[Mrtrix-discussion] csdeconv -normalise

[Donald Tournier]

Hi Nyoman,

To be honest, I don't recommend ever using it. Normalising to the b=0 image
might be appealing since it should in theory provide more quantitative
information, but in practice it introduces more errors than it fixes, since
even a tiny amount of partial voluming with CSF will cause a big drop in
the normalised DW signal, and hence in the normalised FOD. Also, to use it
correctly, you'd need to supply a 'normalised' response function, for
example one that you might have derived using an axially symmetric tensor
model. I expect you used the data-derived response function, and that this
is probably the reason why your FODs were so small. You could use them for
tracking, but you'd have to use a much lower -cutoff than the default value
of 0.1, since the scaling will be off by roughly the b=0 signal amplitude.

In any case, you shouldn't need to normalise if you're using a response
function derived from the same data (as per the documentation) since the
response function will also be appropriately scaled to the DW signal, and
the spherical deconvolution output will therefore inherently be normalised.
This way, your FODs won't be artificially reduced in regions adjoining CSF,
although you may then have to worry about intensity variations if you have
a significant bias field in your images.

Cheers,

Donald.


On 27 July 2012 13:02, Nyoman Kurniawan <nyoman.kurniawan at cai.uq.edu.au>wrote:

>  Dear Dondald,
> Would you be able to tell us when to use the option -normalise in csdeconv?
> If I use this option, the FOD becomes very very small and I can't use it
> for tracking.
>
> Thank you.
>
> Regards,
> Nyoman
>
>
> _______________________________________________
> Mrtrix-discussion mailing list
> Mrtrix-discussion at www.nitrc.org
> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>
>


-- 
Jacques-Donald Tournier (PhD)
Brain Research Institute, Melbourne, Australia
Tel: +61 (0)3 9035 7033
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.nitrc.org/pipermail/mrtrix-discussion/attachments/20120727/a936a59a/attachment.html