Show all posts by user
Dear AFNI users-
We are very pleased to announce that the new AFNI Message Board framework is up! Please join us at:
https://discuss.afni.nimh.nih.gov
Existing user accounts have been migrated, so returning users can login by requesting a password reset. New users can create accounts, as well, through a standard account creation process. Please note that these setup emails might initially go to spam folders (esp. for NIH users!), so please check those locations in the beginning.
The current Message Board discussion threads have been migrated to the new framework. The current Message Board will remain visible, but read-only, for a little while.
Sincerely,
AFNI HQ
History of AFNI updates
Page 2 of 3
Pages: 123
Results 31 - 60 of 61
Hi,
I'd like to run a group analysis on my nonlinear HRF model (thanks to Rick Reynolds).
For this first test, I only consider one group, but as far as I know 3dMVM requires at least one between-subject variable, so I included 'Group'.
The response function consists of 7 parameters which I would like to test with the multivariate method.
With the code below I get the error,
by
paranoidandroid
-
AFNI Message Board
Hi, I'd just like to report I still get the error about the missing R_io.so when using R3.02
Uninstalling R and installing 2.15.3 from a .deb package solved the issue.
(Everything under ubuntu 13.10 64bit, newest AFNI build, January 2014)
by
paranoidandroid
-
AFNI Message Board
rick reynolds Wrote:
-------------------------------------------------------
> If you want the clusters to be approximately the
> same
> size and in approximately the same place, then I
> would
> not expect varying the 3dClustSim inputs to work
> well
> enough.
>
> Do you really need single subject level
> significance,
> or do you just want to find a
by
paranoidandroid
-
AFNI Message Board
I'd like to apply cluster correction on the single-subject level in order to define individual ROIs for later group analysis.
The problem is, that the inter-individual variation in the statistical maps is so high. For some subjects I would have to set a p<0.001 to get more than one huge blob, while in others I might need p=0.05 to get any voxels at all.
Do you think it would be legit
by
paranoidandroid
-
AFNI Message Board
I needed the same thing, you basically have to write your own script for that.
Thats approximately how I did it:
-create a file with stimulus onset times
-loop 3dTcat over the lines of this file producing data segments for each stimulation period
(as input use the baseline/drift/motion corrected raw time series, which is also not output by default, but can be created as explained in other
by
paranoidandroid
-
AFNI Message Board
Hi,
I couldn't find a documentation about the graph window functions, especially the "IFFT()|" transformation.
It looks like it shows the fourier transform of the voxels time series, which would be nice.
The x-axis has the same number of points as there are volumes in the time-domain, so am I right to assume that the highest point corresponts to frequency TR/2 ?
Surprisingly t
by
paranoidandroid
-
AFNI Message Board
I am also wondering about DOFs, the value in statpar seems to be the number of volumes minus the parameters used.
But doesn't it very much depend on the design that is used? Lets say I have two scans with 1000 repetitions each. In scan A I have a single stimulus block, in scan B I have 100 blocks. Both scans have the same DOF, but according to the definition in wikipedia "DOF is the
by
paranoidandroid
-
AFNI Message Board
no, they were resliced in the coregistration but the voxel size remained unchanged. I can upload you a sample dataset if you are interested (after the ISMRM deadline...).
by
paranoidandroid
-
AFNI Message Board
It happens when I average statistical parameter maps with 3dMean. (I admit, its not the best analysis approach, but I just wanted to get a rough idea how the maps look across subjects). They are all coregistered and have exactly the same number of voxels, so I don't understand how p and q values got affected differently... (In each individual subject, this doesnt happen. There, q is alwa
by
paranoidandroid
-
AFNI Message Board
I wonder, in my case I have q values that are smaller than the p values. How is that possible?
by
paranoidandroid
-
AFNI Message Board
Whenever I'm looking at some data, I need to set up the windows in a specific way. Since it involves a lot of clicks, I was wondering if that can be defined in the default settings or in a startup script.
However, the plugin 'save layout' is quite very rudimentary and only saves the size of the windows.
Is it possible to add the following options?
-open axial image window (
by
paranoidandroid
-
AFNI Message Board
I am wondering, is autocorrelation in the data not a problem for the statistics obtained with 3dNLfim? Can 3dREMLfit be combined with nonlinear regression or is this only necessary for linear regression?
by
paranoidandroid
-
AFNI Message Board
I just found some mistakes (the timing was incorrect because I forgot that I already removed some subbricks in the postprocessing), it looks much better now. I may come back to your offer after more testing...
by
paranoidandroid
-
AFNI Message Board
hi Rick,
Sorry, I was on holidays. Thanks a lot for the help. Now I tried the function on some real, noisy data and it works quite well. Ocasionally, however, I get some fits that look not so optimal by eye (see picture). Even though I set the parameter boundaries high enough, it doesn't seem to make the fit broad enough. Maybe it is trapped in some local minimum. Is there a rule of thumb
by
paranoidandroid
-
AFNI Message Board
I see, in that case i think TR-locked stimulus times are good enough. I just wasn't even aware that 3dNLfim performs a convolution because the help doesn't mention the possibility of entering a stim_file via this environmental variable.
So basically, I would have to start with model_conv_diffgamma as a basis and replace "gamma_model" with my desired model? I can try that..
by
paranoidandroid
-
AFNI Message Board
I have a general question about the usage of 3dNLfim, I could not find so much information about it apart from the manual.
So it seems that there is no way to enter -stim_times.
