Frequently Asked Questions (FAQ)

This page contains a collection of topics that are frequently raised in discussions and on the community forum. If you are seeking an answer to a question that specifically relates to MRtrix3 performance issues or crashes, please check the relevant documentation page.

Processing of HCP data

We expect that a number of users will be wanting to use MRtrix3 for the analysis of data from the Human Connectome Project (HCP). These data do however present some interesting challenges from a processing perspective. Here I will try to list a few ideas, as well as issues that do not yet have a robust solution; I hope that any users out there with experience with these data will also be able to contribute with ideas or suggestions.

Do my tracking parameters need to be changed for HCP data?

Probably. For instance, the default parameters for length criteria are currently set based on the voxel size rather than absolute values (so e.g. animal data will still get sensible defaults). With such high resolution data, these may not be appropriate. The default maximum length is 100 times the voxel size, or only 125mm at 1.25mm isotropic; this would preclude reconstruction of a number of long-range pathways in the brain, so should be overridden with something more sensible. The minimum length is more difficult, but in the absence of a better argument I’d probably stick with the default (5 x voxel size, or 2 x voxel size if ACT is used).

Also, the default step size for iFOD2 is 0.5 times the voxel size; this will make the track files slightly larger than normal, and will also make the tracks slightly more jittery, but actually disperse slightly less over distance, than standard resolution data. People are free to experiment with the relevant tracking parameters, but we don’t yet have an answer for how these things should ideally behave.

Is it possible to use data from all shells in CSD?

The default CSD algorithm provided in the dwi2fod command is only compatible with a single b-value shell, and will by default select the shell with the largest b-value for processing.

The Multi-Shell Multi-Tissue (MSMT) CSD method has now been incorporated into MRtrix3, and is provided as part of the dwi2fod command. There are also instructions for its use provided in the documentation.

The image data include information on gradient non-linearities. Can I make use of this?

Again, unfortunately not yet. Making CSD compatible with such data is more difficult than other diffusion models, due to the canonical response function assumption. To me, there are two possible ways that this could be handled:

  • Use the acquired diffusion data to interpolate / extrapolate predicted data on a fixed b-value shell.
  • Generate a representation of the response function that can be interpolated / extrapolated as a function of b-value, and therefore choose an appropriate response function per voxel.

Work is underway to solve these issues, but there’s nothing available yet. For those wanting to pursue their own solution, bear in mind that the gradient non-linearities will affect both the effective b-value and the effective diffusion sensitisation directions in each voxel. Otherwise, the FODs look entirely reasonable without these corrections…

The anatomical tissue segmentation for ACT from 5ttgen fsl seems even worse than for ‘normal’ data…?

The combination of high spatial resolution and high receiver coil density results in a pretty high noise level in the middle of the brain. This in turn can trick an intensity-based segmentation like FSL’s FAST into mislabeling things; it just doesn’t have the prior information necessary to disentangle what’s in there. I haven’t looked into this in great detail, but I would very much like to hear if users have discovered more optimal parameters for FAST, or alternative segmentation software, for which they have been impressed by the results.

Generating Track-weighted Functional Connectivity (TW-FC) maps

This example demonstrates how these maps were derived, precisely as performed in the relevant NeuroImage paper. Assumes that you have a whole-brain tractogram named tracks.tck, and a 3D volume named FC_map.mif representing an extracted FC map with appropriate thresholding.

Initial TWI generation:

$ tckmap tracks.tck temp.mif <-template / -vox options> -contrast scalar_map -image FC_map.mif -stat_vox mean -stat_tck sum

Deriving the mask (voxels with at least 5 streamlines with non-zero TW values):

$ tckmap tracks.tck - -template temp.mif -contrast scalar_map_count -image FC_map.mif | mrcalc - 5 -ge mask.mif -datatype bit

Apply the mask:

$ mrcalc temp.mif mask.mif -mult TWFC.mif

Handling SIFT2 weights

With the original tcksift command, the output is a new track file, which can subsequently be used as input to any command independently of the fact that SIFT has been applied. SIFT2 is a little trickier: the output of the tcksift2 command is a text file. This text file contains one line for every streamline, and each line contains a number; these are the weights of the individual streamlines. Importantly, the track file that was used as input to the tcksift2 command is unaffected by the execution of that command.

There are therefore two important questions to arise from this:

How do I use the output from SIFT2?

Any MRtrix3 command that receives a track file as input will also have a command-line option, -tck_weights_in. This option is used to pass the weights text file to the command. If this option is omitted, then processing will proceed as normal for the input track file, but without taking the weights into consideration.

Why not just add the weight information to the track data?

