I also found that the volume of both are different... $ fslstats mask_1mm.nii.gz -V 2964388 2964388.000000 $ fslstats test.nii.gz -V 388896 3111168.000000 2013/2/25 Tseng Mark <[log in to unmask]> > I used fslinfo to check the mask_1mm and test and the information is below: > > $ fslinfo mask_1mm.nii.gz > dim1 157 > dim2 189 > dim3 156 > dim4 1 > datatype 16 > pixdim1 1.0000000000 > pixdim2 1.0000000000 > pixdim3 1.0000000000 > pixdim4 0.0000000000 > > $ fslinfo test.nii.gz > data_type FLOAT32 > dim1 91 > dim2 109 > dim3 91 > dim4 1 > datatype 16 > pixdim1 2.0000000000 > pixdim2 2.0000000000 > pixdim3 2.0000000000 > pixdim4 0.0000000000 > > > 2013/2/25 Tseng Mark <[log in to unmask]> > >> Hi Mark, >> >> I then did: >> >> flirt -in mask_1mm -ref >> /usr/local/fsl/data/standard/MNI152_T1_2mm_brain.nii.gz -applyxfm -init >> /usr/local/fsl/etc/flirtsch/ident.mat -out test >> >> Then I open test with standard 2mm MNI template (attached fig). Why is it >> shifted? >> >> The mask_1mm is actually a gray matter mask from VBM analysis implemented >> in SPM (It has been normalized to standard space with voxel size of 1x1x1mm >> during VBM analysis). I want to use it to mask a fMRI map in FSL. Do I need >> to do something else? >> >> Thanks again. >> >> 2013/2/25 Mark Jenkinson <[log in to unmask]> >> >>> Hi, >>> >>> You can find the identity matrix in: >>> $FSLDIR/etc/flirtsch/ident.mat >>> which has the contents: >>> >>> 1 0 0 0 >>> 0 1 0 0 >>> 0 0 1 0 >>> 0 0 0 1 >>> >>> There is also a button for this on the ApplyXFM menu. >>> And by default the identity matrix (representing no translation or >>> rotation) is used when you leave out the transformation matrix in a >>> resampling step. >>> >>> All the best, >>> Mark >>> >>> >>> On 25 Feb 2013, at 17:37, Tseng Mark <[log in to unmask]> >>> wrote: >>> >>> Hi Mark, >>> >>> My mask_1mm is from T1 structure image and has been transformed to a >>> 1x1x1mm standard space. I want to use it to mask a fmri map (mask_2mm), >>> which is in 2x2x2mm standard space. >>> >>> Since you suggested "If the images are in the same space (e.g. all the >>> MNI templates of different resolutions are all in the same space) then you >>> can use an identity matrix for the transformation," what is the identity >>> matrix you mean here? >>> >>> Thanks in advance. >>> >>> >>> >>> >>> 2013/2/25 Mark Jenkinson <[log in to unmask]> >>> >>>> Hi, >>>> >>>> I thought a summary might be helpful here: >>>> >>>> - If the images are in different spaces then you need to use an >>>> appropriate transformation matrix (as derived from a registration) in order >>>> to make sure that the post-transformed image is aligned correctly (e.g. >>>> example_func2standard) >>>> >>>> - If the images are in the same space (e.g. all the MNI templates of >>>> different resolutions are all in the same space) then you can use an >>>> identity matrix for the transformation >>>> >>>> - The reference image controls both the resolution and the FOV of >>>> the output image, and both of these must be the same in order to overlay >>>> images in FSLView >>>> >>>> - I personally do not recommend using -applyisoxfm as it often does >>>> not match the FOV correctly. It is usually better to choose an appropriate >>>> reference image (either from the standard template directory or from >>>> amongst the appropriate images in your study, e.g. example_func or highres) >>>> >>>> - If you are transforming a mask between spaces then we recommend >>>> using trilinear interpolation followed by a threshold and binarise >>>> operation with fslmaths. To keep the mask a similar size use a threshold >>>> of 0.5. >>>> >>>> - You can use applywarp instead of flirt if you prefer (it is a >>>> little easier and does not need the -noresampblur option) and it can also >>>> be used with nonlinear warps. >>>> >>>> I hope this information helps clarify things a bit. If you are still >>>> unsure I recommend looking at the FSLCourse lecture and practical on >>>> registration. >>>> >>>> All the best, >>>> Mark >>>> >>>> >>>> On 23 Feb 2013, at 15:48, Tseng Mark <[log in to unmask]> wrote: >>>> >>>> The EPI image had been normalized to standard MNI space during >>>> preprocessing, which is 2x2x2. >>>> >>>> Mask_1mm is a T1 mask that is 1x1x1, which