open-discussion > Visual check of the DTI data
Showing 1-8 of 8 posts
Feb 28, 2020 06:02 PM | ameliaqian
Visual check of the DTI data
I have ran the DTIPrep automated pipeline and going to have visual
check on QCed data. I noticed that the exclusion in DTIprep is
gradient-based. I wonder whether I need to check every slice in one
gradient to make sure there is no artifacts, or one slice in each
gradient is enough. Also, I wonder if the same artifact occurs in
multiple gradient (~10-20%), should I code this whole scan as
failed?
Mar 23, 2020 02:03 PM | ameliaqian
RE: Visual check of the DTI data
In the *_QCReport.txt file, there are "Slice-wise Check Artefacts".
Besides a column of excluded gradient, it includes a column of
"Slice#". I wonder what the slice mean? Does it imply the gradient
has high correlation in this specific slice, so this gradient is
excluded?
Also, I want to ask about the "Region" column. What do "whole", "region1", "region3" mean?
Also, I want to ask about the "Region" column. What do "whole", "region1", "region3" mean?
Mar 23, 2020 08:03 PM | Martin Styner
RE: Visual check of the DTI data
Originally posted by ameliaqian:
Martin
In the *_QCReport.txt file, there are
"Slice-wise Check Artefacts". Besides a column of excluded
gradient, it includes a column of "Slice#". I wonder what the slice
mean? Does it imply the gradient has high correlation in this
specific slice, so this gradient is excluded?
Also, I want to ask about the "Region" column. What do "whole", "region1", "region3" mean?
The Slice number indicates the slice
number where the slice wise check detected an artifacts (mainly
based on a low correlation of this slice to its neighboring slice).
You can disregard the Region column as it is unused.Also, I want to ask about the "Region" column. What do "whole", "region1", "region3" mean?
Martin
Mar 23, 2020 08:03 PM | Martin Styner
RE: Visual check of the DTI data
Originally posted by ameliaqian:
If you do a visual assessment, you need to do more than 1 slice per gradient. Best look at the orthogonal directions, i.e. if your data is axial, look at the sagittal and coronal slices, as you can see across-slice differences best in those views.
You can use the visual QC mode in DTIPrep (though it is quite clunky) to remove DWI gradients with bad slices.
With respect to failing a whole scan, we usually use a 30% threshold as compared to the full number of gradients (i.e. not based on the number of gradients in the QC output of DTIPrep, but the original number of gradients).
Martin
I have ran the DTIPrep automated pipeline and
going to have visual check on QCed data. I noticed that the
exclusion in DTIprep is gradient-based. I wonder whether I need to
check every slice in one gradient to make sure there is no
artifacts, or one slice in each gradient is enough. Also, I wonder
if the same artifact occurs in multiple gradient (~10-20%), should
I code this whole scan as failed?
If you do a visual assessment, you need to do more than 1 slice per gradient. Best look at the orthogonal directions, i.e. if your data is axial, look at the sagittal and coronal slices, as you can see across-slice differences best in those views.
You can use the visual QC mode in DTIPrep (though it is quite clunky) to remove DWI gradients with bad slices.
With respect to failing a whole scan, we usually use a 30% threshold as compared to the full number of gradients (i.e. not based on the number of gradients in the QC output of DTIPrep, but the original number of gradients).
Martin
Mar 23, 2020 10:03 PM | ameliaqian
RE: Visual check of the DTI data
The slice parameter is unclear for me. Does Slice# refer to the
slice in the coronal, sagittal or axial view?
Mar 23, 2020 10:03 PM | Martin Styner
RE: Visual check of the DTI data
Originally posted by ameliaqian:
Martin
The slice parameter is unclear for me. Does
Slice# refer to the slice in the coronal, sagittal or axial
view?
It 's the z-axis slice direction
(usually axial for diffusion sequences)Martin
Mar 24, 2020 12:03 AM | ameliaqian
RE: Visual check of the DTI data
Thank you for your fast response. Here are my follow-up questions:
1. In the DTIPrep manual, denoising is not default. One collaborator recommended me to denoise the DTI data. I wonder in which circumstance I should implement denoise procedure and how it influences the final analysis?
2. One participant are failed from DTIPrep pipeline with the reason "Single b value DWI without a b0/baseline". But when I checked this participant, it does have b0 and 64 gradients. What does this failing reason mean?
3. You mentioned "slice wise check detected an artifacts (mainly based on a low correlation of this slice to its neighboring slice)". I want to know whether DTIPrep calculates the correlation of this slice to its previous neighboring slice or more?
