Hi Luca,<div><br></div><div>There are several issues you raised;</div><div><br></div><div>1) Our MI algorithm: I forward this question to Kegang Hua who implemented an algorithm written by Jinsuh Kim in Iowa U.</div><div>2) How the gradient table updated: There is an option to re-calculate b-matrix if there is any rotation involved in the post-processing registration (a check box with "adjust gradient table ...."), although my lab is not using this function. It is my personal opinion that if a subject rotates his/her head to a degree that this option makes a difference, there should be so many other quality issues that can't be solved by post-processing reorientation of b-matrix such as signal drop, mis-registration of odd and even number slices within one 3D data, through-slice rotation within one 3D data, etc. Also consider following things;</div>
<div>> In regular data, the amount of rotations is usually less than 1 degree within 5-10 min scan time</div><div>> With 5-10 min scanning, SNR is such that test-retest reproducibility of fiber orientation calculation is about 3-5 degree</div>
<div>> I'm not sure if post-processing realignment algorithm has a precision of 1 degree for rotation detection, which is difficult to differentiate from sheering.</div><div>> Eddy current sheering should not require b-matrix rotation</div>
<div>> If there are 3-5 degree rotation during scan, that data is very outlier and I'm not sure if you want to salvage it by applying b-matrix reorientation.</div><div>3) What does it happen with resampling: Any post-processing alignment leads to pixel-interpolation and resultant blurring. This is a separate issue from any resampling to enhance the resolution. The post-processing realignment can be done without changing image resolution.</div>
<div><br></div><div>Susumu</div><div><br><div class="gmail_quote">On Mon, Dec 27, 2010 at 3:38 AM, <span dir="ltr"><<a href="mailto:luca.binotto@unipd.it">luca.binotto@unipd.it</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
Hello Susumu,<br>
<br>
I would like to know something about the way DtiStudio performs the update<br>
of the gradient table after the realignment of the dti dataset with<br>
respect to the mean_b0 volume.<br>
<br>
The registration algorithm implemented in DtiStudio using mutual<br>
information (MI):<br>
1) does it uses normalized mutual information or which other formulation ?<br>
2) how is the gradient table updated ? (rough simple theory)<br>
After realignment of the DWI volumes, b-matrix (the gradient table) must<br>
be updated.<br>
<br>
Patient motion and eddy currents can cause errors in the calculation of<br>
parameters that describe diffusion in each voxel of a DWI dataset.<br>
In normal conditions eddy currents contribution should be<br>
very small (< 1% ?).<br>
But the b-matrix (gradient table) is calculated with respect to a fixed<br>
coordinate system (at the acquisition time) so the b-matrix must be<br>
rotated accordingly to the transformation applied for image registration<br>
otherwise we should get poor fiber statistics and some fibers possibly<br>
deviate.<br>
<br>
So the gradient table is affected by rotations. Translations don't<br>
change it.<br>
But what does it happen with resampling (applying a scaling factor: es.<br>
affine transformation) to a dwi dataset?<br>
According to you is normally its effect small (as for eddy currents<br>
effect) or can be more important?<br>
For example a scaling factor such that a volume with a voxel size of<br>
2x2x2 mm3 is resliced to a 1x1x1 mm3 voxel size can be important its<br>
effect or "only" a lower image quality (eg. I read somewhere of a<br>
"red background" superimposed to the correct color map)?<br>
<br>
<br>
thanks in advance,<br>
best regards,<br>
<br>
Luca<br>
<br>
<br>
<br>
<br>
_______________________________________________<br>
mristudio-users mailing list<br>
<a href="mailto:mristudio-users@mristudio.org">mristudio-users@mristudio.org</a><br>
<a href="http://lists.mristudio.org/mailman/listinfo/" target="_blank">http://lists.mristudio.org/mailman/listinfo/</a><br>
Unsubscribe, send a blank email to: <a href="mailto:mristudio-users-unsubscribe@mristudio.org">mristudio-users-unsubscribe@mristudio.org</a><br>
</blockquote></div><br></div>