DWI Weighted Masking about slicerdmri HOT 9 CLOSED

Thank you!

I think we have seen this before specifically with the tutorial dataset. Isaiah will know.

Lauren

On Jan 28, 2016, at 11:56 AM, spujol <[email protected]mailto:[email protected]> wrote:

from slicerdmri.

@spujol thanks, the issue is that the "B0" images in the tutorial dataset have a gradient magnitude exceeding the B0 tolerance. See debug output here, whereas the tolerance is 1e-6.

It seems that this tutorial dataset was updated recently , so I would like to investigate the original data to see if there was an issue in the initial conversion or possibly roundoff error in subsequent rewrites. Would it be possible to share the original DICOM images?

from slicerdmri.

The DWI Volume masking module with this tutorial dataset was working with Slicer4.4.0.2015-05.21.

Below is the link to Slicer4.4.0.2015-05.21 if this can help:
https://www.dropbox.com/s/u9lwqu0slweyn47/Slicer-4.4.0-2015-05-21-macosx-amd64.dmg.zip?dl=0
![dwi-mask-slicer4 4 0-2015-05-21

Let me know if you still need the dicom images.
(https://cloud.githubusercontent.com/assets/1847492/12658184/7d93b9bc-c5d5-11e5-8115-deeb1aad4e49.png)

from slicerdmri.

The gradient sum for tolerance check was previously not calculated correctly, leading to bugs for some datasets. This was fixed between 4.4 and 4.5 release, see:
Slicer/Slicer#380
http://www.na-mic.org/Bug/view.php?id=3855

Let me know if you still need the dicom images.

Yes, please. I would prefer to investigate from the source first before considering relaxing the tolerance (which has been in place for a long time, but was applied incorrectly).

from slicerdmri.

I agree looking at the DICOMs is an important check.

Regarding the tolerance, the threshold looks too small. For masking, any scan with b less than 10 for example would be reasonable. All we are really looking for is strong signal so even b in the low hundreds should work. Let's ask around lab about different protocols and see how often the baseline is a bit above zero.
Lauren

from slicerdmri.

@spujol sent the dataset, and the conversion looks fine, the low b-values are from 3-5 (directly from the BMatrix).

Sounds like we should check the actual b-value in the masking routine rather than just the gradient magnitude, and threshold on that. And also add an option to control the threshold. This was previously not an issue because the gradient components were all rounded to zero, so we considered all images to be baseline and averaged them.

from slicerdmri.

I ran the Diffusion Weighted Volume Masking in Slicer4.5.0-2016-02-23 nightly build with the Diffusion MRI tutorial dataset. I tried the default value (100) and max value (2147483647) of the Baseline B-Value Threshold Parameter.
In both cases, the mask is generated correctly, which is great.

The slider has a range of values up to the maximum values for int variables. Could the maximum value of the threshold parameter be set to the max B-value of the dataset?

from slicerdmri.

That's a good idea. Alternatively if that can't be done for whatever reason (CLIs can't modify UI based on input) it could be set to a max that is reasonable for b values, say whatever is the max of the connectome scanner

from slicerdmri.

Right, no way to make the value conditional, unfortunately. Updated in 0513bff to Connectome max b-value of 20k (per this). Will bump the Slicer tag again in a couple days.

from slicerdmri.