The .tck file format was developed quite a long time ago, and doesn’t have the capability of storing such data. Therefore, combining per-streamline weighting data with the track data itself would require either modifying this format (which would break compatibility with MRtrix 0.2, and any other non-MRtrix code that uses this format), using some other existing format for track data (which, given our experiences with image formats, can be ill-devised), or creating a new format (which would need to support a lot more than just per-streamline weights in order to justify the effort, and would likely become a fairly lengthy endeavour).

Furthermore, writing to such a format would require duplicating all of the raw track data from the input file into a new output file. This is expensive in terms of time and HDD space; the original file could be deleted afterwards, but it would then be difficult to perform any operations on the track data where the streamline weight information should be ignored (sure, you could have a command-line option to ignore the weights, but is that any better than having a command-line option to input the weights?)

So, for now, it is best to think of the weights file provided by tcksift2 as accompanying the track file, containing additional data that must be explicitly provided to any commands in order to be used. The track file can also be used without taking into account the streamline weights, simply by not providing the weights.

Making use of Python scripts library

In addition to the principal binary commands, MRtrix3 also includes a number of Pyton scripts for performing common image processing tasks. These make use of a relatively simple set of library functions that provide a certain leven of convenience and consistency for building such scripts (e.g. common format help page; command-line parsing; creation, use and deletion of temporary working directory; control over command-line verbosity).

It is hoped that in addition to growing in complexity and capability over time, this library may also be of assistance to users when building their own processing scripts, rather than the use of e.g. Bash. The same syntax as that used in the provided scripts can be used. If however the user wishes to run a script that is based on this library, but is not located within the MRtrix3 scripts/ directory, it is necessary to explicitly inform Python of the location of those libraries; e.g.:

$ export PYTHONPATH=/home/user/mrtrix3/lib:$PYTHONPATH
$ ./my_script [arguments] (options)

(Replace the path to the MRtrix3 “lib” directory with the location of your own installation)

tck2connectome no longer has the -contrast X option…?

The functionalities previously provided by the -contrast option in this command can still be achieved, but through more explicit steps:

tck2connectome -contrast mean_scalar

$ tcksample tracks.tck scalar.mif mean_scalars.csv -stat_tck mean
$ tck2connectome tracks.tck nodes.mif connectome.csv -scale_file mean_scalars.csv -stat_edge mean

The first step samples the image scalar.mif along each streamline, calculates the mean sampled value along each streamline, and stores these values into file mean_scalars.csv (one value for every streamline). The second step then assigns the value associated with each streamline during connectome construction to be the values from this file, and finally calculates the value of each edge to be the mean of the values for the streamlines in that edge.

tck2connectome -contrast meanlength

$ tck2connectome tracks.tck nodes.mif connectome.csv -scale_length -stat_edge mean

For each streamline, the contribution of that streamline to the relevant edge is scaled by the length of that streamline; so, in the absence of any other scaling, the contribution of that streamline will be equal to the length of the streamline in mm. Finally, for each edge, take the mean of the values contributed from all streamlines belonging to that edge.

tck2connectome -contrast invlength_invnodevolume

$ tck2connectome tracks.tck nodes.mif connectome.csv -scale_invlength -scale_invnodevol

Rather than acting as a single ‘contrast’, scaling the contribution of each streamline to the connectome by both the inverse of the streamline length and the inverse of the sum of node volumes is now handled using two separate command-line options. The behaviour is however identical to the old usage.

Visualising streamlines terminations

I am frequently asked about Figures 5-7 in the Anatomically-Constrained Tractography article, which demonstrate the effects that the ACT method has on the locations of streamlines terminations. There are two different techniques used in these figures, which I’ll explain here in full.

  • Figure 6 shows streamlines termination density maps: these are 3D maps where the intensity in each voxel reflects the number of streamlines terminations within that voxel. So they’re a bit like Track Density Images (TDIs), except that it’s only the streamlines termination points that contribute to the map, rather than the entire streamline. The easiest way to achieve this approach is with the tckmap command, using the -ends_only option.
  • Figures 5 and 7 display large dots at the streamline endpoints lying within the displayed slab, in conjunction with the streamlines themselves and a background image. This can be achieved as follows:
    • Use the tckresample command with the -endpoints option to generate a new track file that contains only the two endpoints of each streamline.
    • Load this track file into mrview.
    • Within the mrview tractography tool, for the “Geometry” option, select “Points”.

This will display each streamline point as a dot, rather than drawing lines between each streamline point. Since this track file contains only two points per streamline, corresponding to the streamline endpoints, this means that a dot is drawn at each streamline endpoint.

