[Mrtrix-discussion] Difficulty in estimating ODFs using CSD on the NKI_enhanced sample

[A.S. KANAAN]

Hi Dr. Touriner,

I am currently running a Diffusion analysis on the
NKI_enhanced<https://www.dropbox.com/home/NKI_enhanced>DWI sample
using CSD for estimating the fODFs.
This is the data acquisition protocol:
http://fcon_1000.projects.nitrc.org/indi/pro/eNKI_RS_TRT/DIff_137.pdf .

In my process, I have used different maximum harmonic degrees (lmax 6,8,14)
for estimating the response function and running CSD. The profile of the
estimate response function looked normal in all three cases and in three
different subjects
(Example<https://docs.google.com/file/d/0B836ny4-YXrTR2lMWEREazFGbTA/edit?usp=sharing>),
but the fODF estimates were problematic. In all three subjects (ages:
46,36,21) several voxels on only one side of the Corpus Collosum
contained weird
fODFs representing crossing
fibers.<https://docs.google.com/file/d/0B836ny4-YXrTUUludXhsVzU1S00/edit?usp=sharing>As
evident from the glyphs, tractography using both the deterministic and
probabilistic algorithms
(SD_STREAM<https://docs.google.com/file/d/0B836ny4-YXrTTEFablowQjRKQkk/edit?usp=sharing>and
SD_PROB<https://docs.google.com/file/d/0B836ny4-YXrTUkVyM1NsOG9Sbk0/edit?usp=sharing>)
did not generate anatomically sounds tracts. I tried using different radii
of curvature and step lengths but the results were similar.

Here is my process...

Masks: A single fiber voxel mask (SF_mask.mif) and a White matter mask
(WM_mask.mif) were generated from the FA map at thresholds of 0.7 and 0.2
respectively.
ERF:  >> estimate_response -grad grad.txt -lmax 8 DWI_ED.nii SF_mask.mif
ERF_lmax8.txt
CSD: >> csdeconv -grad grad.txt -mask WM_mask.mif -seed WM_mask.mif -lmax 8
DWI_ed.nii ERF_lmax8.txt CSD_lmax8.mif

Looking at the RAW data, we noticed motion artifacts, some noise and
straight lines that cross the brain on coronal sections (see link below).

Have you encountered this sort of a problem before for an acquisition with
more than 100 diffusion encoding gradient directions? Is my process sound
or do you think its a problem with the RAW data?

You can download the raw data as well as my results
here<https://www.dropbox.com/sh/xbh77szqmlmrkoj/pAPjPDXcnd>.


Viele Grüße,
Ahmad Kanaan
__________________________________________________
Max-Planck-Institut für Kognitions- und Neurowissenschaften
Stephanstrasse 1a
04103 Leipzig, Germany
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.nitrc.org/pipermail/mrtrix-discussion/attachments/20130712/616b179d/attachment.html