Hi MRtrix3 experts,
For the past years we have been acquiring single-shell data, but we now want to
improve our DWI acquisition in order to use MSMT-CSD correctly.
I searched for multi-shell acquisition references and will list some of them here,
as they may be useful to others as well:
Thijs Dhollander:
What sequence parameters do you recommend for multi-shell CSD?
- 6x b=0
- 20x b=1000
- 60x b=3000
B.A.T.M.A.N. Tutorial *(not 100% certain on the b=0 total volumes):
- 8x b=0 (5 AP; 3 PA)
- 17x b=1000
- 31x b=2000
- 50x b=3000
Tahedl, M., Tournier, J.-D., Smith, R.E., 2025 — Structural connectome construction using constrained spherical
deconvolution in multi-shell diffusion-weighted MRI. Nat Protoc 20, 2652–2684.
https://doi.org/10.1038/s41596-024-01129-1
- 5x b=0 ( AP )
- 1x b=0 ( PA )
- 20x b=1000 (AP)
- 32x b=2000 (AP)
- 50x b=3000 (AP)
Given that we do not have multiband capability and have time constraints within our
overall protocol, we plan to acquire a two-shell acquisition (b=0, b=1000, b=X).
Since increasing the maximum b-value increases acquisition time, we are debating
between two options:
- Acquiring more volumes at a lower b-value (b=2000)
- Acquiring fewer volumes at a slightly higher b-value (b=2500)
Our uncertainty comes from the understanding that higher b-values require more volumes
to compensate for the reduced signal intensity. We would therefore also like to ask for
your recommendation regarding the number of volumes at b=2500.
For reference, Tahedl et al. (2025) report the following minimum volume requirements
for each shell:
- b=2000 s/mm²: 32 volumes (where lmax=6 requires ≥28 volumes)
- b=3000 s/mm²: 50 volumes (where lmax=8 requires ≥45 volumes)
With this in mind, would the following two-shell protocols be considered appropriate,
or would their quality be questioned?
Option 1 (fewer volumes, higher b-value):
- 6x b=0 (5 AP; 1 PA)
- 20x b=1000 (AP)
- 40x b=2500 (AP)
Option 2 (more volumes, lower b-value):
- 6x b=0 (5 AP; 1 PA)
- 20x b=1000 (AP)
- 60x b=2000 (AP)
We are particularly interested in which option would better support tractography using MSMT-CSD.
Thank you in advance,
Sebastián