In Figure 7, each streamline endpointis coloured red. Figure 5 is slightly trickier, and requires some image editing trickery:

  • Display only the streamline endpoints track file.
  • Colour each “track” according to direction (this will colour each point according to the “direction” between the two endpoints).
  • Adjust the brightness / contrast of the background image so that it is completely black, and disable all other tools/geometry (so that only the termination points are visible).
  • Take a screenshot.
  • Disable the streamline endpoints track file within the tractography tool, set up whatever background image / tracks you wish to combine within your image, and take another screenshot. Make sure to not move the view focus or resize the mrview window, so that the two screenshots will overlay directly on top of one another.
  • Open the two screenshots using image editing software such as GIMP.
  • In Figure 5, a trick I used was to take the endpoint termination screenshot, and rotate the hue by 180 degrees: this provides a pseudo-random coloring of the termination points that contrasts well against the surrounding tracks.
  • Within the image editing software, make the termination point screenshot transparent where the termination points are not drawn, and then overlay it with the second screenshot (in GIMP, you can use “Copy” -> “Paste as new layer”).

Unusual result following use of tcknormalise

Sometimes, following the use of the tcknormalise command, an unusual effect may be observed where although the bulk of the streamlines may be aligned correctly with the target volume / space, a subset of streamlines appear to converge very ‘sharply’ toward a particular point in space.

This is caused by the presence of zero-filling in the non-linear warp field image. In some softwares, voxels for which a proper non-linear transformation cannot be determined between the two images will be filled with zero values. However, tcknormalise will interpret these values as representing an intended warp for the streamlines, such that streamline points within those voxels will be spatially transformed to the point [0, 0, 0] in space - this results in the convergence of many streamlines toward the singularity point.

The solution is to use the warpcorrect command, which identifies voxels that contain the warp [0, 0, 0] and replaces them with [NaN, NaN, NaN] (“NaN” = “Not a Number”). This causes tcknormalise to _discard_ those streamline points; consistently with the results of registration, where appropriate non-linear transformation of these points could not be determined.

Encountering errors using 5ttgen fsl

The following error messages have frequently been observed from the 5ttgen fsl script:

FSL FIRST has failed; not all structures were segmented successfully
Waiting for creation of new file "first-L_Accu_first.vtk"
FSL FIRST job has been submitted to SGE; awaiting completion
  (note however that FIRST may fail silently, and hence this script may hang indefinitely)

Error messages that may be found in the log files within the script’s temporary directory include:

Cannot open volume first-L_Accu_corr for reading!
Image Exception : #22 :: ERROR: Could not open image first_all_none_firstseg
terminate called after throwing an instance of 'RBD_COMMON::BaseException'
/bin/sh: line 1:  6404 Aborted                 /usr/local/packages/fsl-5.0.1/bin/fslmerge -t first_all_none_firstseg first-L_Accu_corr first-R_Accu_corr first-L_Caud_corr first-R_Caud_corr first-L_Pall_corr first-R_Pall_corr first-L_Puta_corr first-R_Puta_corr first-L_Thal_corr first-R_Thal_corr

These various messages all relate to the fact that this script makes use of FSL’s FIRST tool to explicitly segment sub-cortical grey matter structures, but this segmentation process is not successful in all circumstances. Moreover, there are particular details with regards to the implementation of the FIRST tool that make it awkward for the 5ttgen fsl` script to invoke this tool and appropriately detect whether or not the segmentation was successful.

It appears as though a primary source of this issue is the use of FSL’s flirt tool to register the T1 image to the DWIs before running 5ttgen fsl. While this is consistent with the recommentation in the Anatomically-Constrained Tractography (ACT) documentation, there is an unintended consequence of performing this registration step specifically with the flirt tool prior to 5ttgen fsl. With default usage, flirt will not only _register_ the T1 image to the DWIs, but also _resample_ the T1 to the voxel grid of the DWIs, greatly reducing its spatial resolution. This may have a concomitant effect during the sub-cortical segmentation by FIRST: The voxel grid is so coarse that it is impossible to find any voxels that are entirely encapsulated by the surface corresponding to the segmented structure, resulting in an error within the FIRST script.

If this is the case, it is highly recommended that the T1 image not be resampled to the DWI voxel grid following registration; not only for the issue mentioned above, but also because ACT is explicitly designed to take full advantage of the higher spatial resolution of the T1 image. If flirt is still to be used for registration, the solution is to instruct flirt to provide a transformation matrix, rather than a translated & resampled image:

$ flirt -in T1.nii -ref DWI.nii -omat T12DWI_flirt.mat -dof 6

That transformation matrix should then applied to the T1 image in a manner that only influences the transformation stored within the image header, and does not resample the image to a new voxel grid:

$ transformconvert T12DWI_flirt.mat T1.nii DWI.nii flirt_import T12DWI_mrtrix.txt
$ mrtransform T1.nii T1_registered.mif -linear T12DWI_mrtrix.txt

If the T1 image provided to 5ttgen fsl has _not_ been erroneously down-sampled, but issues are still encountered with the FIRST step, another possible solution is to first obtain an accurate brain extraction, and then run 5ttgen fsl using the --premasked option. This results in the registration step of FIRST being performed based on a brain-extracted template image, which in some cases may make the process more robust.

For any further issues, the only remaining recommendations are:

  • Investigate the temporary files that are generated within the script’s temporary directory, particularly the FIRST log files, and search for any indication of the cause of failure.
  • Try running the FSL run_first_all script directly on your original T1 image. If this works, then further investigation could be used to determine precisely which images can be successfully segmented and which cannot. If it does not, then it may be necessary to experiment with the command-line options available in the run_first_all script.

How do I use atlas / parcellation “X”?

Whether dealing with individual subject data, or a population-specific template, it can be desirable to obtain spatial correspondence between your own data and some other atlas image. This includes taking a parcellation that is defined in the space of that atlas and transforming it onto the subject / template image.

Our recommended steps for achieving this are:

  1. Perform registration from image of interest to target atlas

    • Since this registration is not always intra-modal, and image intensities may vary significantly, here we recommend using FSL flirt.
    • 12 degrees of freedom affine registration is performed to account for gross differences in brain shape.
    • flirt must be explicitly instructed to provide a transformation matrix, rather than a transformed & re-gridded image.
    flirt -in my_image.mif -ref target_atlas.mif -omat image2atlas_flirt.mat -dof 12
  2. Convert the transformation matrix estimated by flirt into MRtrix3 convention

    transformconvert image2atlas_flirt.mat my_image.mif target_atlas.mif flirt_import image2atlas_mrtrix.txt
  3. Invert the transformation matrix to obtain the transformation from atlas space to your image

    transformcalc image2atlas_mrtrix.txt invert atlas2image_mrtrix.txt
  4. Apply this transformation to the parcellation image associated with the atlas

    • Due to the use of a full affine registration (12 degrees of freedom) rather than a rigid-body registration (6 degrees of freedom), it is preferable to re-sample the parcellation image to the target image voxel grid, rather than altering the image header transformation only.
    • When re-sampling a parcellation image to a different image grid, nearest-neighbour interpolation must be used; otherwise the underlying integer values that correspond to parcel identification indices will be lost.
    mrtransform target_parcellation.mif -linear atlas2image_mrtrix.txt -template my_image.mif parcellation_in_my_image_space.mif

Transformation issues with the NIfTI image format

In the NIfTI & NIfTI-2 (.nii) image format, there is not one, but _two_ items stored in the image header that provide information regarding the localisation of the image within physical space (the image :ref:<transform>):

  • The “qform” entry is intended to specify a _rigid_ transformation from voxel coordinates into a spatial location corresponding to the subject’s position within the scanner.
  • The “sform” entry is intended to specify an _affine_ transformation from voxel coordinates into a spatial location within the space of some template image.

While the storage of both of these fields was well-intentioned, the practical reality is that there is no robust way for any software to know which of the two transforms should be used. Furthermore, while software developers will typically ensure that the handling of these fields is internally consistent within their own software, this handling can differ _between_ packages, causing problems for those who utilise tools from multiple sources.

The handling of this information in _MRtrix3_ is as follows:

  • When _reading_ NIfTI images:

    • If _only_ _one_ of these transformations is present and valid in the image header (as indicated by the “sform_code” and “qform_code” flags within the NIfTI image header), then the valid transformation will be used.
    • If _both_ sform_code and qform_code are set, then _MRtrix3_ will _compare_ these two transformations. If they are identical, then _MRtrix3_ can proceed using either data without ambiguity. If they differ, then _MRtrix3_ will issue a _warning_, and by default will use the ``qform`` data as the image transform. This is congruent with _MRtrix3_ using the convention of storing data and performing processing within the “real” / “scanner” space of the individual subject.
  • When _writing_ NIfTI images, _MRtrix3_ will:

    • Write the _same_ _transformation_ to both the qform and sform fields, based on _MRtrix3_’s internal representation of the header transformation for that particular image. This provides the least ambiguity, and ensures maximal compatibility with other software tools.
    • Set both sform_code and qform_code to 1, indicating that they both provide transformations to scanner-based anatomical coordinates.