Hi Colin,
I see you have discovered the metric autoloader prototype that is in Caret5. This was mainly designed for our Human Connectome Project grant application & some of my early surface-constrained tractography work and is essentially a kludge to get something working quickly. We don't really support this feature (and perhaps it shouldn't have even been included in the release). The good news is that the next generation of Caret (Caret7/Connectome Workbench) has a much improved version of this feature and a new file format to hold the connectivity data. Instead of storing the connectivity data as thousands of metric files and NIFTI volumes, we store it all in single CIFTI dense Connectome files. You can read more about CIFTI on the NITRC http://www.nitrc.org/projects/cifti/ The FSL group are implementing CIFTI on their end as well, so we will have tractography outputs from probtrackx that are viewable in Caret7 as CIFTI files. As far as visualizing trajectories, this is also planned, though we are not intending to support streamlines. The reason is we would rather show trajectories by selecting the uODFs that the fibers took along the way so we can easily see what the underlying data was that drove the tractography result. The uODFs would be stored separately from the fiber counts as vector files (in fact it is possible to convert FSL formatted "dyads" vector files to something Caret5 can display, but again, this is an unsupported and undocumented feature). So, the bad news is while we are making a lot of progress on these fronts as a part of the Human Connectome Project, we don't have anything to release to you right at this moment for tractography analysis. Jenn is in charge of who will receive the next version of Caret for beta testing, so please contact her if you want to sign up. Best, Matt Glasser. Van Essen Lab. _____ From: [email protected] [mailto:[email protected]] On Behalf Of Colin Reveley Sent: Friday, December 16, 2011 12:51 PM To: [email protected] Subject: Re: [caret-users] DWI, trackvvis, metric autoload re: that, I left something out; the nodes connected by streamlines to an ROI would ideally not just "make themsleves known" they would color according to the path of the streamline (which is in the niftii volume) in the standard RGB (as in green rostrocaudal, red mediolateral, blue inferiorsuperior) by path midpoint, path mean or whatever. The niftii volume provides the coloring information. hence, giving quick, accessable clues as to what bundle of fibres might be mediating the putative connection between regions. this global sketch can then be used to guide seeds, waypoint masks and targets for more serious analysis of the DWI data, which you then express as a metric. On 16 December 2011 18:36, Colin Reveley <[email protected]> wrote: Hi - Caret has a feature "metric auto load" intended for tractography data. Essentially, for each voxel in a volume there's an associated metric. When you click a surface node, the metric for the node associated with that voxel is loaded. That suits me well, and works well for probabilistic tractography which is easily expressible as a metric, in terms of caret volume->surface operations from a niftii volume. So, cool. But, there is also merit in deterministic tractography from tensors or ODFs. It gives a global picture more cheaply (in terms of memory and computation). I guess the standard for this sort of thing is trackvis. Freesurfer has a command dmri_trk2trk what this does is take a DTK/Trackvis .trk file (which is a load of vectors describing streamlines, fancy those colored visuals you always see), applies a transform (from FLIRT) to get it into (in this case) the equivalent voxel to voxel relations as the underlying CARET anatomical image, and outputs a new trk, but also a niftii of vectors. Now is there anyway to take that vector niftii and somehow load it as connectivity (rather than metric) data? If not, similar code may be in CARET already, because of the cocomac interface. This is different, but not that different. It would be useful to many I would think. one would define an ROI of nodes, and the nodes connected to it by streamlines would make themsleves known. thanks Colin, Sussex
_______________________________________________ caret-users mailing list [email protected] http://brainvis.wustl.edu/mailman/listinfo/caret-users
