Hi Ajay,
> Thanks for the explanation. I think some of the reason for confusion came from the fact that the pixelshift map was not always an integer number and so I thought there is potential for more than one voxel being mapped into a single voxel space in the undistorted space (either fully or partially depending on the value). I was not sure in this case how the intensity is handled when just using pixelshift method (without intensity correction)? For example in the frontal lobe where the image is compressed (if acquired A >>P), then we see signal loss in the front of the brain and signal stacked together in the concave shape. If this is the case, how does the pixel shift map know how much of the intensity belongs to each voxel in undistorted space when there is compression of multiple voxels, or does it assume the same whole intensity for each pixel in undistorted space that maps to it?
Yes, if distortion causes the signal from a number of voxels to collapse into a single voxel, the intensity from that single voxel in the distorted image will be used for multiple voxels in the corrected image. As the shiftmap is real-valued, this will never be exactly the case, but compression is a problem in areas like the frontal lobe. If you use intensity correction, the intensities will be divided by the amount of stretching that occurs when going from distorted to undistorted coordinates.
> Also there was a note that the icorr method is ill conditioned for fieldmaps, however is this also true of matched EPI resolution fieldmaps that are more recent as opposed to lower resolution fieldmaps from the past? There was not much detail on why it is ill conditioned so I was curious if you had any details on this.
I am not completely sure what this statement in the documentation refers to, but I am guessing that it is about situations where the finite differences that FUGUE uses to find the amount of stretching break down. These should be more reliable with higher resolution images (as long as SNR is good), but you will probably want to have a close look at your corrected images if you decide to use this option - in particular, look at regions where the fieldmap varies quickly and around the edges of the brain. Most of the time the lack of intensity correction doesn’t really matter though, so depending on your analysis you may be fine just leaving it off.
Cheers,
Eelke
|