Hi Mike,

Thank you for your email. This looks great. Is there a way to compute the corresponding adc map (to help sort out the shine through)? I have a bunch of acquisitions where the isotropic images are distorted and am hoping to generate isotropic images from dti acquisitions that can benefit from eddy_correct while showing the same acute infarcts as the isotropic images, but at higher resolution / better spatial fidelity.

Thanks!

-Salil

Date:    Mon, 16 Feb 2015 11:32:41 -0500
From:    Michael Dwyer <[log in to unmask]>
Subject: Re: Possible to output the trace image from FSL processing for DTI

Hi Salil,

The trace image is not used much quantitatively because it confounds T2
effects with diffusion effects ("T2 shine-through"). That being said, it
can still be very useful for picking out early ischemia with restricted
diffusion.

As far as I know, it is not a direct output from dtifit or other FSL tools.
However, you can compute it yourself pretty easily. The trace image is just
the geometric mean (not arithmetic mean) of the individual DWI-weighted
direction images (without the b0 image included).

trace = (bimg_1 * bimg_2 * bimg_3 * ... * bimg_n) ^ (1/n)

Because there are a lot of images, you will likely run into floating point
numerical issues if you try to calculate it like that, though. However, you
can also calculate it as the exponential of the average of the logarithms
of the individual images. You can do this using fslmaths as follows:

First, put together all your non-b0 images. Assuming the b0s are at the
beginning, you can do:

> fslroi dwi_series dwi_nonb0 5 -1

If they aren't at the beginning, you'll have to take them out more manually
and splice things back together.

Then take the logarithm:

> fslmaths dwi_nonb0 -log dwi_nonb0_log

Then average:

> fslmaths dwi_non_b0_log -Tmean dwi_non_b0_log_mean

Then exponentiate:

> fslmaths dwi_non_b0_log_mean -exp trace


You can do the last three steps in one line if you like:

fslmaths dwi_nonb0 -log -Tmean -exp trace


Hope this helps!

Best,
Mike





On Sun, Feb 15, 2015 at 6:49 PM, Matteo Bastiani <
[log in to unmask]> wrote:

> Dear Salil,
>
> when you process your diffusion weighted dataset using the dtifit routine
> implemented in FSL, you will get, amongst the other output files, one that
> ends with “_MD.nii.gz”.
> That is the mean diffusivity volume, which is what you are probably
> looking for.
>
>
> Cheers,
> Matteo
>
>
> > On 15 Feb 2015, at 19:43, Salil Soman <[log in to unmask]> wrote:
> >
> > Dear List Members,
> >
> > I was wondering if there is a way to output the trace image from a multi
> direction DTI acquisitoin (e.g. 30 directions with single b value of 1000
> with 5 b0 images)? To clarify, the trace image to which I am referring is
> the diffusion weighted image that is routinely used in clinical practice,
> generated automatically by the MRI scanner, for evaluating the presence of
> strokes.
> >
> > Thank you for your time and consideration.
> >
> > Best wishes,
> >
> > Salil Soman, MD, MS
>
>


--
Michael G. Dwyer, Ph.D.
Assistant Professor of Neurology and Biomedical Informatics
Director of Technical Imaging Development
Buffalo Neuroimaging Analysis Center
University at Buffalo
100 High St. Buffalo NY 14203
[log in to unmask]
(716) 859-7065