Dear Colleagues,
I wonder if anyone can recommend a specific pipeline for diffusion MRI data preprocessing with subsequent reconstruction of adjacency matrices. Preferably, with step-by-step instructions and straightforward strategy for scripting/running multiple subjects.
I would also appreciate if you could suggest an optimal tractography algorithm to do this for 70-direction, b=2200 s/mm2 diffusion data.
Thank you very much in advance!
I can't answer but know 2 individuals here at LLU that might know; Grant MacCauley and Sam Barnes- they can be reached by e-mail: [email protected] [email protected]
I'd like to suggest camino, a JAVA-based toolbox for dMRI.
http://camino.cs.ucl.ac.uk/index.php?n=Main.Tutorials
There is a lot of info on DTI here: http://www.diffusion-imaging.com
DTI-Prep (http://www.ncbi.nlm.nih.gov/pubmed/24523693) is a toolbox dedicated to quality control in DTI analysis worth looking into.
Dear All,
Thanks a lot for your suggestions and links! I'm currently going through Camino tutorials. Everything seems quite straightforward, so far.
Please give feedback on your experience. In this paper they test many different approaches for diffusion MRI analysis: http://www.ncbi.nlm.nih.gov/pubmed/25555998
You can try MRtrix's tools, is not tensor model but it can suit your diffusion images.
http://mrtrix.readthedocs.io/en/latest/tutorials/hcp_connectome.html
I've been looking into the Connectome Mapper, which is a sort of unified toolbox linking MRtrix, camino, and some others. It's exactly what you need in theory, but I have to tell you I've been struggling to make it work in practice. Maybe you'll have better luck?
Many thanks again for the references and suggestions!
Eventually, I ended up with the following pipeline:
1. DTIPrep*: QC, initial preprocessing, removal of bad gradients and eddy-current correction
2. FSL-FDT | BedpostX: estimation of diffusion parameters and tensor reconstruction
3. FSL-FLIRT: normalization (Do you suggest FNIRT instead?)
4. FSL-FDT| ProbTrackX or CAMINO: reconstruction of connectivity matrices
Does this look ok to you? Is there anything else you would recommend?
===
* - As far as I understood, the output from DTIPrep contains already rotated bvecs. Right?
That looks like a good protocol, but make sure the output of DTIPrep has rotated bvecs. Just open your original bvecs (e.g. in Excel) and the output bvecs of DTIPrep, and then compare them side-by-side. If the bvecs output of DTIPrep is truly rotated, then the values will be slightly different to the original bvecs. If not, then I suggest you use FSL's eddy command (available from version 5.0.9). If you have blipped data, then make sure to first run topup.
I would use mrtrix and CSD-based tractography for this kind of acquisition.
-Timo
I also recommend MRtrix3 using CSD-based probabilistic tractography. My ebook has step-by-step instructions; http://jeromemallershandbookofstructuralbrainmrianalysis.yolasite.com/?
-Jerome
I would strongly recommed MRtrix3 ( https://www.mrtrix.org/ ), which is based on constrained spherical deconvolution (CSD) instead of diffusion tensor imaging. In terms of structural connectivity construction, MRtrix further provides useful methods for correcting the biases, such as Anatomically Constrained Tractography (ACT) and Spherical-deconvolution informed filtering of tractograms (SIFT). In addition, there is also a very helpful tutorial available online: https://osf.io/fkyht/wiki/home/.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude