| Author |
 Topic  |
|
|
mcclernon
New Member
USA
14 Posts |
 Posted - Jul 12 2005 : 10:59:05 AM
|
We are attempting to run tstatprofile2 on a dataset of the size and type we have run before, but are having difficulty. We are using matlab 7 with latest bxh and biac tools.
When I run tstatprofile2 on realigned files (64x64x28)-- no problems. However, when I attempt to run on the larger normalized files (79x95x68)-- we run into memory errors. this is a bit odd since we have successfully run tstatprofile on datasets of the same size in the past on the same machines.
I thought it might help to resample the data during normalization and run tstat on the resulting smaller files. We reran normalization with 3x3x3 voxel size and did end up with smaller files (53x63x46). However, neither tstatprofile2 or readmr will read these files. We get the following error with readmr:
??? Error using ==> readmr
Input size array doesn't match size of data!
Any thoughts on 1) why tstatprofile2 won't work on our larger nra*.img files and/or 2) why readmr doesn't like the resampled nra*.img files?
Thanks,
Joe |
Edited by - mcclernon on Jul 12 2005 10:59:45 AM |
|
|
petty
BIAC Staff
USA
453 Posts |
Posted - Jul 12 2005 : 11:05:59 AM
|
the first thing to check is the BXH header. if you resampled the size of the data without writing new BXH headers, then you dimensions are wrong, and therefore "Input size array doesn't match size of data!".
you can manually change the BXH header in wordpad, or just write out a new one using bxhabsorb.
|
 |
|
|
mcclernon
New Member
USA
14 Posts |
Posted - Jul 12 2005 : 11:14:38 AM
|
| We were careful to rerun bxhabsorb after resampling. bxhabsorb runs successfully and when I look at the resulting header, the parameters match what we expect to find. however, when I use readmr to read in the data using the bxh header, I get the same error "input size array doesn't match...". |
|
|
|
petty
BIAC Staff
USA
453 Posts |
Posted - Jul 12 2005 : 11:16:40 AM
|
did you use the wildcard checkbox? without this it only makes the header for nraV0001.img, instead of the whole series.
if you look at the header it'll either list all of the images within it, or just the 1 |
|
|
|
diaz
BIAC Alum
USA
212 Posts |
Posted - Jul 12 2005 : 11:19:22 AM
|
| alternatively, is it possible that some files *.txt are included in the bxh-header that were not supposed to be. Typically, when I've gotten this error, there has been some kind of mismatch between the data and the header (wrong dimensions, wrong files listed, incomplete number of files listed - when this is the case, it may have been that the previous preprocessing step did not complete). |
Michele T. Diaz, Ph.D.
Associate Director
Brain Imaging and Analysis Center |
|
|
|
gregory.mccarthy
BIAC Alum
USA
26 Posts |
Posted - Jul 12 2005 : 1:37:32 PM
|
| Although this is not directly relevant to the problem raised, you might consider switching from tstatprofile2 to the new eventstats program. Eventstats was designed to supercede tstatprofile, and takes advantage of new programs that extract timing information directly from Eprime, Presentation, and CIGAL files. Eventstats also uses the files created from the automatic QA (quality assurance) procedure to exclude from analysis those epochs marred by artifact. |
|
|
|
mcclernon
New Member
USA
14 Posts |
Posted - Jul 12 2005 : 1:53:41 PM
|
in response to Michele and Chris-- I don't know that this is a bxh header related problem. I've had no problems running tstatprofile2 on the a*.img and ra*.img files which do not have bxh headers. I'm starting to think this may be machine related, but it doesn't make sense given that we've run the same analyses before (i've maxed VM, the machine has 2GB ram etc.). I may bring some files over to the biac lab and see if I can get it run over there.
Joe |
|
|
|
syam.gadde
BIAC Staff
USA
421 Posts |
Posted - Jul 12 2005 : 2:00:22 PM
|
| To clarify, in a separate email discussion, we determined that there indeed was a problem with the .bxh headers for the "re-sliced" data (the image data files were larger than expected and had extra data at the end). However, the root cause of the memory errors (which motivated the reslicing in the first place) seems not to involve the .bxh headers. |
|
|
|
mcclernon
New Member
USA
14 Posts |
Posted - Jul 12 2005 : 4:39:03 PM
|
OK. Just an update. We tried to run tstatprofile2 on our normalized files over at the BIAC lab. Same errors as we saw on our machines. As at our lab, worked fine on ra*.img files.
We've been able to get tstatprofile2 to work on individual runs so that's what we'll do using batch commands. After processing all runs, we'll average the results across runs and then ultimately across subjects. Not elegant, but it will work for now.
thanks to everyone who helped us work through this.
Joe |
|
|
| |
Topic |
|
difficulty with tstatprofile
