>> Michael Chappell MEng DPhil
>>> W: http://www.ibme.ox.ac.uk/qubic
>>> T: +44 1865 617657
>>> Associate Professor, Institute of Biomedical Engineering, University
>>> of Oxford.
>>> http://www.ibme.ox.ac.uk <http://www.ibme.ox.ac.uk/>
>>> Director of Training, EPSRC-MRC CDT in Biomedical Imaging
>>> http://www.onbi.ox.ac.uk <http://www.onbi.ox.ac.uk/>
>>> Governing Body Fellow, Wolfson College, Oxford.
>>> http://www.wolfson.ox.ac.uk <http://www.wolfson.ox.ac.uk/>
>>> Research Fellow, FMRIB Centre
>>> http://www.fmrib.ox.ac.uk <http://www.fmrib.ox.ac.uk/>
>>>
>>>
>>>
>>>
>>
>
> ---
> Michael Chappell MEng DPhil
> W: http://www.ibme.ox.ac.uk/qubic
> T: +44 1865 617657
> Associate Professor, Institute of Biomedical Engineering, University
> of Oxford.
> http://www.ibme.ox.ac.uk
> Director of Training, EPSRC-MRC CDT in Biomedical Imaging
> http://www.onbi.ox.ac.uk
> Governing Body Fellow, Wolfson College, Oxford.
> http://www.wolfson.ox.ac.uk
> Research Fellow, FMRIB Centre
> http://www.fmrib.ox.ac.uk
>
>
>
>
------------------------------
Date: Thu, 29 Jan 2015 14:22:02 +0100
From: Rosalia Dacosta Aguayo <[log in to unmask]>
Subject: Re: Warning: Data demeaning selected, but at least one design column has non-zero mean
Hi Anderson,
Now I am testing for correlations for two groups regarding HST....that is
demeaned....
Here it is,
Thanks again,
Rosalia.
2015-01-29 13:02 GMT+01:00 Anderson M. Winkler <[log in to unmask]>:
> Hi Rosalia,
> Could you show your design matrix and contrasts?
> Thanks
> Anderson
>
>
> 2015-01-28 15:14 GMT+00:00 Rosalia Dacosta Aguayo <[log in to unmask]>:
>
>> Hi Mathew,
>>
>> I did not demean any EV in my design. I am testing for correlations....is
>> there any problem with this?
>>
>> Kind regards,
>>
>> Rosalia.
>>
>> 2015-01-28 14:41 GMT+01:00 Matthew Webster <[log in to unmask]
>> >:
>>
>>> Hello,
>>> Have you demeaned all of the columns in design.mat. If so, then
>>> you can safely ignore the message. If you haven't, then although randomise
>>> will have demeaned the columns for you, you may want to consider exactly
>>> what your design is testing.
>>>
>>> Kind regards
>>> Matthew
>>>
>>>
>>> Dear FSL experts,
>>>
>>> Anyone of you could explain me this message:
>>>
>>> randomise -i all_FA_skeletonised -o tbss_FA_D -m mean_FA_skeleton_mask
>>> -d design.mat -t design.con -n 5000 --T2 -V -D
>>> randomise options: -i all_FA_skeletonised -o tbss_FA_D -m
>>> mean_FA_skeleton_mask -d design.mat -t design.con -n 5000 --T2 -V -D
>>> Loading Data:
>>> Data loaded
>>> *Warning: Data demeaning selected, but at least one design column has
>>> non-zero mean* - therefore invoking automatic demeaning of design matrix
>>> 2.3865e+50 permutations required for exhaustive test of t-test 1
>>>
>>> Kind regards,
>>>
>>> Rosalia
>>>
>>>
>>>
>>
>
------------------------------
Date: Thu, 29 Jan 2015 14:24:59 +0100
From: Rosalia Dacosta Aguayo <[log in to unmask]>
Subject: Re: Warning: Data demeaning selected, but at least one design column has non-zero mean
Hi Anderson,
I am worried. The problem I have:
I am not sure when I have to use -D and when not...the same for demeaning,
and if I use demeaning...should I use the option -D too?
