The surface grayordinates simply don’t exist in a 3D cartesian space, they exist on a 2D surface mesh made up of triangular tiles and their neighbors are determined by the topology of that mesh. Most vertices have six neighbors, but a very few have 5 (also those along the medial wall may have fewer valid neighbors for a different reason). You find the neighbors of surface grayordinates by looking at the topology section of a GIFTI surface file. The first data array of a GIFTI surface file are the 3D coordinates of each vertex, the second data array are the topological relationships between the vertices that form each triangular tile. Individual subject standard mesh surfaces may have different 3D coordinates across subjects, but always have the same topology and hence are said to be on the same mesh. It is this standard mesh that is the equivalent of a standard voxel grid in volume space. We provide the command wb_command -surface-geodesic-rois that you can use to find the nearest surface vertices within some number of mm of a given vertex. There is the command wb_command -surface-geodesic-distance which you can use to compute the distances between a vertex and the rest of the surface (you can limit this to a max distance to speed it up). Finally, there is the command wb_command -cifti-export-dense-mapping that will tell you which grayordinate indices correspond to which surface vertex indices or subcortical voxels. Perhaps one or more of these will address your request.
If for some reason you need to know about an individual subject’s voxels, you can work with the data before it has been resampled to grayordinates space and then resample any results to grayordinates space after you’re done (however you’ll want to be sure you are only working with voxels inside the greymatter ribbon). You’ll also need to be careful you don’t accidentally average across subjects in volume space or smooth the data in volume space. At some point we’ll make it possible to completely reverse the grayordinates projection (e.g. go from group average results in grayordiantes space to individual subject native or MNI volume space), but this has not yet been a high enough priority for our developers, as most people don’t need to run their analysis in that direction. I’ll be honest that despite your lengthy e-mail, I’m not really clear what you are doing and hence why you need what you are requesting. If you don’t want to post sensitive details in public, you can e-mail me off the list. If you don’t want to tell me specifically what you’re up to, I’m more limited in my ability to help. Peace, Matt. On 3/28/15, 8:44 AM, "Alpay Özcan" <al...@vt.edu> wrote: >Matt, >The focus of the project is the resting state data, at this point >all of the requests are concerned about these data sets, rfMRI, no >other data sets are involved. > >As you indicated in your previous e-mail, in rfMRI data >'cross-subject correspondence is maintained through having the same >vertex number mean the same anatomical location in each subject.' > >My basic question remains the same: If I pick up one grayordinate, >I would like to know analytically or in a Cartesian manner which >anatomic vertex or voxel this grayordinate belongs to. > >I would like have the location information of the grayordinate on the >91x109x91 data set, which has cross-subject correspondence. > >Stated in an another manner, when I load the whole grayordinate data >with home build code, it is a bunch of time series but I don't exactly >know which series belong to which anatomical location. > >This becomes extremely important when one wants to automatically chose >regions of interest around a voxel. To be able to pick neighbors using >grayordinates, I need to know where the neighbors are located in >grayordinates. For example, from the grayordinate description in your >paper, it is quite possible to jump from left cortex to the right cortex >by incrementing by one around the 30Kth grayordinate. Surely, those >voxels are not neighbors. > >I went through the discussion list. There are suggestions that point >to moving back and forth between software (e.g., freesurfer, workbench >etc.) when selections are hand made. In my case, this will not >be feasible as I am trying to home build an automated pipeline. So >a translation table defining rigorously seems to be the appropriate >solution. > >Grayordinates are extremely useful. In addition to space saving, thanks >to them the need to run cortex segmentation algorithms have been >eliminated. This also eliminated potential ambiguities >that would be created by various segmentation algorithms. This is >in the same spirit as the atlas registered fMRI data for the connectome >project. > >For the grayordinates to reach their full potential, it would be >really helpful to have their Cartesian description available. >As the grayordinate data are constructed from the 91x109x91 full set >by way of a mask(?) and appropriate ordering afterwords, I am hoping >that this will not be a big bother. > >With my apologies for the iteration, I was wondering if there is a >translation available in a general readable format such as the ones >I indicated earlier: > >translation table, for example >42,50,50 ---> 1st entry in grayordinates >43,50,50 ---> 2nd entry in grayordinates > >or > >a 91x109x91 mask containing grayordinate numbers at the appropiate >voxels and zeros elsewhere > >or > >any other appropriate common format. > > >thanks in advance for your help, >Alpay > >On 27-Mar-15 12:44, Glasser, Matthew wrote: >> That will depend on the surface you choose to use, as the cortical >> grayordinates are particular surface vertices on a standard surface mesh >> topology. The 3D volume coordinates of these vertices can change across >> subjects, but cross-subject correspondence is maintained through having >> the same vertex number mean the same anatomical location in each >>subject. >> Perhaps you can give me some more details on what you are doing and I >>can >> be more helpful then. >> >> Peace, >> >> Matt. >> >> On 3/27/15, 11:25 AM, "Alpay Özcan" <al...