An additional note on downsampling: when using the ADAP_BARY_AREA surface resampling, you do not need to smooth before downsampling (it will not lose data that falls between the sampling points). However, with downsampling volume data, the available methods do require some presmoothing to avoid this problem.
Tim On Tue, May 5, 2015 at 4:40 PM, Glasser, Matthew <[email protected]> wrote: > I don¹t recommend using data with large amounts of spatial smoothing > (downsampling is actually a more forthright acknowledgement of what you > are doing to the data and saves space and computational time). Also, it¹s > worth thinking if your use case would be better served by using a brain > parcellation. If correctly made, a parcellation achieves the > dimensionality reduction of downsampling in a more neurobiologically > informed way. If you still need to downsample the grayordiantes space, > you can use an approach like this (example parameters were for creating an > experimental higher resolution 1.6mm grayordinates space for 7T data): > > Batch script call: > > GitRepo="/media/2TBB/Connectome_Project/Pipelines² #Where you have > installed the HCP Pipelines code > https://github.com/Washington-University/Pipelines > TemplateFolder="${GitRepo}/global/templates/standard_mesh_atlases" > NumberOfVertices=³59000² #set to your desired mesh vertex number > OriginalMesh="164" > NewMesh=³59² #what you want to call it > NewResolution=³1.6² #your desired voxel resolution > Caret7_Command=³wb_command² #Requires that you have installed connectome > workbench > SubcorticalLabelTable="${GitRepo}/global/config/FreeSurferSubcorticalLabelT > ableLut.txt" > > ${GitRepo}/CreateNewTemplateSpace.sh ${TemplateFolder} ${NumberOfVertices} > ${OriginalMesh} ${NewMesh} ${NewResolution} ${Caret7_Command} > ${SubcorticalLabelTable} > echo "set -- ${TemplateFolder} ${NumberOfVertices} ${OriginalMesh} > ${NewMesh} ${NewResolution} ${Caret7_Command} ${SubcorticalLabelTable}" > > > ${GitRepo}/CreateNewTemplateSpace.sh: > > > TemplateFolder=${1} > NumberOfVertices=${2} > OriginalMesh=${3} > NewMesh=${4} > NewResolution=${5} > Caret7_Command=${6} > SubcorticalLabelTable=${7} > > ${FSLDIR}/bin/flirt -interp spline -in ${TemplateFolder}/Avgwmparc.nii.gz > -ref ${TemplateFolder}/Avgwmparc.nii.gz -applyisoxfm ${NewResolution} -out > ${TemplateFolder}/Atlas_ROIs.${NewResolution}.nii.gz > ${FSLDIR}/bin/applywarp --rel --interp=nn -i > ${TemplateFolder}/Avgwmparc.nii.gz -r > ${TemplateFolder}/Atlas_ROIs.${NewResolution}.nii.gz > --premat=$FSLDIR/etc/flirtsch/ident.mat -o > ${TemplateFolder}/Atlas_ROIs.${NewResolution}.nii.gz > ${Caret7_Command} -volume-label-import > ${TemplateFolder}/Atlas_ROIs.${NewResolution}.nii.gz > ${SubcorticalLabelTable} > ${TemplateFolder}/Atlas_ROIs.${NewResolution}.nii.gz -discard-others > -drop-unused-labels > > ${Caret7_Command} -surface-create-sphere ${NumberOfVertices} > ${TemplateFolder}/R.sphere.${NewMesh}k_fs_LR.surf.gii > ${Caret7_Command} -surface-flip-lr > ${TemplateFolder}/R.sphere.${NewMesh}k_fs_LR.surf.gii > ${TemplateFolder}/L.sphere.${NewMesh}k_fs_LR.surf.gii > ${Caret7_Command} -set-structure > ${TemplateFolder}/R.sphere.${NewMesh}k_fs_LR.surf.gii CORTEX_RIGHT > ${Caret7_Command} -set-structure > ${TemplateFolder}/L.sphere.${NewMesh}k_fs_LR.surf.gii CORTEX_LEFT > > for Hemisphere in L R ; do > ${Caret7_Command} -metric-resample > ${TemplateFolder}/${Hemisphere}.atlasroi.${OriginalMesh}k_fs_LR.shape.gii > ${TemplateFolder}/fsaverage.${Hemisphere}_LR.spherical_std.${OriginalMesh}k > _fs_LR.surf.gii > ${TemplateFolder}/${Hemisphere}.sphere.