I have mixed reports from my fellows and I do not have things clear....
Thank you,
Rosalia.
2015-01-29 14:22 GMT+01:00 Rosalia Dacosta Aguayo <[log in to unmask]>:
> Hi Anderson,
>
> Now I am testing for correlations for two groups regarding HST....that is
> demeaned....
>
> Here it is,
>
> Thanks again,
>
> Rosalia.
>
> 2015-01-29 13:02 GMT+01:00 Anderson M. Winkler <[log in to unmask]>:
>
>> Hi Rosalia,
>> Could you show your design matrix and contrasts?
>> Thanks
>> Anderson
>>
>>
>> 2015-01-28 15:14 GMT+00:00 Rosalia Dacosta Aguayo <[log in to unmask]>:
>>
>>> Hi Mathew,
>>>
>>> I did not demean any EV in my design. I am testing for
>>> correlations....is there any problem with this?
>>>
>>> Kind regards,
>>>
>>> Rosalia.
>>>
>>> 2015-01-28 14:41 GMT+01:00 Matthew Webster <
>>> [log in to unmask]>:
>>>
>>>> Hello,
>>>> Have you demeaned all of the columns in design.mat. If so,
>>>> then you can safely ignore the message. If you haven't, then although
>>>> randomise will have demeaned the columns for you, you may want to consider
>>>> exactly what your design is testing.
>>>>
>>>> Kind regards
>>>> Matthew
>>>>
>>>>
>>>> Dear FSL experts,
>>>>
>>>> Anyone of you could explain me this message:
>>>>
>>>> randomise -i all_FA_skeletonised -o tbss_FA_D -m mean_FA_skeleton_mask
>>>> -d design.mat -t design.con -n 5000 --T2 -V -D
>>>> randomise options: -i all_FA_skeletonised -o tbss_FA_D -m
>>>> mean_FA_skeleton_mask -d design.mat -t design.con -n 5000 --T2 -V -D
>>>> Loading Data:
>>>> Data loaded
>>>> *Warning: Data demeaning selected, but at least one design column has
>>>> non-zero mean* - therefore invoking automatic demeaning of design
>>>> matrix
>>>> 2.3865e+50 permutations required for exhaustive test of t-test 1
>>>>
>>>> Kind regards,
>>>>
>>>> Rosalia
>>>>
>>>>
>>>>
>>>
>>
>
------------------------------
Date: Thu, 29 Jan 2015 14:21:40 +0000
From: "Anderson M. Winkler" <[log in to unmask]>
Subject: Re: Warning: Data demeaning selected, but at least one design column has non-zero mean
Hi Rosalia,
Have a look at Jeanette Mumford's page on mean centering (you have seen
this already, no?): http://mumford.fmripower.org/mean_centering/
There is no need to use -D, so you can just remove the -D and all warnings
will disappear.
Anderson
2015-01-29 13:24 GMT+00:00 Rosalia Dacosta Aguayo <[log in to unmask]>:
> Hi Anderson,
>
> I am worried. The problem I have:
>
> I am not sure when I have to use -D and when not...the same for demeaning,
> and if I use demeaning...should I use the option -D too?
>
> I have mixed reports from my fellows and I do not have things clear....
>
> Thank you,
>
> Rosalia.
>
> 2015-01-29 14:22 GMT+01:00 Rosalia Dacosta Aguayo <[log in to unmask]>:
>
>> Hi Anderson,
>>
>> Now I am testing for correlations for two groups regarding HST....that is
>> demeaned....
>>
>> Here it is,
>>
>> Thanks again,
>>
>> Rosalia.
>>
>> 2015-01-29 13:02 GMT+01:00 Anderson M. Winkler <[log in to unmask]>:
>>
>>> Hi Rosalia,
>>> Could you show your design matrix and contrasts?
