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
The first week of December 2013. We will have a registration web page up sometime in September.
by
Bob Cox
-
AFNI Message Board
The chopping part is the non-trivial procedure, since the GM mask is a complex beast. That is, simply chopping 3D space into pieces and then taking the intersection of those pieces with the GM mask would give regions with variable sizes, depending on the overlap of the GM mask with the spatial subdivision. I could whomp up something like this (using truncated octahedra as the spatial basis -- t
by
Bob Cox
-
AFNI Message Board
Modifying 3dvolreg is probably not going to happen -- I wrote it in 1998 (I think) and after such a length of time, I'm wary of touching a code that works.
3dAllineate could get the -maxdisp options, but not until HBM 2013 is over (at the earliest).
by
Bob Cox
-
AFNI Message Board
It is certainly possible to use 3dTfitter to solve a linear system from a matrix (e.g., from 3dDeconvolve) using L1 regression.
The difficulty with such an approach is ascribing statistics to the resulting beta values. There is no good theoretical closed-form way to do that, so some sort of resampling method must be used -- which is very computationally intensive.
You don't need to use
by
Bob Cox
-
AFNI Message Board
The 'calc' functions don't like bad inputs, and so convert them to good values. For example:
ccalc '1/0'
ccalc 'log(-3)'
ccalc 'asin(3)'
ccalc 'y0(0)'
None of these results make much sense, but the alternative would be to produce perhaps a million error messages as bad values transpire in a big volume.
by
Bob Cox
-
AFNI Message Board
The next scheduled bootcamp is 02-06 Dec 2013, here at the NIH. That's the earliest we could get the auditorium here for an entire week. We will announce registration on this web site's message board sometime in September, I imagine. (First we have to survive the HBM 2013 meeting, at least!)
Also, we are having trouble with our server right now, so access to this web site will be
by
Bob Cox
-
AFNI Message Board
For the first question: there's no paper describing IM -- it's too simple a concept to publish IMHO. You can just cite the original AFNI 1996 paper, I guess.
For the second question, you'll have to provoke Gang Chen to answer.
by
Bob Cox
-
AFNI Message Board
I'll have to take a look at your datasets. Since you are at the NIH, the simplest thing is for you to put them on a flash drive and bring them to me -- Bldg 10, Room 1D73. Please call or send email before you just show up at my office.
by
Bob Cox
-
AFNI Message Board
Ah, I see -- the shape is constrained to be the same between different classes of conditions, just the amplitude varies.
As they point out in the paper, in the general case this is a nonlinear problem, which is outside the computational range of 3dDeconvolve. I once started to write such a program, but abandoned in in December 2007 when no one I talked to was interested in it.
So there are
by
Bob Cox
-
AFNI Message Board
I'm afraid that at first glance, I don't see a fundamental difference between the analysis in the linked arxiv paper and that done with 3dDeconvolve+3dREMLfit using (say) CSPLIN basis functions.
by
Bob Cox
-
AFNI Message Board
It is possible to do what you want (create a dataset with the baseline model estimate and censored TRs) removed. You use the 3dSynthesize program after running 3dDeconvolve.
by
Bob Cox
-
AFNI Message Board
When you get stuck with a problem for many days, ask us for help! Often there is a solution already created, as in this case.
by
Bob Cox
-
AFNI Message Board
If you have an up-to-date version of AFNI (less than a month old), you can try this first
3dUnifize -input ANAT+orig -prefix ANATuni
This will (approximately) unshade the T1-weighted volume, and then you can try 3dSkullStrip on the ANATuni+orig dataset.
All this is assuming the original volume is at about 1x1x1 mm resolution, which is what we are used to.
by
Bob Cox
-
AFNI Message Board
In the second command, the input dataset is specified as
'TS.1+orig.BRIK TS.1+orig.HEAD[48]'
What this means is to take the 2 dataset specifications inside the quotes, read both datasets (all sub-bricks from the first one, and #48 only from the second one), take the median of all sub-bricks at each voxel, and then do the automask procedure.
