 |
 |
Dear Paul, I'm not quite sure why you are going to all this effort here. Is the correction via the FEAT GUI working correctly or not? If it is then you shouldn't need to do anything else except just use the filtered_func_data output by FEAT in your resting-state analysis. Also note that the MELODIC GUI offers the same options. I really wouldn't ever expect people to be applying forward warps themselves to any data. This is done internally to get a good alignment between the fieldmap and EPI data spaces, although it is complicated by the fact that each has a distorted and undistorted space. If there is some problem with the correction then please let us know. Otherwise, you shouldn't need to try and replicate all the internal steps, and I would certainly never recommend you inverting the forward warp, as these warps may not be 1-to-1 as the EPI distortion process in no way guarantees that topology needs to be preserved (although in practice it is never seen as this amount of warping is accompanied by massive signal loss and so there is nothing left to see). All the best, Mark On 17 Dec 2010, at 17:27, Paul Guillod wrote: > Hi there, > > I am having trouble optimizing the field map corrections on my EPI resting-state data. I ran the pre-stats through feat which applied the B0 correction decently well, but I noticed when you forward warp the magnitude image in fugue to align with the base of the EPI series it is warped in the exact opposite direction than it should be (at least when you visually inspect FM_D_fmap_mag_brain_siglossed with EF_D_example_func in the feat output/warp directory - since my EPI set was phase encoded in the x-direction it is apparent on a coronal slice). Is this intentional for alignment purposes? > > Normally when I forward warp the magnitude image for alignment with my EPI base I have to specify the opposite direction in fugue (x instead of x-) as when I save an undistorted output. That said, when I forward warp the magnitude image it aligns *incredibly* well with the distorted EPI. Since I already resampled the field map to the EPI grid, I was hoping to save the forward warp, invert it, and apply that to the EPI set (certainly potential for a novice error). > > In the feat logs the shift map from fugue is saved as a warp to apply simultaneously with the rigid motion correction matrix using applywarp, so I figured that taking the forward shift, inverting it, and applying it to the EPI data would be more accurate than running the correction using a rigidly aligned undistorted field map to distorted data. However, my modest attempts have failed to achieve this. Even when I apply the inverted warp to the forward warped field map magnitude image (from fugue) it doesnt output the original magnitude image as one would expect ( f-1(f(x))=x assuming warp is 1-to-1 function, right?). So I checked the files where the shift map is converted to a warp and noticed a discrepancy between the output of the data corrected in fugue and the data corrected using the shiftmap *from* fugue. > > For instance, in the main feat output directory the image example_func_orig_distorted is identical to warp/EF_D_example_func. The former is corrected using applywarp specifying the shift map from fugue for the output example_func whereas the latter is corrected using fugue directly to undistort the image into warp/EF_UD_example_func. Since the two are identical and corrected using the same shift map (one using fugue and the other using the converted warp map), the undistorted outputs should be identical as well, right? But when you compare the outputs they were shifted differently (warp/EF_UD_example_func was shifted more than example_func). Did I miss something or does fugue undistort data differently than applywarp? Or does the error lie in convert warp? > > I apologize for the wordiness in the message, but would appreciate any assistance. No doubt I raised concerns that have formerly been addressed so please don't hesitate to redirect me. Still new to most of this stuff. > > Thanks, > > -paul >