Dear jeongchan,
> i'm doing human fmri study with stimulation on the subject.
>
> mostly whenever we applied pressure, subjects' head move aroud 0.8 sometimes ~1.8 (mm? voxel? shift parameter in the *.par file from mcflirt-fsl). i know i'd better change the design might not cause much motions, but it'll be critical for me if i change.
>
> probably, the raw data in itself have a distorted slice timing and/or too much motion so that the data will be interpolated alot.
>
>
> here's my question...
>
> is there any way to use priori information about the movement(shift mostly) so that i can shift each volume without any interpolation before i run mcflirt-fsl, as if the 4D data collected without much motion?
> the reason that i'm asking is that i don't want the data to be interploated too much, so it would be better if i can change the initial location before motion correction.
> (i've looked through the mcflirt options '-init' but don't know in detail.)
>
> my idea: put some markers on the subject's head during the scan then detect the marker so that i can quantifiy the movement (using mcflirt only based on the marker not the brain) during the run. then shift the each volume using the shift-information from my marker then perform motion correction. but don't know more better way.
the interpolation errors increase up to a half of a voxel, after which they decrease again and become zero/negligible at one voxel. They do not get biger and bigger as the distance/movement increases. Therefore I do not think that your strategy would buy you anything.
There are other movement related affects that do increase approximately linearly with movement (specifically with rotation around an axis non-parallel to the main magnetic field) but those cannot be corrected by the strategy you suggest.
Good luck Jesper
|