>>> Thanks
>>> Anderson
>>>
>>>
>>> 2015-01-28 15:14 GMT+00:00 Rosalia Dacosta Aguayo <[log in to unmask]>:
>>>
>>>> Hi Mathew,
>>>>
>>>> I did not demean any EV in my design. I am testing for
>>>> correlations....is there any problem with this?
>>>>
>>>> Kind regards,
>>>>
>>>> Rosalia.
>>>>
>>>> 2015-01-28 14:41 GMT+01:00 Matthew Webster <
>>>> [log in to unmask]>:
>>>>
>>>>> Hello,
>>>>> Have you demeaned all of the columns in design.mat. If so,
>>>>> then you can safely ignore the message. If you haven't, then although
>>>>> randomise will have demeaned the columns for you, you may want to consider
>>>>> exactly what your design is testing.
>>>>>
>>>>> Kind regards
>>>>> Matthew
>>>>>
>>>>>
>>>>> Dear FSL experts,
>>>>>
>>>>> Anyone of you could explain me this message:
>>>>>
>>>>> randomise -i all_FA_skeletonised -o tbss_FA_D -m mean_FA_skeleton_mask
>>>>> -d design.mat -t design.con -n 5000 --T2 -V -D
>>>>> randomise options: -i all_FA_skeletonised -o tbss_FA_D -m
>>>>> mean_FA_skeleton_mask -d design.mat -t design.con -n 5000 --T2 -V -D
>>>>> Loading Data:
>>>>> Data loaded
>>>>> *Warning: Data demeaning selected, but at least one design column has
>>>>> non-zero mean* - therefore invoking automatic demeaning of design
>>>>> matrix
>>>>> 2.3865e+50 permutations required for exhaustive test of t-test 1
>>>>>
>>>>> Kind regards,
>>>>>
>>>>> Rosalia
>>>>>
>>>>>
>>>>>
>>>>
>>>
>>
>
------------------------------
Date: Thu, 29 Jan 2015 16:19:36 +0100
From: Pavel Hok <[log in to unmask]>
Subject: Re: Repeated measures ANOVA 2x3
Hi Anderson,
thanks, will this work with FEAT? It used to complain when two or more
non-zero values in one EV had different group.
Cheers
Pavel
On 29 January 2015 at 10:44, Anderson M. Winkler <[log in to unmask]>
wrote:
> Hi Pavel,
>
> Please, see attached an example design. Note that each subject is
> allocated on its own exchangeability block (EB).
>
> - For contrasts C1-C3, use the option "-e design.grp" in randomise.
> - For contrasts C4-C6, and for the F-test, use the option "-e design.grp"
> and also the option "--permuteBlocks" in randomise.
>
> You may need to split the contrast file (*.con) into two, one with just
> C1-C3, another with just C4-C6, then make the necessary changes to the
> *.fts file (F-contrast), and run randomise twice (in separate directories
> to avoid overwriting results).
>
> All the best,
>
> Anderson
>
>
> 2015-01-28 14:41 GMT+00:00 Pavel Hok <[log in to unmask]>:
>
>> Dear FSL Experts,
>>
>> I know that similar questions have been already answered many times, but
>> I could not find any satisfactory conclusions on the forum.
>>
>> I was trying to run 2x3 repeated measures ANOVA (i.e. 3 groups, 2
>> timepoints, all subjects have complete data) in FEAT. Since this design is
>> not mentioned in the GLM manual page, I tried to combine two other designs:
>> the factors effect approach (
>> http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#ANOVA:_2-factors_2-_.26_3-levels_.282x3_between-subjects_ANOVA.29)
>> and mixed effects ANOVA (
>> http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#ANOVA:_2-groups.2C_2-levels_per_subject_.282-way_Mixed_Effect_ANOVA.29
>> ).
>>
>> Unfortunately, the Feat complains about rank deficiency. Is there a way
>> to model such a study design? My second guess would be to get the
>> within-subject contrasts using a FE analysis and then run simple T-test. My
>> original design matrix is attached.
