[Neuroimaging] [DIPY] Comparing DIPY tensor fitting and other tools

[Samuel St-Jean]

Dipy by default uses one iteration of weighted least square, so ordinary
least square, then a refit based on that. The difference probably comes
from implementation details in solving the system of equations as mentioned
by the previous poster, but in the end they should be mostly similar.
Largest difference should be in partial volume voxel. I think there was an
ismrm 2015 abstract discussing the issue of various least square fit on the
fa values.
On Nov 27, 2015 19:59, "Reid, Robert I. (Rob)" <Reid.Robert at mayo.edu> wrote:

> Hi,
>
>
>
> > FDT and DTIK give me identical maps (except slight difference when
> these two apply the mask. I see some voxels having FA>1 on FDT result).
> DIPY gives a map very similar FA map compared to the other two FA maps I
> have.  Shouldn’t I get an exactly same map (within rounding error maybe).
>
>
>
> You did not say how the data was weighted, so my guess is that you used
> the defaults, which are different.  FDT (or at least dtifit, which I am
> more familiar with) for historical consistency defaults to no weighting
> when it fits the tensor to ln(S/So), which is not optimal since ln(S/So) is
> heteroscedastic.  dtifit’s -w option uses a heuristic to weight the data to
> account for that.  IIRC, dipy determines the weights iteratively, which is
> theoretically better but in practice usually close to the heuristic.
>
>
>
>      Rob
>
>
>
> --
>
> Robert I. Reid, Ph.D. | Sr. Analyst/Programmer, Information Technology
>
> Aging and Dementia Imaging Research | Opus Center for Advanced Imaging
> Research
>
> Mayo Clinic | 200 First Street SW | Rochester, MN 55905 | mayoclinic.org
> <http://www.mayoclinic.org/>
>
>
>
> _______________________________________________
> Neuroimaging mailing list
> Neuroimaging at python.org
> https://mail.python.org/mailman/listinfo/neuroimaging
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/neuroimaging/attachments/20151127/c35ab9c7/attachment.html>