That is generally true, however, one have to take into the account that the noise in magnitude images is not Gaussian, but rather Rician. Averaging noisy, magnitude only data like a high-B value DTI image will not lead to effective noise suppression as one would expect with a Gaussian dataset. In fact, it could actually cause a built up of erroneous signal. To do the proper averaging, one should actually average both magnitude and phase images, this however, could be not practical with modern, multichannel scanner setups.
Another argument against averaging raw data is that patient motion, gradient errors and a built up of eddy currents could produce an erroneous offset in the directions and thus you won't be avering what expect to average. On the other hand, FA values are scalar and Gaussian, thus their avering avoids all the lengthy discussion above.
cheers
peter
________________________________
From: FSL - FMRIB's Software Library on behalf of Scott Kolbe
Sent: Sun 11/2/2008 5:54 PM
To: [log in to unmask]
Subject: Re: [FSL] DTI Average
you should always average your raw data.
noise exerts a greater effect on images with higher diffusion weighting
(less signal) so it is better to average first and then calculate the
tensor/bedpost parameters.
scott
Matt Glasser wrote:
> Why is that? Regardless, it is necessary to align your averages with some
> rigid body registration before averaging.
>
> Peace,
>
> Matt.
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
> Of Peter Kochunov
> Sent: Friday, October 31, 2008 10:42 AM
> To: [log in to unmask]
> Subject: Re: [FSL] DTI Average
>
> It would be safer to average the scalar output of the dtifit, rather than
> individual, diffusions-sensitized images.
> pk
> ----- Original Message -----
> From: "Departament de Psiquiatria i Psicobiologia Cli -nica UB Mundet"
> <[log in to unmask]>
> To: <[log in to unmask]>
> Sent: Friday, October 31, 2008 10:07 AM
> Subject: [FSL] DTI Average
>
>
> Hi, we would like to know how to average two or more DTI adquisitions in FSL
> software. Can you create an average image before preprocessing? Or you can
> only average the values you calculate from a ROI in a FA map (for example)
> of
> every adquisition from every subject after preprocessing?
>
> Thank you.
>
--
========================
Scott Kolbe
MS Imaging Group
Howard Florey Institute &
Centre for Neuroscience
University of Melbourne
VIC, Australia, 3010.
ph: +61 3 8344 1887
email: [log in to unmask]
website: http://www.neuroimaging.org.au/index.php?id=383
|