>>
>> Thanks for any ideas!
>>
>> Kind regards
>>
>> Pavel
>>
>>
>>
>> --
>>
>> -------------------------------------------------------------------------------
>> -- MUDr. Pavel Hok
>>
>> -------------------------------------------------------------------------------
>> -- Laboratoř funkční magnetické rezonance
>> -- Neurologická klinika
>> -- Lékařská fakulta
>> -- Univerzita Palackého v Olomouci
>> -- Fakultní nemocnice Olomouc
>>
>> -------------------------------------------------------------------------------
>> -- Laboratory of functional magnetic resonance imaging
>> -- Department of Neurology
>> -- Faculty of Medicine and Dentistry
>> -- Palacky University Olomouc
>> -- University Hospital Olomouc
>> -- Czech Republic
>>
>> -------------------------------------------------------------------------------
>> -- I.P. Pavlova 6, 775 20 Olomouc
>> -- web: fmri.upol.cz
>> -- tel.: +420 588 443 418
>> -- e-mail: [log in to unmask]
>>
>> -------------------------------------------------------------------------------
>>
>
>
--
-------------------------------------------------------------------------------
-- MUDr. Pavel Hok
-------------------------------------------------------------------------------
-- Laboratoř funkční magnetické rezonance
-- Neurologická klinika
-- Lékařská fakulta
-- Univerzita Palackého v Olomouci
-- Fakultní nemocnice Olomouc
-------------------------------------------------------------------------------
-- Laboratory of functional magnetic resonance imaging
-- Department of Neurology
-- Faculty of Medicine and Dentistry
-- Palacky University Olomouc
-- University Hospital Olomouc
-- Czech Republic
-------------------------------------------------------------------------------
-- I.P. Pavlova 6, 775 20 Olomouc
-- web: fmri.upol.cz
-- tel.: +420 588 443 418
-- e-mail: [log in to unmask]
-------------------------------------------------------------------------------
------------------------------
Date: Thu, 29 Jan 2015 10:35:55 -0500
From: Dylan Nielson <[log in to unmask]>
Subject: large change in waytotal values
Hi all,
I ran some tractography using probtrackx2 in March of 2014 and I'm trying
to replicate those results now, but I'm getting wildly different waytotal
values using exactly the same seeds, target masks, and data. Using
probtrackx gives me results that are roughly similar to the results I was
getting originally with probtrackx2. I unfortunately don't know what
version of probtrackx2 I was using in March 2014, but the current
probtrackx2 and probtrackx are both updated to FSL build 508. I can't find
any reference to changes in probtrackx2 in the FSL change log. Does anyone
know if there have been changes to probtrackx2 or is there anything else
that could explain this behavior?
Thanks,
Dylan
Here are the commands from probtrackx.log and waytotal amounts:
probtrackx2 March 17, 2014
waytotoal: 76357
/usr/local/fsl/bin/probtrackx2 -x
/home/dylan/RP0642/RP0642_003b/approxVTAs_2/spherical_seeds_T1_nodif/R1_4_T1.nii.gz
-l -c 0.1 -S 2000 --steplength=0.5 -P 5000 --forcedir --opd --pd -f
--modeuler -s /home/dylan/RP0642/RP0642_003b/dti.bedpostX/merged -m
/home/dylan/RP0642/RP0642_003b/dti.bedpostX/nodif_brain_mask -V 1
--dir=/home/dylan/RP0642/RP0642_003b/approxVTAs_2_track/VTA/R1_4_T1/gm_mask
--waypoints=/home/dylan/RP0642/RP0642_003b/subject_masks_nodif/gm_mask.nii.gz
probtrackx2 January 28, 2015
waytotal: 5913
/usr/local/fsl/bin/probtrackx2 -x
