Hi Alessandro,
you say you tried to register a map of one IC in native space to the atlas. That won't work because there is little anatomical information in such a map.
You need to register the first functional volume of the scan to the atlas:
flirt -in <native_bold_volume> -ref <your_atlas> -out <native_bold_volume_in_atlas> -omat <native_bold_to_atlas_params>.mat
When you are happy with the spatial normalization of the native bold volume to the atlas you can apply that matrix to an IC map:
flirt -in <IC_map_in_native_space> -ref <your_atlas> -out <IC_map_in_native_space_flirt_to_atlas> -nosearch -applyxfm -init <native_bold_to_atlas_params>.mat
You can try some flirt flags and see what fits your needs: -paddingsize 0.0 and -interp nearestneighbour, etc.
Hope this helps,
koene
On Fri, 11 May 2012 17:31:41 +0200, Alessandro Bruni <[log in to unmask]> wrote:
>
> Hi, there,
>could you let
> me give a little more information about the question I posted
> earlier: the input for MELODIC was everytime a 4D image (25
> slices, every slice being 64x64, and a total of 144 timepoints)
> taken from 4 different subjects. These 4D images were all just
> smoothed(8mm) and realigned(to the first one) but not using FSL.
> Neither coregistration to any high res image nor normalization
> to any standard space were run. I just run single-session ICA on
> these 4 subjects separately, each time getting prob maps and
> z-score maps for every ICs, obviously taking into account just
> for task-related ones. Now i'm asked to compare these activation
> maps with Oxford atlas, which is characterized by 91 slices,
> each one being 91x109, dimensionally different from mine. I
> tried to register an activation map using flirt with
> MNI_2mm_brain image as reference, trying different dof, but
> everytime i got bad output...maybe i was wrong with the advanced
> options, which i left untouched since i'm not so much used with
> registration tools (the preprocessing was already done by a
> former thesist)...i'm in a deadlock...i
> hope to have you provided the right info to get me some hints...
>
> Thanks anyway,
>
> Alessandro
>
>Date: Fri, 11 May 2012 10:56:22 +0200
>From: [log in to unmask]
>Subject: [FSL] how to compare low res activation map with atlases?
>To: [log in to unmask]
>
>
>
>
>
>Hi, there.
>I'm new to FSL and i'm using it for my thesis. I think my problem is very simple for you all, and is that: i've got my activation maps using MELODIC (which were not normalised to any standard space) and now i should compare them with oxford-harvard atlas for getting the amount of activations in each cortical area. but since the dimensions are not equal (being my z-maps 64x64x25) i was wondering how may i solve this matter...i tried to run a registration using FLIRT but seems not work...
>Thanks in advance,
>Alessandro
>
|