Thanks very much for the tip - this helps enourmously (using FLIRTs
rectangluar filter requires 17 seconds per individual instead of 442
seconds for the default sinc filter).
By the way, I notice that when I use FLIRT it will preserve the origin of
the [refvol]. On the other hand, applyxfm4d appears to put random values
into these fields. I recognise that FSL does not typically use the origin
values, but it certainly would not hurt to retain the values of the
reference volume (this makes it easier to overlay images with programs like
SPM and MRIcro that do assume the origin is meaningful).
On Tue, 24 Aug 2004 12:39:47 +0100, Mark Jenkinson <[log in to unmask]>
>You can apply a single matrix transform to an entire 4D file
>using flirt. It also lets you select the interpolation method.
>However, if you need to use different matrices for different
>volumes then you'll need to write a tiny script to:
>(a) use avwsplit to get the individual 3D volumes,
>(b) apply the transform to each 3D volume using flirt, and
>(c) reassemble into a 4D volume using avwmerge.
>All the best,
>Chris Rorden wrote:
>>I have noticed that applyxfm4d always uses the slow sinc interpolation
>>method. It would be nice if this tool allowed the user to optionally
>>other interpolations (e.g. trilinear would be much faster and appropriate
>>for situations like Russ Poldrack's ROI tools). Does anyone know whether
>>there is a way to transform a 4D volume using a 4x4 transformation matrix
>>without using sinc interpolation?