Hi - I think there is something wrong with your processing of the
Philips data.
0.2 seems a much more likely whole-brain average FA than 0.4-0.5.
My guess is that you have artefacts in some of your volumes, causing
big signal change globally,
and therefore high FA. You should look carefully through your
volumes for the Phillips data.
Even if you find the cause of the problem, and are able to solve it,
you need to be very careful doing a study across different scanners.
slight differences in SNR or protocol parameters can cause
significant changes in FA across a group. You should certainly ensure
that your two groups are balanced across scanners (i.e. you have the
same numbers of controls and MCI patients on each scanner).
Cheers
T
On 26 Oct 2006, at 21:21, Chunchun Ni wrote:
> Hi,
> I want to compare FA in specific area between MCI patients and control
>
> groups. But some are scanned on Philips, others are scanned on
> Siemens. I
>
> looked at mean FA of the whole brain. It's around 0.4~0.5 on Philips
> machine and around 0.2 on Siemens machine. That's huge difference. If
> there is big system bias, how can I combine data got from two
> machine? If
>
> the difference is caused by mistake during processing, how can I
> find it
>
> out? I double check every step, the processing is straightforward
> as long
>
> as I input the correct bvecs. Anybody have same problem and have
> hints?
>
> Thanks for your time.
>
>
> Chunchun
|