4. In the DTIPrep manual, the procedure first implement automated pipeline then visual check. In visual check, I found some DTIPrep-excluded gradient looks similar to the other unexcluded ones, while the other have the severe artifacts. I wonder whether there is a strong reason that I must exclude the "look-not-so-bad" gradients which has been detected? My rationale is to keep as much DTI gradients as possible at the beginning rather than losing them. Furthermore, could I perform visual check then automated pipeline?
5. I found I cannot implement visual check 2: "Directional Spatial Distribution QC". I can only see the lines starting from the centre of sphere, rather than dots. The lines change color when I move the sphere. My version of DTIPrep is 1.2.4.
5. I notices b0 in one brain scan have artifacts. Should I excluded this participant in this case?
1. In the DTIPrep manual, denoising is not default. One collaborator recommended me to denoise the DTI data. I wonder in which circumstance I should implement denoise procedure and how it influences the final analysis?
2. One participant are failed from DTIPrep pipeline with the reason "Single b value DWI without a b0/baseline". But when I checked this participant, it does have b0 and 64 gradients. What does this failing reason mean?
3. You mentioned "slice wise check detected an artifacts (mainly based on a low correlation of this slice to its neighboring slice)". I want to know whether DTIPrep calculates the correlation of this slice to its previous neighboring slice or more?
4. In the DTIPrep manual, the procedure first implement automated pipeline then visual check. In visual check, I found some DTIPrep-excluded gradient looks similar to the other unexcluded ones, while the other have the severe artifacts. I wonder whether there is a strong reason that I must exclude the "look-not-so-bad" gradients which has been detected? My rationale is to keep as much DTI gradients as possible at the beginning rather than losing them. Furthermore, could I perform visual check then automated pipeline?
5. I found I cannot implement visual check 2: "Directional Spatial Distribution QC". I can only see the lines starting from the centre of sphere, rather than dots. The lines change color when I move the sphere. My version of DTIPrep is 1.2.4.
5. I notices b0 in one brain scan have artifacts. Should I excluded this participant in this case?
Mar 29, 2020 10:03 PM | ameliaqian
RE: Visual check of the DTI data
1. In the DTIPrep manual, denoising is not default. One
collaborator recommended me to denoise the DTI data. I wonder in
which circumstance I should implement denoise procedure and how it
influences the final analysis?
2. One participant are failed from DTIPrep pipeline with the reason "Single b value DWI without a b0/baseline". But when I checked this participant, it does have b0 and 64 gradients. What does this failing reason mean?
3. You mentioned "slice wise check detected an artifacts (mainly based on a low correlation of this slice to its neighboring slice)". I want to know whether DTIPrep calculates the correlation of this slice to its previous neighboring slice or more?
4. I found I cannot implement visual check 2: "Directional Spatial Distribution QC". I can only see the lines starting from the centre of sphere, rather than dots. The lines change color when I move the sphere. So I cannot identify the location of excluded gradients. My version of DTIPrep is 1.2.4.
5. I notices b0 in one brain scan have artifacts in raw data but corrected in QCed.nrrd. Should I excluded this participant in this case?
6. I am trying to test the effect of NC on the DTIPrep output. Is it the parameter "SLICE_correlationDeviationThresholdGradient"? Also, I wonder if I change this parameter from 3.5 to 1, whether I need to change the parameter "SLICE_correlationDeviationThresholdbaseline" correspondingly?
2. One participant are failed from DTIPrep pipeline with the reason "Single b value DWI without a b0/baseline". But when I checked this participant, it does have b0 and 64 gradients. What does this failing reason mean?
3. You mentioned "slice wise check detected an artifacts (mainly based on a low correlation of this slice to its neighboring slice)". I want to know whether DTIPrep calculates the correlation of this slice to its previous neighboring slice or more?
4. I found I cannot implement visual check 2: "Directional Spatial Distribution QC". I can only see the lines starting from the centre of sphere, rather than dots. The lines change color when I move the sphere. So I cannot identify the location of excluded gradients. My version of DTIPrep is 1.2.4.
5. I notices b0 in one brain scan have artifacts in raw data but corrected in QCed.nrrd. Should I excluded this participant in this case?
6. I am trying to test the effect of NC on the DTIPrep output. Is it the parameter "SLICE_correlationDeviationThresholdGradient"? Also, I wonder if I change this parameter from 3.5 to 1, whether I need to change the parameter "SLICE_correlationDeviationThresholdbaseline" correspondingly?