| T O P I C R E V I E W |
| cyt |
Posted - Jul 11 2006 : 5:23:54 PM
I'm getting a left-right flipping of fMRI images in SPM2, as indicated by motor neuron response in the hemisphere opposite of what should have occurred. It is more than a simple mirroring of images, as the coordinates of search volumes are also appropriately mirrored. I presume there is a setting within SPM that is causing this problem, so if anyone knows where to find this switch, or what the cause of this flipping is, I'd greatly appreciate it. |
| 6 L A T E S T R E P L I E S (Newest First) |
| josh.bizzell |
Posted - Jul 28 2006 : 10:14:19 AM
I think what you will want to do is either flip the raw data before preprocessing or flip the preprocessed data before estimation. Both of these can be done in SPM:
- Open SPM
- Click display and find a sample image (raw data or preprocessed data)
- In the "resize {x}" field, enter "-1"
- Click "Reorient images..." button, hit "OK" button, and select all images you want to flip
This will create .mat files in SPM99 and SPM2 if you're working on raw data, or it will edit the existing .mat files if you're working on preprocessed data. If you're using SPM5, it will edit the transformation matrix in the header file, so you won't see any .mat files.
You might also be able to do the steps above on the beta maps, if you don't want to re-run the estimation. Personally, I'd probably just flip the preprocessed images, and re-run the estimation, (you might not have to re-run the modelling, if you have SPM.mat files already).
-Josh |
| cyt |
Posted - Jul 27 2006 : 12:07:06 PM
They are indeed preprocessed. Are there any ways, short of re-preprocessing, to reverse this flipping then? |
| petty |
Posted - Jul 12 2006 : 3:13:33 PM
are the inputs to your stats already pre-processed? if so then they should be flipped because spm has already performed the R->L flip that syam mentioned above. |
| syam.gadde |
Posted - Jul 12 2006 : 1:41:38 PM
I'm not too familiar with SPM, but maybe if you bring the data up in the GUI, you might be able to explicitly specify the input orientation (which I think gets written to a .mat file). |
| cyt |
Posted - Jul 12 2006 : 1:34:57 PM
My inputs are .img files associated with .hdr files, so there aren't any BXH files to work with. They go straight to modeling and estimation, if that affects anything. Is there a specific point at which SPM determines image orientation? I'd edit the underlying .m files if that's what it took. |
| syam.gadde |
Posted - Jul 11 2006 : 7:23:23 PM
What input are you giving to SPM2? I believe SPM2 assumes that the input is in LAS orientation (i.e. x-dimension goes from R->L, y from P->A, etc.). I also believe that during the normalization step, it flips the R->L axis and the output orientation is RAS -- this is expected behavior I believe. So, you may want to make sure your input data is in LAS format, and if not, and if you have a .bxh file, you can do something like: bxh_reorient --orientation=LAS input.bxh output.bxh and then use bxh2analyze to create the .hdr/.img files that SPM2 requires. This is what bputil does under the hood. |
|
|
