New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Upgrade on DKI implementations #2009
Conversation
in future we can incorporte t-designs for different number of directions
evecs for perpendicular direction calculation was not properly selected this was an issue if data is masked
mean kurtosis tensor has to be equal to standard MK in isotropic diffusion a bug in parameter indexing for mean kurtosis tensor calculation was fixed
…echniques based on cumulant expansion other cumulant expansion includes dki and future techniques such as the correlation tensor imaging (CTI)
…'s non-linear fit
subfuntions of OLS and WLS fit have to appear first than main functions minor doc corrections
clip of maximum has to be done after applying normalization factor of 1/5
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Awesome stuff. One thing that's not crystal clear to me: what is the advantage of the "numerical" solutions?
@arokem - answering your question. In theory, numerical and analytical solutions should give the exact same solution; however, numerical solution avoids some instabilities for voxels corresponding to the singularities of the analytical solution. |
…oks (non-linear and restore fits)
…tical solution tests
Codecov Report
@@ Coverage Diff @@
## master #2009 +/- ##
==========================================
+ Coverage 90.51% 91.14% +0.62%
==========================================
Files 242 246 +4
Lines 31023 31539 +516
Branches 3254 3302 +48
==========================================
+ Hits 28081 28745 +664
+ Misses 2231 2081 -150
- Partials 711 713 +2
|
Hello @RafaelNH, Thank you for updating !
|
Just updated the documentation to address issue #2003 . All done at my side. Let me know if there is something else that you want me to address. |
Off we go then! |
Hi all! Four years have passed since we've implemented the DKI module, so I've decided to do a review on its code and I've decided to give it an upgrade.
Here are some features added:
Non-linear and Restore DKI fits - Basically I've made the DTI's non-linear and RESTORE fits compatible with DKI, so now dki.py can call these fitting procedures directly from dti.py.
Added some numerical solutions to the estimation of axial, radial and mean kurtosis. Theoretically, these should give the same results than the analytical solutions that we previously implemented. However, numerical solutions can be more stable for some voxels with parameters near to the analytical solution singularities.
Our previous implementations did not have a quantification of kurtosis anisotropy. Therefore, I've decided to add the Kurtosis Fractional Anisotropy index according to Glenn et al. (2015). Also added the mean kurtosis tensor as proposed by Hansen & Jespersen (2013)
New metrics are already added in DKI example.