${NewMesh}k_fs_LR.surf.gii > BARYCENTRIC > ${TemplateFolder}/${Hemisphere}.atlasroi.${NewMesh}k_fs_LR.shape.gii > -largest > ${Caret7_Command} -surface-cut-resample > ${TemplateFolder}/colin.cerebral.${Hemisphere}.flat.${OriginalMesh}k_fs_LR. > surf.gii > ${TemplateFolder}/fsaverage.${Hemisphere}_LR.spherical_std.${OriginalMesh}k > _fs_LR.surf.gii > ${TemplateFolder}/${Hemisphere}.sphere.${NewMesh}k_fs_LR.surf.gii > ${TemplateFolder}/colin.cerebral.${Hemisphere}.flat.${NewMesh}k_fs_LR.surf. > gii > done > > > > You can then create the template grayordinates space with wb_command > -cifti-create-dense-scalar <TemplateGrayordinateSpace> -volume > ${TemplateFolder}/Atlas_ROIs.${NewResolution}.nii.gz > ${TemplateFolder}/Atlas_ROIs.${NewResolution}.nii.gz -left-metric > ${TemplateFolder}/L.atlasroi.${NewMesh}k_fs_LR.shape.gii -roi-left > ${TemplateFolder}/L.atlasroi.${NewMesh}k_fs_LR.shape.gii -right-metric > ${TemplateFolder}/R.atlasroi.${NewMesh}k_fs_LR.shape.gii -roi-right > ${TemplateFolder}/R.atlasroi.${NewMesh}k_fs_LR.shape.gii > > You can use wb_command -cifti-resample from the 2mm standard grayordiantes > space to your new one (it¹s recommended that you use the individual > subject¹s midthickness surfaces as area surfaces or group average vertex > areas when doing the resampling so that you treat each square mm of cortex > the same). If you are doing a simple resampling (no registration), you > don¹t need to specify any volume transformation. CUBIC is the recommended > volume resampling method and ADAP_BARY_AREA is the recommended surface > resampling method for continuous data. If you need to resample > midthickness surfaces to use as area surfaces, you can do that with > wb_command -surface-resample (for surfaces use the BARYCENTRIC resampling > method). In most cases though, rather than downsampling parcellation is a > better approach. > > Peace, > > Matt. > > On 5/5/15, 4:15 PM, "Eric Wong" <[email protected]> wrote: > > >Hello, > >I am interested in looking at the HCP grayordinate space data, but at > >various levels of reduced resolution. My understanding is that various > >levels of in-plane smoothing are available through FSL, but that the > >resulting smoothed data is still on the full grayordinate mesh? Are there > >coarser (decimated) meshes to which one could downsample the data to > >perform lower resolution surface based analyses? I could always apply > >standard mesh decimation methods to the data, but if coarser meshes > >already exist, we may as well use the same ones. Alternatively, if I was > >to generate my own decimated meshes (which I would be happy to share), > >presumably it would make the most sense to decimate the mesh using > >whatever canonical surface geometry was used to define the original gray > >ordinate mesh in the first place. Is this surface geometry available? > >Thank You, > >Eric Wong > >_______________________________________________ > >HCP-Users mailing list > >[email protected] > >http://lists.humanconnectome.org/mailman/listinfo/hcp-users > > > ________________________________ > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you > are not the intended recipient, be advised that any unauthorized use, > disclosure, copying or the taking of any action in reliance on the contents > of this information is strictly prohibited. If you have received this email > in error, please immediately notify the sender via telephone or return mail. > > _______________________________________________ > HCP-Users mailing list > [email protected] > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
