If you see significant differences among the tracts of different
populations, can they be interpreted in any meaningful way? Are they
not just systematic mis-registration? If tracts can be exactly aligned,
then there should not be any differences.
IMHO, what would be most useful from tractography is a model of how the
brain is connected up, possibly involving some average plan of brain
connectivity. Unfortunately, such a wiring diagram would be a bit too
complicated for the results section of the average paper, so (unless we
go down the road of systems biology modelling and all that 4th paradigm
stuff) the field is limited to finding some publishable subset of the
data. This typically involves finding localisable differences among
populations.
My own favoured approach to understanding the anatomical differences
between populations would be to use the diffusion data to achieve more
accurate inter-subject alignment, and then study the shapes of the
brains. This would involve assumptions about every brain having the
same basic set of tracts, which may not be the case. The template model
may therefore need to deal with missing or extra tracts, so this would
complicate things a bit.
Best regards,
-John
On Wed, 2010-03-31 at 11:45 +0100, Richard Binney wrote:
> Hi Maria
>
> I would add that currently one of the biggest challenges for diffusion
> weighted imaging is how to perform a quantitative
> (statistical) assessment of tractography data - for example, the
> nature of the data is such that it can't really be treated with
> same the type of parametric analysis we would use for functional
> imaging data.
>
> Currently, in the case of probabilistic tractography, the usual
> approach is to merely threshold maps in order to remove false
> positives either at the individual or group level (this is often done
> using a [sometimes fairly arbritrary] predefined value or
> one emprically derived from the distribution of probabilities observed
> in the dataset). Group level maps/data are generated by averaging the
> spatially preprocessed maps or region-pair connectivity values across
> all subjects. Final results are interpreted in
> a principally qualitative way. I've yet to come across any attempts to
> take analyses further than this.......I would be interested to know if
> anyone on the SPM list has any new thoughts/ideas regarding potential
> statistical approaches......
>
> All thes best
>
> Richard
>
>
> On Wed, Mar 31, 2010 at 8:01 AM, Volkmar Glauche
> <[log in to unmask]> wrote:
> Dear Maria,
>
> the SPM Diffusion toolbox provides only a not very
> sophisticated tracking algorithm. If you want to do
> tractography (both streamline and probabilistic) and get the
> results as NIfTI, I would suggest looking at the DTI and
> Fibertools developed at the MR Physics department in Freiburg:
>
> http://www.uniklinik-freiburg.de/mr/live/arbeitsgruppen/diffusion/fibertools_en.html
>
> These tools use their own internal data format (called
> mrStruct), but results, mask images etc. can be converted
> between NIfTI and mrStruct. These tools can be integrated with
> the SPM8 batch system.
> The only limitation is that you have to import your DWI data
> directly from DICOM (i.e. you can't use Diffusion Toolbox for
> spatial preprocessing).
>
> Hope this helps,
>
> Volkmar
>
> <https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=SPM;5637c12f.1003>
>
--
John Ashburner <[log in to unmask]>
|