| T O P I C R E V I E W |
| Bethany |
Posted - Aug 03 2007 : 3:17:06 PM
So, I just tried to run a third-level (across subejcts) analysis in FSL and hit a snag: the analysis seemed to run all right, but when I went to look at the data, part of the brain was missing. All the axial slices between above the corpus callosum down to halfway through the cerebellum are just not there. The second-level (within subjects, across runs) brains look fine, at least the ones I've looked at so far.
Help?
EDIT: Further analysis suggests that, although all the second-level analyses were run in the same way with the same template, the brains are oriented in a different manner in each one. I presume this explains the problem with my third-level analysis, but then raises the question of why each individual subject is oriented differently and what can be done about it. Do I need to change the individual header files? How do I do this? And, do I need to re-run my first-level analyses again after this or just re-do the registration on the problem subject? |
| 5 L A T E S T R E P L I E S (Newest First) |
| vinod |
Posted - Aug 06 2007 : 2:35:26 PM
The problem turned out to be an issue with the batch script referencing the orient file. The orient file was not being referenced in some of the sessions. |
| Bethany |
Posted - Aug 06 2007 : 11:45:36 AM
I'm not sure what you mean by using an initialization matrix. I made a different orient.mat file for each subject, if that's what you mean. |
| vinod |
Posted - Aug 06 2007 : 09:58:27 AM
My suspicion is the same too. You are probably using the same initialization matrix for all subjects rather than generate one for each of them.
Also, is there anything similar/consistent about these three subjects: were they the first three or last three (their exam numbers look quite close to each other)? If the same scan parameters were used for all your subjects, using the same initialization matrix should not affect things much.
Now to the question on whether you need to run your whole first level analysis or registration, I am tempted to say that just running registrations should do, but I am not certain yet. FSL runs all stats in the first level on non-registered data. And uses the registration parameters only when combining sessions at higher level. But, when I asked a similar question on the FSL mainling list, I got a vague reply about not being encouraged to do that since it messes up the file structure in the individual directories. |
| petty |
Posted - Aug 06 2007 : 08:45:25 AM
is it possible that they started off in different orientations, and this is the result of trying to merge all of them together?
did you use an initialization matrix for these subjects, from their own anatomicals, seperately? |
| Bethany |
Posted - Aug 04 2007 : 12:07:24 AM
To expand on the problem, all the anatomical and functional files appear (as far as I can tell, anyway) to be properly aligned before the first-level analysis. After the first level analysis, the subjects wind up in several different orientations (although they were all processed through the same .fsf template and batch script.)
Here's a picture of three of my RiskTime.01 subjects (31404, 31412, and 31420) in three different orientations
(backwards, proper, and really messed up, respectively), plus the original, properly BET-extracted anatomical from 31420 (the one that appears to be at a diagonal after processing).
Link to picture of misoriented brains so it doesn't stretch out forums
|
|
|
