<p><strong>B.A.T.M.A.N. - Basic And Advanced Tractography with MRtrix for All Neurophiles</strong></p>
<p>Welcome to B.A.T.M.A.N. - Basic And Advanced Tractography with MRtrix for All Neurophiles. If you want to learn about the latest approaches in diffusion weighted imaging (DWI), this is where you want to be :). Most students and researchers that have been working with DWI and/or tractography will likely have worked with diffusion tensor imaging (DTI) (Basser et al., 1994) to estimate how nerve fibers are oriented within the brain. However, the diffusion tensor model, on which DTI is based on, does not perform well in brain regions containing crossing or kissing (i.e. tangentially touching) fibers. The reason for this is that the tensor model approaches fiber orientation with an ellipsoid shape. In a region where fibers cross, the orientation estimation of the tensor model will approach a sphere and thus cannot capture the orientation of two separate fibers. This is an especially severe problem when considering that up to 90% of all brain voxels contain crossing fibers (Jeurissen et al., 2013). Several strategies have been suggested to deal with this problem, and most recently, the development of Constrained Spherical Deconvolution (CSD) (Tournier et al., 2004, 2007) has raised the attention of the field, since it evidently outperforms DTI or other alternatives in regions of crossing-fibers (Farquharson et al., 2013; Tournier et al., 2008).
Based on the improvements CSD has introduced to tractography, the main drivers of this method developed an easy-to-use and freely-available software to perform CSD-based tractography: MRtrix (visit the website on [<a href="http://mrtrix.org" rel="nofollow">http://mrtrix.org</a>]). Following the success of CSD, the MRtrix developers have put forth more algorithms to improving the biological plausibility of fiber tracking: These algorithms include Anatomically Constrained Tractography (ACT) (Smith et al., 2012), which rejects such streamlines that end in biologically implausible tissue (e.g. the cerebrospinal fluid, CSF); spherical-deconvolution informed filtering of tractograms (SIFT) (Smith et al., 2013), which corrects for the fact that longer streamlines tend to be overestimated in tractography; and finally multi-shell multi-tissue CSD (MSMT) (Jeurissen et al., 2014), which – among others – improves tractography in voxels containing partial volumes by exploiting the differences in b-value sensitivity of different tissue types. In this tutorial, we aim to make the user familiar with how to implement all those algorithms into their own tractography processing pipeline. For that, we (hopefully!) provide easily understandable instructions for each step, with ready-to-run bash commands. We also provide a tutorial dataset, using state-of-the-art scanning parameters (Jeurissen et al., 2014; Tournier et al., 2013), which we are happy to share for the purpose of this tutorial.</p>
<p>We provide two versions of the tutorial – an extended as well as a trimmed version. You can find both tutorials as pdfs in the "Files" section ("BATMAN_tutorial.pdf" and "BATMAN_trimmed_tutorial.pdf"). The extended version is admittedly pretty long, but also very informative. If you want to have background information on any of the analysis steps, you should take the time and work through that version. If, however, you are already familiar with DWI and just want to learn some MRtrix "terminology", you might rather choose the trimmed version. Importantly, the tutorial is based on some tutorial data, which we are happy to make available here as well. The tutorial data directory is called BATMAN (you will also find it in the "Files" section), and it has several subfolders. We recommend you download the entire "BATMAN" directory from the OSF webpage as a zip. Then, unzip it on your own machine. After that download either one of the tutorials as pdf (the original or the trimmed version, whatever fits you best) and there you go! If you have any issues, don't hesitate to contact me (either on OSF or via email: email@example.com<a href="http://-regensburg.de" rel="nofollow">-regensburg.de</a>).</p>
<p>Visit us also on our <a href="https://www.uni-regensburg.de/medizin/psychiatrie-psychotherapie/forschung/biomedizinische-bildgebung/index.html" rel="nofollow">lab's website</a> and on <a href="https://www.researchgate.net/lab/Jens-Volkmar-Schwarzbach-lab" rel="nofollow">researchgate</a>. </p>
<p>Note: Lately, some users have had problems with step 4.1.1 ("Preparing a mask for streamline termination"), with the command:</p>
<p>"flirt -in mean_b0_preprocessed.nii.gz -ref 5tt_nocoreg.nii.gz -interp nearestneighbour -dof 6 -omat diff2struct_fsl.mat"</p>
<p>getting an error like</p>
<p>"Image Exception : #75 :: 3D only method called by higher-dimensional volume.
3D only method called by higher-dimensional volume."</p>
<p>The problem here is that in the tutorial uses the 5tt-image as reference for the registration. This is a 4D-image, and FSL's latest version does not support a 4D-reference image any more. What you can do is either split that 4D-volume into several 3D volumes and use one of those as a new reference image (e.g. using FSL's "fslsplit"). Or probably the best way around this is to use the T1-image as a reference. For more details on this, you can take a look at the discussion on the <a href="http://community.mrtrix.org/t/mrtrix-tutorial-available-on-osf/1942/25" rel="nofollow">MRtrix Blog's Webpage</a>.</p>
<p>Note2: The files hcpmmp1_ordered.txt and hcpmmp1_original.txt had a few typos in it (see also <a href="https://github.com/MRtrix3/mrtrix3/pull/1722#pullrequestreview-289931625" rel="nofollow">https://github.com/MRtrix3/mrtrix3/pull/1722#pullrequestreview-289931625</a>). Therefore, using labelconvert with reference to this yields some warnings and will be wrong for those regions with typos... Really sorry about that! I now uploaded the correct versions, thanks to Paul Triebkorn (<a href="https://github.com/paul-triebkorn" rel="nofollow">https://github.com/paul-triebkorn</a>), Robert Smith (<a href="https://github.com/Lestropie" rel="nofollow">https://github.com/Lestropie</a>) and Donald Tournier (<a href="https://github.com/jdtournier" rel="nofollow">https://github.com/jdtournier</a>). [October 7th 2019]</p>