If you need more info, would you like me to post your query on [email protected]? This list tends to be more caret5 inquiries, although I suppose it also supports Caret workbench. Questions about the HCP dataset and pipeline clearly fall under the [email protected] scope.
On May 7, 2013, at 5:40 PM, Timothy Coalson <[email protected]> wrote: > There is downsampling, yes, but not by mesh decimation. We have an atlas > with surfaces at 164k and 32k that uses spheres generated specifically for > regular topology and even spacing, registered individuals to the atlas, and > resampled the surfaces to that mesh using that registration. > > I don't know that much about the data itself. > > Tim > > > > On Thu, Apr 25, 2013 at 8:29 AM, Matthias Ekman <[email protected]> > wrote: > Dear all, > > my apologies in case this is not the right list to ask my question. > > I recently attended a presentation given by David, talking about functional > connectivity analysis performed on the surface. To me that made a lot of > sense and caught my interest. I downloaded the connectome-workbench > application and the HCP tutorial dataset [1]. Thanks for making this freely > available, the application looks very promising. > > The HCP tutorial dataset contains a functional gifti file > "rfMRI_REST1_LR_s2.atlasroi.L.32k_fs_LR.func.gii" with 32492 vertices and > 1200 time-points. > > This is the point where I was wondering, how you guys managed to reduce the > number of vertices to such a low number (32k per hemisphere). In my pipeline, > using freesurfer [2], I end up with about 127k vertices, which makes the > analysis computationally much more demanding. > > I assume, that the analysis scripts are not freely available yet, but it > would help me already to get an idea whether the processing contains any > downsampling steps of the data (e.g. in freesurfer it is possible to decimate > the mesh in order to reduce the number of vertices). > > Any pointer into that direction would be of great help and much appreciated. > > Related to that, I noticed that the provided time-series data contains in the > 2nd half (>Volume 600) some very "ugly" looking spikes [3]. But I assume you > guys are already aware of that. > > regards, > Matthias > > > [1] http://humanconnectome.org/connectome/get-connectome-workbench.html > [2] this basically involves surface reconstruction of the T1 (recon-all) > followed by a projection of the functional data onto the white matter mesh > (mri_surf2surf) > [3] https://dl.dropboxusercontent.com/u/38470419/spikes.png > _______________________________________________ caret-users mailing list [email protected] http://brainvis.wustl.edu/mailman/listinfo/caret-users