is the whole brain gray >>>> matter mask. >>>> >>>> We want to calculate a cluster size threshold to correct for our EPI >>>> image. This threshold is going to be calculated only in whole brain gray >>>> matter by 3dClustSim in AFNI. That's why we want to transform Mask_1mm to >>>> standard space. >>>> >>>> 2013/2/23 Chou Paul <[log in to unmask]> >>>> >>>>> Dear Mark >>>>> >>>>> One more following question, in previous email, you said your >>>>> mask_1mm is locate in individual native space. is the script you used (flirt >>>>> -in mask_1mm -ref /usr/local/fsl/data/standard/MNI152_T1_2mm_brain >>>>> -applyisoxfm 2 -interp nearestneighbour -out mask_2mm_int) correctly >>>>> for EPI image (which located in MNI space)? >>>>> >>>>> Best >>>>> >>>>> Paul >>>>> >>>>> ------------------------------ >>>>> Date: Sat, 23 Feb 2013 11:57:12 +0000 >>>>> From: [log in to unmask] >>>>> Subject: Re: [FSL] changing voxel size in a mask >>>>> >>>>> To: [log in to unmask] >>>>> >>>>> Hi Wolf, >>>>> >>>>> Yeah I think it's the interpolation process. I then tried 2 methods >>>>> below, which yielded the same result: >>>>> >>>>> flirt -in mask_1mm -ref >>>>> /usr/local/fsl/data/standard/MNI152_T1_2mm_brain -applyisoxfm 2 -interp >>>>> nearestneighbour -out mask_2mm_int >>>>> >>>>> and >>>>> >>>>> flirt -in mask_1mm -ref >>>>> /usr/local/fsl/data/standard/MNI152_T1_2mm_brain -applyisoxfm 2 >>>>> -noresampblur -out mask_2mm_nor >>>>> >>>>> mask_1mm >>>>> 1226669 / 1226669 >>>>> >>>>> mask_2mm_int >>>>> 153206 / 1225648 >>>>> >>>>> mask_2mm_nor >>>>> 153206 / 1225648 >>>>> >>>>> So now the voxel size in mask_2mm is 1/8 that in mask_1mm, and their >>>>> volume is nearly the same! >>>>> >>>>> 2013/2/23 wolf zinke <[log in to unmask]> >>>>> >>>>> Hi, >>>>> >>>>> When using flirt for downsampling, it does apply a blurring which can >>>>> be switched off by the option -noresampblur. Anyway, I guess resampling >>>>> data requires some kind of interpolation, and when resampling binary data >>>>> you have to apply a threshold and binarize the mask again. If you just >>>>> binarize the blurred unthresholded mask, you likely end up with a slightly >>>>> dilated mask. >>>>> >>>>> For your kind of downsampling you might look into the fslmaths option >>>>> -subsamp2. Since you apply it on a mask, it does not matter, in which way >>>>> the data of the old voxel is averaged into the lower resolution voxel. >>>>> >>>>> good luck, >>>>> wolf >>>>> >>>>> >>>>> On 02/23/2013 03:02 AM, Tseng Mark wrote: >>>>> >>>>> Thank you Paul. >>>>> >>>>> My mask_1mm is a T1 structure image and I want to use it to mask a >>>>> fmri map, which is in 2x2x2 dimension. If I want to transform mask_1 into >>>>> the space of the fmri map, I need the transformation matrix produced during >>>>> the preprocessing of my fmri data. Right? >>>>> >>>>> The problem is that the matrix data are not with me now....that's >>>>> why I tried using "flirt -applyisoxfm" but unsure if the problem I raised >>>>> is "normal" or not..... >>>>> >>>>> Mark >>>>> >>>>> 2013/2/23 Chou Paul <[log in to unmask]> >>>>> >>>>> Dear Mark >>>>> >>>>> Have you ever tried the "ApplyXFM" with identity transformation to >>>>> resample your "mask_1mm" image to "mask_2mm" image ? I think this tool may >>>>> do the job for you. >>>>> >>>>> Best >>>>> >>>>> Paul >>>>> >>>>> > Date: Sat, 23 Feb 2013 00:45:57 +0000 >>>>> > From: [log in to unmask] >>>>> > Subject: [FSL] changing voxel size in a mask >>>>> > To: [log in to unmask] >>>>> >>>>> > >>>>> > Hi, >>>>> > >>>>> > I have a mask (mask_1mm) with voxel size of 1x1x1mm. I want to use >>>>> it to mask a 2x2x2mm map so I use below command to change it into 2x2x2 >>>>> (mask_2mm): >>>>> > >>>>> > flirt -in mask_1mm -ref >>>>> /usr/local/fsl/data/standard/MNI152_T1_2mm_brain -applyisoxfm 2 -out >>>>> mask_2mm >>>>> > >>>>> > Then I check the voxel number/volume of the 2 masks: >>>>> > >>>>> > mask_1mm.nii.gz >>>>> > 1226669 / 1226669 >>>>> > >>>>> > mask_2mm.nii.gz >>>>> > 181444 / 1451552 >>>>> > >>>>> > I found the voxel size in mask_2mm is not 1/8 that in mask_1mm, and >>>>> the volume of the two masks are different. >>>>> > >>>>> > Is it normal? >>>>> > >>>>> > Thanks. >>>>> > >>>>> > Mark >>>>> >>>>> >>>>> >>>>> >>>>> >>>> >>>> >>> >>> >> >