/home/dylan/RP0642/RP0642_003b/approxVTAs_2/spherical_seeds_T1_nodif/R1_4_T1.nii.gz
-l -c 0.1 -S 2000 --steplength=0.5 -P 5000 --forcedir --opd --pd -f
--modeuler -s /home/dylan/RP0642/RP0642_003b/dti.bedpostX/merged -m
/home/dylan/RP0642/RP0642_003b/dti.bedpostX/nodif_brain_mask -V 1
--dir=/home/dylan/RP0642/RP0642_003b/approxVTAs_2_track/VTA/R1_4_T1/test
--waypoints=/home/dylan/RP0642/RP0642_003b/subject_masks_nodif/gm_mask.nii.gz
probtrackx January 28, 2015
waytotal: 68242
/usr/local/fsl/bin/probtrackx -x
/home/dylan/RP0642/RP0642_003b/approxVTAs_2/spherical_seeds_T1_nodif/R1_4_T1.nii.gz
-l -c 0.1 -S 2000 --steplength=0.5 -P 5000 --forcedir --opd --pd -f
--modeuler -s /home/dylan/RP0642/RP0642_003b/dti.bedpostX/merged -m
/home/dylan/RP0642/RP0642_003b/dti.bedpostX/nodif_brain_mask -V 1
--dir=/home/dylan/RP0642/RP0642_003b/approxVTAs_2_track/VTA/R1_4_T1/test_probtrackx
--waypoints=/home/dylan/RP0642/RP0642_003b/subject_masks_nodif/gm_mask.nii.gz
------------------------------
Date: Thu, 29 Jan 2015 17:27:59 +0000
From: "Anderson M. Winkler" <[log in to unmask]>
Subject: Re: Repeated measures ANOVA 2x3
Hi Pavel,
The design as shown is for randomise (with the leftmost column, group,
representing exchangeability blocks).
It will also run in FEAT, but then replace that column for all-ones (just
one big group, as that represents then variance groups).
All the best,
Anderson
2015-01-29 15:19 GMT+00:00 Pavel Hok <[log in to unmask]>:
> Hi Anderson,
>
> thanks, will this work with FEAT? It used to complain when two or more
> non-zero values in one EV had different group.
>
> Cheers
>
> Pavel
>
> On 29 January 2015 at 10:44, Anderson M. Winkler <[log in to unmask]>
> wrote:
>
>> Hi Pavel,
>>
>> Please, see attached an example design. Note that each subject is
>> allocated on its own exchangeability block (EB).
>>
>> - For contrasts C1-C3, use the option "-e design.grp" in randomise.
>> - For contrasts C4-C6, and for the F-test, use the option "-e design.grp"
>> and also the option "--permuteBlocks" in randomise.
>>
>> You may need to split the contrast file (*.con) into two, one with just
>> C1-C3, another with just C4-C6, then make the necessary changes to the
>> *.fts file (F-contrast), and run randomise twice (in separate directories
>> to avoid overwriting results).
>>
>> All the best,
>>
>> Anderson
>>
>>
>> 2015-01-28 14:41 GMT+00:00 Pavel Hok <[log in to unmask]>:
>>
>>> Dear FSL Experts,
>>>
>>> I know that similar questions have been already answered many times, but
>>> I could not find any satisfactory conclusions on the forum.
>>>
>>> I was trying to run 2x3 repeated measures ANOVA (i.e. 3 groups, 2
>>> timepoints, all subjects have complete data) in FEAT. Since this design is
>>> not mentioned in the GLM manual page, I tried to combine two other designs:
>>> the factors effect approach (
>>> http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#ANOVA:_2-factors_2-_.26_3-levels_.282x3_between-subjects_ANOVA.29)
>>> and mixed effects ANOVA (
>>> http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#ANOVA:_2-groups.2C_2-levels_per_subject_.282-way_Mixed_Effect_ANOVA.29
>>> ).
>>>
>>> Unfortunately, the Feat complains about rank deficiency. Is there a way
>>> to model such a study design? My second guess would be to get the
>>> within-subject contrasts using a FE analysis and then run simple T-test. My
>>> original design matrix is attached.