Then if I have a block design with multiple blocks per time series, how is the approach to analyze this?
Should I first average all blocks and create a new time series with stimulus begin at subbrick 0? That could be
by
paranoidandroid
-
AFNI Message Board
Hi Rick,
Thanks for your answer. In my case I would definitely want to use 3dNLfim with all 7 parameters and define upper- and lower limits for each.
One could still take the output of this fit and enter it as regressor in 3dDeconvolve for fixed shape, but I am interested in the variability of the parameters across brain regions.
by
paranoidandroid
-
AFNI Message Board
this has become a priority for me as well, and I'd like to implement the following model in afni. However, I am not at all familiar with the afni/linux/C syntax, can someone help me to translate this function in the language of waver/3dcalc?
the code that I use in matlab:
t1=[0:sampling_rate:h1];
y1=f1/2*cos(2*pi*1/(2*h1).*t1)-f1/2; %initial dip
t2=[0:sampling_rate:h2-sa
by
paranoidandroid
-
AFNI Message Board
I know that fsl has such a function (in the FLOBS toolbox), it's sometimes called four half-cosine model. I don't know how to incorporate this in afni, but if you can obtain the HRF itself (which I think is this tool doing) you could manually convolve it with your stimulus in matlab or R and enter it in afni as a regressor.
by
paranoidandroid
-
AFNI Message Board
Gang Wrote:
-------------------------------------------------------
> > is there an option for unequal spacing of the
> intervals, i.e. to have more beta
> > weights for the interesting first 20s than for
> the post-stimulus response?
>
> One approach I can think of is to artificially
> break each block into two trial types (as well as
> their timings): the
by
paranoidandroid
-
AFNI Message Board
ah ok, I didn't know align_epi_anat.py already concatenates the matrices automatically.
I adjusted the code below, sot that the Allineate calculcation is only done on the sub-brick #0 of the source. Since I already use volreg for motion correction, I think it would be a bit double-stitched (and extremely slow) to run allineate also on every repetition.
3dTcat -prefix first_only_$subj
by
paranoidandroid
-
AFNI Message Board
I want to coregister scans of the same modality (EPI to EPI), combined with 3dvolreg for motion correction. After a lot of reading I pasted together this code. It works - although not overwhelmingly well, especially the outlines of the brains sometimes don't match, it could need a little bit more stretch..., and as you can see below the images don't have a lot of details (because they
by
paranoidandroid
-
AFNI Message Board
thanks for the answers,
I have a very related questions, instead of opening a new thread I hope someone will read it here:
How critical is the choice of duration for TENT(zero)/SPLIN(zero) etc. models?
What should keep me from setting the duration as long as the inter-stimulus interval? Because I would like to see how the hemodynamic response returns (and stays) to baseline, because if you
by
paranoidandroid
-
AFNI Message Board
Hi,
Lets say in one experiment, I give single stimulus pulses and want to model the response function for ~15s after each pulse, this is no problem with around 6-8 parameters for the TENT or SPLIN function.
But for a block design of 20s stimulus duration and expecting a fairly long BOLD response after stimulus offset (under- or overshoot..) I would need a huge amount of parameters, and since
by
paranoidandroid
-
AFNI Message Board
Dear Afni experts, as you can appreciate from the picture, there seems to be a strong trend in my time series, which I don't want to see in my plots.
This is the command (is used a very high polynomial degree because the scans are 1200repetitions)
3dDeconvolve -input pb04.$subj.r*.scale+orig.HEAD \
-censor motion_${subj}_censor.1D
by
paranoidandroid
-
AFNI Message Board
Thanks for your answer, the 3ddeconvolve command looks like this:
# run the regression analysis
3dDeconvolve -input pb04.$subj.r*.scale+orig.HEAD \
-censor motion_${subj}_censor.1D \
-polort 5 \
-num_stimts 7
by
paranoidandroid
-
AFNI Message Board
(sorry for another thread but google couldn't help me out)
You have to look at the picture below to understand my dilemma. When displaying the graph of scaled time series I can very clearly see a strong activation to the stimulus in the sensory area, the red marker shows 6.8% at the peak.
But why on earth is the beta-value (OLay=0.49) so low? As I understood it should roughly represent
by
paranoidandroid
-
AFNI Message Board
When I try AFNIs 3Dautomask with the default settings I get something like in the attached picture. I must admit, we have quite some ghosting artefacts here (and the EPIs are from mouse brains), but isn't it also a bit silly of afni to assume that there are two brains which are clearly separated from each other?
Can you recommend me some parameters that might alleviate this problem? But i
by
paranoidandroid
-
AFNI Message Board
Has there been any progress in automatically reading bruker 2dseq into afni?
I would be very interested in that in order to avoid additional transformation steps to DICOM or Nifti...
will try with some to3d options...
Edit: it worked very well with the folowing command: to3d -time:zt 12 740 1000 alt+z 3D:0:0:90:60:8880:"2dseq.seq"
(Bruker EPI are stored in the order slices(z)
by
paranoidandroid
-
AFNI Message Board
Thanks a lot rick! These tent functions sound really awesome.
I found a very good tutorial here afni.nimh.nih.gov/pub/dist/edu/latest/afni06_decon/afni06_decon.ppt but I am not at all familiar with afni, is there a sample code and some sample data that would do exactly this?
But one important question: Does this also allow negative beta values for the individual functions? Because I am e
by
paranoidandroid
-
AFNI Message Board
Page 2 of 3
Pages: 123