Hi Vivek - I suspect that you are specifying the DWI dicom of one of the subjects, and trac-all uses the default 60-direction Siemens gradient table. It's hard to know without seeing your configuration file.

How are you merging the DWIs of the two subjects? You only mention how you co-register the T1s. Co-registering DWIs is problematic, because the intensity values in those images encode the amount of water diffusion in specific directions and, after you've transformed the volumes to a different space, these directions are no longer valid. It is preferable to analyze the DWIs in each subject's native space and then transform the outputs of that analysis (ODFs, tensors, FA maps, etc), instead of transforming the DWIs themselves. Ultimately, the philosophy of TRACULA is to do the tractography and extract stats in each subject's native space. You are trying to circumvent that, and I'm not sure that will work well.

Best,
a.y

From: freesurfer-bounces@nmr.mgh.harvard.edu [freesurfer-bounces@nmr.mgh.harvard.edu] on behalf of Vivek Subramanian [vivek.subramanian@duke.edu]
Sent: Tuesday, December 06, 2016 3:57 PM
To: freesurfer@nmr.mgh.harvard.edu
Subject: [Freesurfer] Treating multiple subjects as a single subject in TRACULA

Hi,

I have DWI data from two subjects acquired using b-values of 0 for the B0's and 700 for the DWI's. There are 10 B0's and 60 DWI's in each dataset. I was successfully able to follow the instructions to use TRACULA on data from a single subject (recon-all; then trac-all -prep, -bedp, and -path), but I now want to infer a "group tractography" by running TRACULA on data from both subjects in the group as if they were from the same subject. This should yield one tractography for the entire group.

I have registered all T1's to a common MNI template so they have the same dimensions. Cortical reconstruction was performed on the average of the subjects' T1s. When I run TRACULA, however, rather than pulling the bvecs from the DWI's, it takes them from a file called gradient_mgh_dti60.gdt, which is located under the Freesurfer installation. This file only contains 60 diffusion gradient directions. These closely match those of the 60 DWI files from one subject (but are not exactly the same), but because the contents of gradient_mgh_dti60.gdt are only being copied over once to the bvecs file in the subject directory, the bvecs file has only 60 sets of entries (+10 sets of 0's for the set of B0's). In other words, the file contains three rows of data (one for each spatial dimension), each with 10 zeros followed by 60 bvec entries. I expect there to be three rows of data, with 10 0's, followed by 60 bvec entries, followed by 10 0's, followed by 60 bvec entries. The bvals file, however, is fine; i.e. there are 10 0's, followed by 60 700's, followed by 100 0's, followed by 60 700's.

This causes the -prep phase of trac-all to exit with errors when dtifit fails because the bvals and bvecs files don't have the same number of entries. I'm wondering what I need to do to get around this issue so the data can all be treated as if it came from the same subject and the bvecs are taken from the DWI's rather than from the gradient_mgh_dti60.gdt file.

If you'd like, I can upload the bvecs and bvals files that trac-all -prep generates and also the output stream to the terminal.

Thank you,

Vivek