Hi Peter -
For various reasons, it is better not to do this. The key one being
that FA is not an unbiased estimator. Therefore white noise in your
image leads to biases in FA (not just more variance). For example, if
you average lots of different low SNR FA maps, the output will look
ok, but you will have consistently overestimated FA in low FA regions
(like grey matter) compared to computing a single FA image either
from all your data, or from your averaged data. There are also some
effects in high FA regions, but they are a bit harder to predict.
It is not a big issue if all your subjects are processed in the same
way, but it is better to use as much of your data as possible in
computing a single FA value.
T
On 9 Jun 2007, at 02:25, Kochunov, Peter wrote:
> I've averaged the FA maps for up to 8 sessions on the same subject
> and this works quite alright. Just flirt them all to one session.
> pk
>
> ________________________________
>
>
> From: David Kideckel [mailto:[log in to unmask]]
> Sent: Friday, June 08, 2007 10:48 AM
> Subject: Two separate DTI Acquisitions
>
>
>
> Dear FSLers,
>
>
>
> Four our DTI acquisitions, we have 2 separate runs completed back
> to back on each subject, same parameters, etc. What would be the
> best approach to handle this? Signal averaging? My aim is to use
> TBSS with these data.
>
>
>
> I have read several posts in the archives but it doesn't seem like
> anyone is in complete agreement.
>
> If anybody in the same situation as us has tried something that has
> worked out well, please let me know!
>
>
>
> Best,
>
>
>
> David Kideckel, PhD Candidate
|