In the first command, there is only
by
Bob Cox
-
AFNI Message Board
macosx_10.5_Intel_64.tgz is available -- it is compiled for a 64 bit machine.
by
Bob Cox
-
AFNI Message Board
Simple -- every measurement is duplicated. Since the t statistic (1 sample, say) is something like
t = (sample mean) / sqrt( sample variance / N )
if you accidentally duplicate datasets (either .HEAD or .BRIK can be used as input filenames, but using both gives each dataset TWICE), then N is bigger, so t is bigger -- artificially.
by
Bob Cox
-
AFNI Message Board
It is possible with Group InstaCorr, but not with single subject InstaCorr.
by
Bob Cox
-
AFNI Message Board
I have just further modified the compression stuff in AFNI to use pigz and pbzip2 by default in place of gzip and bzip2, if the former programs are present. AFNI pipes the data through these external programs -- it doesn't actually do the compression or decompression internally using library code.
AFNi doesn't supply pigz or pbzip2, but I recommend that users install these programs
by
Bob Cox
-
AFNI Message Board
1dplot will ignore lines that start with the '#' character. If '-ok_1D_text' doesn't work to give the plot for you, you'll either have to edit the file so the header lines start with '#', or these lines are just deleted, or you'll have to use some other plotting program.
by
Bob Cox
-
AFNI Message Board
You have 2 options for dealing with this type of spatial non-stationarity, inside the AFNI bubble-universe.
(1) Ignore it. At the group analysis level, it is not clear at all that the inhomogeneity in smoothness of the FMRI noise is important, relative to to all the other statistical "sins" that are committed along the way.
(2) Eradicate it. That is, blur the time series data, b
by
Bob Cox
-
AFNI Message Board
And at some point in the nebulous future (but before the Andromeda Galaxy hits us), I will make this graph Fading the default -- at that point, you'll have to turn Fading off manually or by setting AFNI_GRAPH_FADE to NO.
by
Bob Cox
-
AFNI Message Board
Second try on including an image, since attachment didn't seem to work:
by
Bob Cox
-
AFNI Message Board
I recently (last Friday) added a requested feature to the AFNI time series graph viewer -- the ability to mark sub-graphs of those voxels that are below threshold -- that is, those voxels that are un-colored when See Overlay is turned on.
To enable this feature in a particular graph window, you can type the F key while the mouse focus is in the graph window, or you can use the new item 'T
by
Bob Cox
-
AFNI Message Board
It's part of afni.c -- a function that the dataset loading function can call if it runs out of memory. Other programs don't provide such a "feed me Seymour" function, so if a dataset fails to load in any of the 3d*** programs, they just explode.
In afni.c, the FMS function (a name and acronym I just invented) goes thru and purges from RAM all datasets that aren't bein
by
Bob Cox
-
AFNI Message Board
Because of laziness -- I mean careful husbandry of programming resources -- when I coded that part of AFNI, back about September 1994.
by
Bob Cox
-
AFNI Message Board
Job Ads - 11 years ago
Here are a couple of neuroscience job openings for PhD-level people that might be of interest:
NINDS extramural positions (closes Monday, 25 Mar 2013)
https://www.usajobs.gov/GetJob/ViewDetails/339670400
PostDoc position in Tulsa OK:
http://www.laureateinstitute.org/upload/file/PostDocHamilton.pdf
by
Bob Cox
-
AFNI Message Board
The filter in 3dBandpass is an FFT-based filter. That is, it does what Paul Taylor says: transforms the data to the frequency domain (from the time domain), sets to zero the values it doesn't "like", then transforms back to the time domain. This type of filter has some ringing artifacts in the time domain -- the impact of these depends on what use you are making of the filtered t
by
Bob Cox
-
AFNI Message Board
So if this were implemented, what should show up in the graph box of a sub-threshold voxel -- a flat line, a blank box, a picture of a random AFNI user's face?
by
Bob Cox
-
AFNI Message Board
This will require adding capabilities to the AFNI driver functions and to the clusterize plugin (to accept outside control). I'll take it under advisement.
by
Bob Cox
-
AFNI Message Board
The next full AFNI FMRI data analysis bootcamp will be held in Lister Hill Auditorium on the NIH campus, from Feb 25 to 01 Mar, 2013. This class will be 4-1/2 days long, ending about noon on Friday. The auditorium can hold over 100 students, and so the class will be open to all takers -- NIH and non-NIH people alike. Students must bring a laptop loaded with AFNI and the course materials and da
by
Bob Cox
-
AFNI Message Board