Apodized disc function #2748
Apodized disc function #2748
Conversation
Provides API equivalent to that already available for generating Apodized Point Spread Functions, but that instead produces a disc of minimal thickness with minimal side lobes.
For first iteration, instead of using ZSH coefficients "1 -1 1" for response function, use the zonal Spherical Harmonic coefficients corresponding to an apodized Disc Function with lmax=4.
|
|
Conflicts: core/math/SH.h python/lib/mrtrix3/dwi2response/tournier.py
|
clang-tidy review says "All clean, LGTM! 👍" |
|
clang-tidy review says "All clean, LGTM! 👍" |
Error introduced in creation of merge 2c6c831.
|
clang-tidy review says "All clean, LGTM! 👍" |
|
RE Here's evolution of the RF across iterations with and without the proposed change.
PR: In this test subject they converge to selection of exactly the same set of voxels. In terms of RF magnitude, there's a strictly monotonic decrease on Subjectively, it seems to me that the set of single-fibre voxels selected following spherical deconvolution using the aDF is closer to the final selection than that chosen based on the current So personally I'm convinced enough to apply the change to |
|
Given absence of feedback to the contrary, and the fact that there's already supra-threshold differences in other |
Just a little tidbit I thought of as part of a collaboration project.
Replicated the same optimisation process as was used for the refined aPSF RH coefficients in the TODI manuscript; but instead of being 1.0 at [0.0, 0.0, 1.0] and 0.0 everywhere else, it is instead 1.0 at [*, *, 0.0] and 0.0 everywhere else. Produces flat discs that have minimal ringing / zero values.
It is primarily just a new class available in the API. I am however proposing one behavioural change: that this is a better choice compared to a flat disc (with severe ringing) for the first iteration of the
dwi2response tournieralgorithm. Don't know the extent of testing that people may demand to see to accept such a change?