@vt.edu> wrote: >> >>> Matt, >>> thank you for your quick reply. >>> >>> Just to clarify, my original question was not implying anything >>> negative about the grayordinates nor any desire to work in MNI space. >>> In fact, in my view, grayordinates are logically addressing many space >>> constraints and analysis problems appropriately. >>> >>> To rephrase the question, I would like to obtain the definition >>> of the grayordinates in 91x109x91 volume. The description does not need >>> to be in the form of a matrix, e.g., a coordinate translation map would >>> do. For example (I am making up the numbers): >>> >>> 42,50,50 ---> 1st entry in grayordinates >>> 43,50,50 ---> 2nd entry in grayordinates >>> etc. >>> >>> This is more rigorous than, for instance, saying the grayordinates >>>start >>>from somewhere in the left cortex. The translation map would help a lot >>> when CIFTI data are analyzed as you suggested in your reply. >>> >>> I hope this is not a big bother. >>> >>> thanks again, >>> >>> Alpay Özcan, D.Sc. >>> Research Assistant Professor >>> Arlington Innovation Center: >>> Health Research >>> Virginia Polytechnic Institute >>> and State University >>> 900 N. Glebe Road >>> Arlington VA, 22203, USA >>> >>> Tel: (571) 858-3204 >>> http://aic.ncr.vt.edu/~alpay >>> >>> On 27-Mar-15 11:54, Glasser, Matthew wrote: >>>> The whole point of grayordinates is not to work in MNI volume space, >>>>but >>>> to work in a combined cortical surface and subcortical volume space >>>>with >>>> better grayordiante-wise correspondence across subjects. The >>>> improvements >>>> gained with grayordinates are simply not possible to achieve in MNI >>>> volume >>>> space (for sheet-like cortical regions) because of the differences in >>>> geometry between a sheetlike structure and a globular nucleus and the >>>> fact >>>> that people simply don¹t have topologically corresponding folding >>>> patterns >>>> over most of the brain and areas are not always on the same places on >>>> folds. >>>> >>>> If your analysis does not require explicit spatial neighborhood >>>> information (most kinds of analysis do not require this) you can >>>>simply >>>> analyze the CIFTI data as a matrix (which you can convert to and from >>>> CIFTI and other formats like NIFTI using wb_command -cifti-convert). >>>> For >>>> example, one can run FSL¹s melodic tool (which does not support CIFTI >>>> but >>>> also does not care about spatial neighborhood relationships) by >>>> converting >>>> the CIFTI data to NIFTI, running melodic, and then converting the >>>> results >>>> from NIFTI back to CIFTI. If you do require explicit spatial >>>> neighborhood >>>> information, have a look at the commands available in wb_command that >>>> work >>>> natively on CIFTI files (e.g. finding clusters, smoothing, gradients, >>>> etc.), or write your own (you will need to interpret the surface >>>> topology >>>> of the 32k standard meshes). >>>> >>>> Peace, >>>> >>>> Matt. >>>> >>>> On 3/27/15, 10:42 AM, "Alpay Özcan" <al...@vt.edu> wrote: >>>> >>>>> Hello, >>>>> There is a need for understanding where exactly grayordinates >>>>>locations >>>>> are placed. The descriptions in the Neuroimage 80 papers (Glasser et. >>>>> al) are nice and helpful but for computational purposes an analytical >>>>> definition is necessary. My search for such definition of the >>>>> grayordinates did not yield much. >>>>> >>>>> With my apologies in advance if the solution already exists, would it >>>>> be >>>>> possible to generate a coordinate volume cube (or 3D matrix) in the >>>>>MNI >>>>> space 91x109x91 where the voxel with the 1st grayordinates will >>>>>have a >>>>> value of 1, 2nd grayordinates have a value 2, 3rd 3 etc. and the >>>>> non-grayordinate voxels will have 0 values? >>>>> >>>>> An accessible format (raw, NIFTI) would also be extremely helpful. >>>>> >>>>> If there is already a solution, pointers will be appreciated. >>>>> >>>>> Thanks in advance for your help. >>>>> >>>>> Best regards, >>>>> >>>>> -- >>>>> Alpay Özcan, D.Sc. >>>>> Research Assistant Professor >>>>> Arlington Innovation Center: >>>>> Health Research >>>>> Virginia Polytechnic Institute >>>>> and State University >>>>> 900 N. Glebe Road >>>>> Arlington VA, 22203, USA >>>>> >>>>> Tel: (571) 858-3204 >>>>> http://aic.ncr.vt.edu/~alpay >>>>> _______________________________________________ >>>>> HCP-Users mailing list >>>>> HCP-Users@humanconnectome.org >>>>> 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 >>> HCP-Users@humanconnectome.org >>> 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. >> ________________________________ 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 HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users
