| Author |
 Topic  |
|
|
jimmy.dias
BIAC Alum
USA
210 Posts |
 Posted - May 20 2004 : 12:51:17 PM
|
Hi all,
There is now a text based menu driven utility on Golgi that allows you to submit batch processing jobs without having to create/modify your own shell scripts. This utility has the ability to do the following:
BIAC QA
SPM99 Time Slice Correction
SPM99 Motion Correction
SPM99 Coregistration
SPM99 Normalization
SPM99 Smoothing
To use this utility, log in to Golgi and type the following:
[username@golgi: ~] bputil
More information about the utility can be found below:
- Prerequisite: Data must be accessible from Golgi within the "net" directories.
- Prerequisite: All data must have a valid BXH header (fairly recent one at that). For those of you that don't have valid BXH headers, you should use the options in the utility to (re)generate BXH headers for the functionals and anatomicals. An alternative is to use bxhabsorb within MATLAB to create the headers.
- Prerequisite: Regeneration of functional BXH headers require that a pfile header (*.pfh) exist in the run directory where the images are located. All data as of last March (2003) should have that pfile header included. If you are trying to reprocess data even further back than that, first check to see if the pfh header is included and if not, send Francis and I email with the date of the study and the exam number so that we can put pfile headers in the right place.
- An alternate configuration file that specifies the parameters for each of the steps can be specified within the utility. I would suggest copying the global configuration file to your home directory and modifying it to your liking. For example:
[username@golgi: ~] cp /etc/SPM99_procd.conf ~/myprocd.conf
[username@golgi: ~] xemacs ~/myprocd.conf &
- The utility searches for functional data in the following directory:
<experiment directory>/Data/Func/200YMMDD_<examno>/run<seriesno>_<runno>
If the data is not found bputil will also look here:
<experiment directory>/Data/Func/200YMMDD_<examno>/run<seriesno>_<runno>
- If the instruction set contains coregistration or normalization, the utility will look for anatomicals in the following directories:
<experiment directory>/Data/Anat/200YMMDD_<examno>/series<seriesno>
<experiment directory>/Data/Anat/200YMMDD_<examno>/series<seriesno>
- Results for the processing utility will be put in an Analysis directory within Data for now.
<experiment directory>/Data/Analysis/200YMMDD_<examno>/run<seriesno>_<runno>
This will eventually be changed to:
<experiment directory>/Analysis/200YMMDD_<examno>/run<seriesno>_<runno>
- Logs are named by the date and exam no, followed by a timestamp for when the process was submitted. Logs are found here:
/data/users/<username>/BPClient/logs
|
Edited by - jimmy.dias on May 20 2004 1:18:21 PM |
|
|
diaz
BIAC Alum
USA
212 Posts |
Posted - Jun 11 2004 : 4:38:55 PM
|
I've been having some difficulty using bputil with one subject. For this subject I've been using option 6 (tralign,motion, norm, smooth, qa). Something seems to be going wrong in the smoothing step. Because normalization and QA both finish and the respective files look fine. From the log (20030514_41988_1086961591.out) it seems like it's not transfering the snra files over to the experiment directory. But I'm not sure. I've tried running this 3 times and had the same problem each time. I've also tried regenerating the *.bxh headers and using an option with tralign, motion, coreg-normalize, and smooth (which works fine). I've also looked at the original V*.img files and the *.bxh headers and nothing seems to be wrong with any of these files. Any thoughts?
Michele |
Michele T. Diaz, Ph.D.
Associate Director
Brain Imaging and Analysis Center |
 |
|
|
jimmy.dias
BIAC Alum
USA
210 Posts |
Posted - Jun 11 2004 : 6:53:15 PM
|
Hi Michele,
I took a look at the log file and it seems as if the smoothing is being done on the original input for that option (maybe other options as well, I'll check) (run*) instead of the traligned, motion corrected, normalized output (nrarun*). I'll fix the bug, test it and report back. Thanks for the heads up.
Jimmy |
Edited by - jimmy.dias on Jun 13 2004 9:16:02 PM |
|
|
|
jimmy.dias
BIAC Alum
USA
210 Posts |
Posted - Jun 15 2004 : 2:54:13 PM
|
Hi All,
The Processing Utility on Golgi is divided up into two parts, one which helps you create an XML-based list of processes to execute (bputil) and another that actually executes these files.
