3dtrackID outputs .niml.dset

Hi there,

I just finished running 3dTrackID. And wanted to analyze the data. However, now what I wanted to do is to analyze the probabilistic tracts (probability of a path) running from one ROI to another.
I read in the documentation online for this command that the FILE.niml.dset contains the matrix information on the connectivity.
When I looked at the file I couldn’t understand the matrix. It wasn’t showing any numbered ROI’s at the top or anything like that.

Before visualizing the data I wanted to take a look at it myself, but I can’t seem to understand the ouput in the file FILE.niml.dset. It seems to be values all over the place.
I’ve attached what the file looks like and I’m having a hard time interpreting the values. Are these the probabilities a tract exists there? And if so, where are the coordinates of the location?

Hi, Sondos-

The *niml.dset format files are mainly for SUMA to read. The human-readable versions with same matrix/tracking result information are in *grid files.

For doing some stats with those with 3dMVM, you can look at the FAT_MVM demo (scripts+example data):


In the newest FATCAT_DEMO2, there are also some examples of viewing those files with fat_mat_sel.py:



The thing about the grid files is that there is information on the number of tracts, fractional number of tracts, physcial volume of tracts, etc but we are not given probabilities of connectivity between each ROI region. I was expecting that maybe FATCAT would output the probabilities of SC between each pair of brain regions.

But I guess what you’re saying is: the file.niml.dset does contain the probabilistic DTT information but it’s not meant to be readable by users, right?
Whereas the .grid files contain additional information to the probabilistic values (ie. NT, PV, NV)

Hi, Sondos-

The *.niml.dset and *.grid files contain the same information (same matrices, etc.), just in different formats.

The FATCAT tracking doesn’t output a quantity “proability of connectivity” between regions. 3dTrackID estimates the most likely locations of WM associated with pairs of target ROIs, based on the diffusion data. Within those voxels, quantities are calculated (each WM connection’s average FA, average MD, number of voxels, average bundle length, etc.). For example, see pages 49 and following here:

I don’t think any tracking program can provide probabilities of connectivity. I don’t see a method for how that could be reliably estimated from this kind of information.