>>>
>>> Thanks for any ideas!
>>>
>>> Kind regards
>>>
>>> Pavel
>>>
>>>
>>>
>>> --
>>>
>>> -------------------------------------------------------------------------------
>>> -- MUDr. Pavel Hok
>>>
>>> -------------------------------------------------------------------------------
>>> -- Laboratoř funkční magnetické rezonance
>>> -- Neurologická klinika
>>> -- Lékařská fakulta
>>> -- Univerzita Palackého v Olomouci
>>> -- Fakultní nemocnice Olomouc
>>>
>>> -------------------------------------------------------------------------------
>>> -- Laboratory of functional magnetic resonance imaging
>>> -- Department of Neurology
>>> -- Faculty of Medicine and Dentistry
>>> -- Palacky University Olomouc
>>> -- University Hospital Olomouc
>>> -- Czech Republic
>>>
>>> -------------------------------------------------------------------------------
>>> -- I.P. Pavlova 6, 775 20 Olomouc
>>> -- web: fmri.upol.cz
>>> -- tel.: +420 588 443 418
>>> -- e-mail: [log in to unmask]
>>>
>>> -------------------------------------------------------------------------------
>>>
>>
>>
>
>
> --
>
> -------------------------------------------------------------------------------
> -- MUDr. Pavel Hok
>
> -------------------------------------------------------------------------------
> -- Laboratoř funkční magnetické rezonance
> -- Neurologická klinika
> -- Lékařská fakulta
> -- Univerzita Palackého v Olomouci
> -- Fakultní nemocnice Olomouc
>
> -------------------------------------------------------------------------------
> -- Laboratory of functional magnetic resonance imaging
> -- Department of Neurology
> -- Faculty of Medicine and Dentistry
> -- Palacky University Olomouc
> -- University Hospital Olomouc
> -- Czech Republic
>
> -------------------------------------------------------------------------------
> -- I.P. Pavlova 6, 775 20 Olomouc
> -- web: fmri.upol.cz
> -- tel.: +420 588 443 418
> -- e-mail: [log in to unmask]
>
> -------------------------------------------------------------------------------
>
------------------------------
Date: Thu, 29 Jan 2015 19:31:27 +0000
From: "Siegel, Jonathan Todd" <[log in to unmask]>
Subject: Re: GLM gui w/ Covariates
Hi Anderson,
Thanks so much for the quick response. That's exactly what I was hoping to hear. I have analyzed the data a couple of different ways including chronological age (centered), but I was just looking to mimic the results that I got from the SPSS analyses that I did when extracting the 48 ROIs defined by using the intersection of the mean FA skeleton and the JHU atlas labels. In essence, if other variables are contained in the matrix, but not given 'weights' in the contrasts, then it would be like covarying out the additional variables, right? Thanks again for all of your help.
Best Regards,
Jon
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Anderson M. Winkler
Sent: Thursday, January 29, 2015 3:52 AM
To: [log in to unmask]
Subject: Re: [FSL] GLM gui w/ Covariates
Hi Jon,
This design isn't great. It isn't wrong, but I believe it's just not a very good one.
The problem is the coding of age. I suggest replacing the age categories for the actual subjects' ages. This should give gains in power -- both for testing the effects of age itself, or to test the effect of the other variables while having age as nuisance.
If you are worried that the relationship between age and diffusion parameters could be non-linear, and wanted to have separate groups to capture this potential effect, then you could still use the three age groups, but then consider using separate EVs for each.
The remaining of the design is fine, as well as the call to randomise.
