Seeding from the thalamus

Hi all,

I’d like to run an experiment seeding from the thalamus but am concerned about the effect of ACT. What do you suggest?

I was thinking of removing the thalamus (and possibly the rest of the deep GM) from the GM and deep GM components of the 5tt image and put it into the WM component. Would this be a good strategy?


Hi Claude,

It depends on exactly what you’re trying to reconstruct; but if you want to reconstruct streamlines from the thalamus to the rest of the brain, then ACT will perform exactly as desired. This was not documented in the ACT manuscript, but I subsequently made changes here (included in 3.0_RC1 update) that permit streamline seeds to be drawn within the sub-cortical GM. The condition is that one (and only one) of the propagations from the seed point (assuming bidirectional tracking) must reach the white matter; if neither propagations reach the white matter, the streamline resides entirely within sub-cortical GM, and is hence rejected. As long as that is achieved, the tracking should otherwise behave as expected given the ACT priors. Certainly let me know if this is not the case.


Thanks Rob, I am assuming that this would also work with unidirectional tracking - assuming that the one end point reaches the white matter? To me it makes more sense to do tracking in a single direction since I want to ensure that the seed point is the start of the “tract” (though I do like that you ensure that one and only one streamline must reach the white matter.


It should – not sure though, @rsmith with no doubt confirm. But…

This depends entirely on how you set your seeds. If you have independent information as to what constitutes the ‘start of the tract’, and that’s the precise location within the thalamus where you want to place your seeds, then that sounds fine to me.

But if you’re planning on seeding from the whole thalamus, whether with random or grid seeding, then there will be no relationship between the location of these seed points and that of the start of the tract – in this case I think it would be more defensible to use bidirectional tracking within the ACT framework, and rely on its truncation procedure to refine the termination point within subcortical GM (case 6, figure 2 in the original ACT paper). It won’t be perfect, in that there is still no guarantee that the thalamus-side termination is genuinely the start of the tract, but it will at least be based on objective criteria that aim to identify this point – and it’ll definitely be much better than random / uniform locations.

Yes, I have a precise location within the thalamus where I would like to start my tracking from.