How to deal with WMH when generating the streamlines

Dear Mrtrix3 experts,

I am dealing with white matter hyperintensities (WMH) in a population of elderly people.
I’m looking around and I found several posts which have been useful, and I’m sorry because I’m opening another thread on this issue. However, I’m still unsure about the approach that I have to use when generating the streamlines (with tckgen and using msmt_csd in dwi2fod).

This is my situation:
After 5ttgen, the WMH have not being misclassified as GM. But also the WMH are not included in any other masks obtained with 55tgen (subcortical GM, CSF or WM). So, I decided to add as fifth volume the WMH lesion mask (BIANCA segmentation + manual correction, so the mask is pretty accurate).

Is this a correct approach? From here SS3T and ACT - #2 by ThijsDhollander I understand that the ACT rules are switched off for the lesion mask in the pathological volume, but I’m not sure if it would be better to not consider them at all, given that WMH are not classified in any other masks of 5ttgen. So, maybe another possibility is to run tckgen without the fifth pathological volume. But, the diffusion properties in the WMH are different from normal appearing white matter.

I’m a bit confused on which can be the best approach.

Thank you.


Dear @Ric ,

Assigning the WMHs to the lesion volume is indeed the recommended approach.

If those voxels are not assigned to any tissue type, I’m not sure any tracks will pass through there - but please test whether this is indeed the case.

If you are able to track through the WMHs without adding them to the 5TT image, then there’s probably no additional advantage to adding them to the lesion volume (but still dilate the masks slightly for optimal tracking).

All the best,

Thank you Nic for your reply.

I’ve decided to use the lesion volume, but I have a couple of questions concerning the results.

Below you can see the results from a subject with a huge WMH lesion load (I’m working with 7T data, so the WMH burden detected on FLAIR and T1 can be higher than at lower field strength). I generated 10 million streamlines and in the figure below, 5 million of them are rendered (the opacity of streamlines is set to 0.3). The picture is showing the fifth volume with the WMH mask and the streamlines.

Some tracts are not passing through the lesion, but in other portion they are passing trough and/or terminating in the lesion. I’m wondering if this can be an expected and acceptable behavior when using the lesion volume, or if I still need to refine my processing steps for this point. What do you think?

I am new to the tractography world, but I would expect streamlines to pass, enter, exit (5ttgen: Recommended approach to dealing with white matter hyperintensities (WMHs) for ACT - #2 by cbajada) all over the lesions, not just in some portions. Maybe, this information is in the 5 million streamlines that I cannot render (my graphic card is “only” 4 GB so I cannot render all the 10 million) or more simply 10 million are not enough streamlines to “cover” everything and I should push the number to let’s say 30-40 millions (or even more, I saw that sometimes also 100M is suggested?)

Thank you in advance for any clarification and suggestion.

Before the picture, I copy and paste the commands I’m using the generate and render the streamlines:

tckgen wmfod_norm.mif tracks_10mio_final.tck -act T1w_5tt_WMHlesion_coreg.mif
-seed_gmwmi gmwmSeed_coreg.mif -backtrack -select 10M -maxlength 250
-minlength 5 -cutoff 0.06 -crop_at_gmwmi -nthreads 8 -debug -force

tckedit tracks_10mio_final.tck -number 5M tracks_5M_from10M.tck -force

mrview T1w_5tt_WMHlesion_coreg.mif -overlay.colourmap 2 -tractography.load \ tracks_5M_from10M.tck -tractography.opacity 0.3 -mode 2



Hi Riccardo,

It seems like you have holes in your WMH masks, which are assigned to another tissue type (likely GM) and preventing tracking through these regions. No scaling up of your number of streamlines will fix this - it needs some filling and dilating hocus pocus and reassignment of the lesion mask.

FYI - you can always track fewer streamlines for test scenarios - if something were wrong it should show up with 1 or 2 million streamlines as well. Once you are happy with the mask, you can scale up the tractogram to as much as you want :slight_smile: