[Mristudio-users] Scanner-specific gradients?

[Igor Yakushev]

Dear Susumu,

Thanks again for your careful answer!
We have ascertained that the scan-specific (extracted from individual 
DICOM headers) gradients were the right ones. I have one more practical 
question:

Our individual gradients have values of 14-15 digits following zero, 
e.g. -0.145791634917259.
To what decimal digit should/may one round these values? It's important 
to know, as we're going to analyse large amount of data and, as i wrote,
we have to introduce a different gradient table for each individual 
case. Of note, there is a rather small difference between the values, 
please see an example of a gradient table in my previous mail and one 
more below (rounded to 3 decimal digits):

 0: 0.000, 0.000, 0.000
 1: 0.696, -0.157, -0.701
 2: -0.718, -0.157, -0.678
 3: -0.014, 0.533, -0.846
 4: 0.009, 0.846, 0.533
 5: 0.704, 0.690, -0.168
 6: -0.710, 0.690, -0.145

Thank you!
Igor

* susumu <susumu at mri.jhu.edu> [Thu, 26 Feb 2009 20:04:51 -0500]:
> Hi Igor,
>
> Unless you explicity change the gradient table, the same imaging
> protocol
> should use the same gradient table.
> However, there are several things that make the matter complicated;
>
> 1) The manufacturers provide gradient tables. For example, Siemens has
> 6-
> and 12-orientation tables. While they provide only one kind of the
> 12-orientation Siemens table, when actual scan happens, they
dynamically
> change the table.
> 2) For example, you may use axial, coronal, or sagital scans. The
> patient
> may go in the scanner head-first or leg-first. Or they could be face
up,
> face down, or facing to the side. There are so many possibilities
> related to
> the relative relationship between the subject, the physical gradients,
> and
> the imaging parameters. Apparently, one gradient table can not be
> applied to
> all these setttings.
> 3) One of the issues due to this type of degrees of freedom is that
the
> +/-
> sign of the gradient table may not represent what the scanner actually
> did.
> As a matter of fact, the +/- sign is arbitrary defined. What is most
> important is the relationship between the diffusion gradients and
> imaging
> gradients, which are changed dynamically depending on your protocol.
> Usually, for siemens, axial, head-first, face-up, the sign of the X
> gradient
> has to be flipped.
> 4) If you look the two tables you got carefully, you notice that the x
> components have opposite signs. I suspect the first one is the one
> provided
> by Siemens and the second one is read from DICOM header. The latter
one
> reflects what was actually done by the scanner, I believe.
> 5) If you use the one provided by the manufacturer (the first one) for
> tensor calculation, it won't affect your FA, ADC, and diffusivity, but
> it
> affects vector angles (and thus fiber tracking). This is why DtiStudio
> has a
> function of "flip" before fiber tracking. Of course, it is much better
> approach to use the second table with the correct +/- sign.
> 5) The second thing you notice is that there is a small differences in
> each
> component. I suspect this is because you used oblique imaging. Please
> check
> if this is actually the case. This is a very tricky issue and you have
> to
> make sure that you are doing a right thing. Not using oblique makes
> situation much simpler, but if you want to use oblique, you have to
> consider
> the following;
> 6) If you are using older operating system of Siemens, you HAVE TO
> recalculate your gradient table based on oblique angles. You can do it
> automatically if you are using DICOM or Mosaic. You have to check
> "rotate
> gradient if applicable" box. Then DtiStudio read the oblique
information
> and
> reorient the table for you.
> 7) If you are using the latest operating system, you don't have to
> recalculate the table. So you don't want to check the "rotate gradient
> if
> appliable" box. Either way, it is advised to image one person with a
> severe
> oblique plane, process the data, and perform tracking. This
immediately
> tells if things are correctly working. Vendors change things without
> notice.
> You'd better learn how to protect yourself.
> 8) If you are not sure, you can send the tracking results to us.
>
> Susumu
>
>
>
>
>
> -----Original Message-----
> From: mristudio-users-bounces at mristudio.org
> [mailto:mristudio-users-bounces at mristudio.org] On Behalf Of Igor
> Yakushev
> Sent: Thursday, February 26, 2009 12:13 PM
> To: DTI Studio, ROI Editor, Landmarker Questions/Support
> Subject: [Mristudio-users] Scanner-specific gradients?
>
> Dear Susumu, dear Users,
> I have a beginners question:Should be the combination of gradients the
> same
> for all subjects imaged with the same scanner (i.e. scanner-specific
> gradients)? Or the gradients may be different for different
> scans/subjects
> (i.e. scan-specific gradients)? Such "individual" gradients (bvec) can
> be
> calculated using e.g. MRIcron.
> For example, gradient table (as provided by radiologists) in the first
> case
> looks like:
> 0: 0, 0, 0 1: -0.7, 0, -0.7 2: 0.7, 0, -0.7 3: 0, 0.7, -0.7 4: 0, 0.7,
> 0.7
> 5: -0.7, 0.7, 0 6: 0.7, 0.7, 0 in the second case (calculated with
> MRIcron)
> completely different:
> 0: 0.000, 0.000, 0.000 1: 0.704, -0.146, -0.695 2: -0.710, -0.146,
> -0.689 3:
> 0.004, 0.546, -0.838 4: 0.003, 0.838, 0.546 5: 0.706, 0.692, -0.149 6:
> -0.708, 0.692, -0.142 Accordingly, results (tracks originated from the
> same
> ROI) are very different.
>
Thanks,Igor_______________________________________________Mristudio-users
> mailing
>
listMristudio-users at mristudio.orghttp://lists.mristudio.org/mailman/listinfo
> /mristudio-users
>
> _______________________________________________
> Mristudio-users mailing list
> Mristudio-users at mristudio.org
> http://lists.mristudio.org/mailman/listinfo/mristudio-users