A copy of the XML files produced from any succesful invocation of bputil can be found in the following directory:
/data/users/<username>/BPClient/ifiles
Multiple submissions are distinguished by a timestamp at the end of the ifile name.
The bug that Michele had mentioned above was due to incorrect parsing of the XML file by the Utility. This bug has been fixed and tested.
Additional notes on directory prefixes:
As some operation is done on a run directory, the scripts on Golgi put the results in a directory that are prepended with certain letters in order to distinguish results from input. Subsequent operations on resulting directories would be put into a directory with a name that is (prefixes) + input directory, i.e. a name that reflects all operations done on the original data.
For example, lets say our original run directory was named run004_01.
Traligning the original run would get a directory named arun004_01
Smoothing traligned run would get a directory named sarun004_01
Accidentally smoothing the original run directory would get srun004_01
Coregistering and Normalizing (one step) smoothed, time-slice corrected data would get ncsarun004_01.
The following table lists all the possible prefixes.
Prefix : Operation : Example (Input->Output)
'a' : Time Slice Correction : blah -> ablah
'r' : Motion Correction : blah -> rblah
'nc' : Coregistration and Normalization (one step) : blah -> ncblah
'n' : Normalization : blah -> nblah
's' : Smoothing: blah -> sblah
Cheers,
Jimmy |
Edited by - jimmy.dias on Jun 15 2004 3:45:47 PM |
|
|
|
diaz
BIAC Alum
USA
212 Posts |
Posted - Jul 12 2004 : 08:35:40 AM
|
Hi,
I've recently collected 3 data sets, submitted them to bputil and am having two problems. I'm missing the smoothed run output and the normalized run output folders are present, but empty. I've looked at the log file, where there is no explicit error messages. The second problem is related to the QA output generated by bputil. The QA done during the scan is fine, but when I look at the QA generated by bputil, the spikiness information is missing (it looks like a web page with missing images). All three of these subjects had runs with missing slices, and for those runs I renamed the folder so that data is not being included in the analyses. Also, I've successfully run bputil on a different subject with missing slices (with the folder renamed and the data not included) which worked fine. For these subjects I am using option 6 (tralign,motion, norm, smooth, qa) and the analyses were run on Friday and Saturday. Any thoughts?
Thanks
Michele
|
Michele T. Diaz, Ph.D.
Associate Director
Brain Imaging and Analysis Center |
|
|
|
syam.gadde
BIAC Staff
USA
421 Posts |
Posted - Jul 12 2004 : 08:56:17 AM
|
quote:
Originally posted by diaz
The QA done during the scan is fine, but when I look at the QA generated by bputil, the spikiness information is missing (it looks like a web page with missing images).
Michele,
Others have noticed this problem too. We are trying to find out the root cause. It seems like the spikiness, mean, and stddev images are not being converted to JPEG images for some reason. They do seem to be converted correctly if you run it manually using fmriqa_generate.pl which is strange. If anyone else has experienced this problem and has any other info that may help, please post here. Thanks! |
|
|
|
jimmy.dias
BIAC Alum
USA
210 Posts |
Posted - Jul 12 2004 : 1:49:40 PM
|
Hi All,
The QA problem described above are related to there being three different versions of jpeg library installed on Golgi. There are those distributed with the java run-time environments, java14 and java131 and those that came with the system. Since bputil uses java under the covers, the convert-to-jpeg utility packaged with the jpeg library that ends up being called is from either one of the two libraries distributed with the java environments (depending on how your user environment is setup) rather than the system jpeg library. We've found that the java131 convert-to-jpeg utility works properly while the java14 convert-to-jpeg utility fails miserably. Ideally, we'd use neither one of the convert utilities distributed with java and instead use the one that is installed on the system. Until I figure out how to do that, I've fixed bputil so that it uses java131 (and subsequently the libjpeg library associated with it) instead of java14. Can someone who has had the mysterious no image problem described above re-run bputil with the QA_only option and report back?
Thanks,
Jimmy |
|
|
|
diaz
BIAC Alum
USA
212 Posts |
Posted - Jul 12 2004 : 3:09:48 PM
|
Hi,
I just re-ran QA via bputil and the spikiness images are now there.
However, normalization is still not finishing (using option 6).
Michele |
Edited by - diaz on Jul 12 2004 3:16:02 PM |
|
|
| |
Topic |
|
|
|
Golgi Processing Utility