All the best,
Anderson
2015-01-28 21:55 GMT+00:00 Jon Siegel <[log in to unmask]<mailto:[log in to unmask]>>:
Hi FSL Experts,
I had a quick question about using the GLM gui and the randomise function with a set of our TBSS data. I read an email regarding different group sizes and I wasn't sure how to account for this, or what the correct answer is. Here's where I'm at, though. I have a 3 groups of subjects (40 young adults, 36 middle-aged adults, and 66 older adults) that I've taken through the TBSS processing stream. One major question of interest that I have is whether there are any tracts that correlate with memory performance (ie. CVLT, WMS, Associative memory) after accounting for age effects (and other covariates like: Years of education, Gender, APOE4 status, etc). I'm curious whether I can address this using the GLM gui and the randomise command. Here is what I've done so far and I'd like to double-check to see whether this is valid or whether I need to change the way I'm doing things (ie. change the permutation blocks so that each age group is done separately instead of 1 entire group?):
AgeVectorC CVLTc EducationC Intercept
-0.18 13.94 -0.5 1
-0.18 6.94 -0.5 1
0.82 -6.06 -1.5 1
0.82 -2.06 1.5 1
-1.18 8.94 -1.5 1
-1.18 -8.06 5.5 1
To clarify-- This was just 6 participants from the design matrix. Age Vector/Group is -1 for young, 0 for middle, and 1 for old. All variables were centered using the entire sample (all 142 participants) and then I created the design matrix using these centered values.
I'm interested in just those voxels that are significantly correlated with CVLT performance after accounting for age and years of education and so I've created my contrasts file to be as follows:
AgeVectorC CVLTc EducationC Intercept
C1: 0 1 0 0
C2: 0 -1 0 0
C1 would then tell me which voxels show a significant positive correlation with FA after accounting for age and years of education and C2 would then tell me which voxels show a significant negative correlation with FA after accounting for age and years of education????
Here is the randomise command that I've used as well:
#randomise -i all_FA_skeletonised -o AgeGroupCVLT_EducationAgeControlled_FA -m mean_FA_skeleton_mask -d AgeGroupCVLT_EducationAgeControlled.mat -t AgeGroupCVLT_EducationAgeControlled.con -n 5000 --T2 &
Thank you in advance for your help. I know there are other ways to do the analyses (ie. splitting up the groups into separate EVs and looking for group differences and such), but I'm curious if this design and contrast file would be okay for looking at correlations within the entire sample?
Best,
Jon
------------------------------
Date: Thu, 29 Jan 2015 20:13:00 +0000
From: Jane Tseng <[log in to unmask]>
Subject: Problem using fix to train classifier
Dear experts,
I have two different data sets and am using fix to train classifiers for each data set. I get a .RData output file for one of them but not for the other. The only difference between the preprocessing of the two data sets is the high pass filter cutoff, though I wouldn't think that would be a problem? For the data set that I haven't been able to generate a .RData file, fix seems to do "running classifier training" ten times and then quit. Does anyone know what the problem could be?
Many thanks!!
------------------------------
Date: Thu, 29 Jan 2015 20:19:13 +0000
From: David McFarlane <[log in to unmask]>
Subject: Install to /usr/local
I used the NeuroDebian installer to install FSL to Ubuntu 14.04 LTS. As far as I can tell, it installed all the programs right into /usr/bin. This is just ugly & wrong. How can I get FSL programs installed to /usr/local/fsl, where they belong? Thanks.
(Sorry if this has been answered before, I looked around and could not find it.)
-- David McFarlane, Systems Designer, Michigan State University
------------------------------
Date: Thu, 29 Jan 2015 20:28:32 +0000
From: Roman M <[log in to unmask]>
Subject: Vertex Analysis: Minimum ROI Volume?
Hi, I've been using the "alternative Vertex Analysis Pipeline" (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;2965a07f.1411)
to try to include new ROIs in a vertex analysis. Is there a minimum
volume that a region needs to have for the analysis to be possible? For
instance, if the region of interest were a 2D surface (therefore had no
volume - because it's some small region) rather than a 3D volume,
would it be impossible to do a vertex analysis?
Cheers
Roman
------------------------------
End of FSL Digest - 28 Jan 2015 to 29 Jan 2015 (#2015-32)
*********************************************************
|
