A Question Regarding Tracking Threshold

[maede...@gmail.com]

Dear Frank

There are 2 issues regarding tracking threshold I have problem with.

First, for some data, I have problem selecting the proper tracking threshold. Specially when there are lots of false positive in CSF, removing all false positives may cause removing the fibers in some parts like cerebellum. I will send one of my fib files to which I have set the threshold 0.08150. I would be grateful if you could check and let me know if it is a good choice or not. It will help me a lot for the other subjects.

The second one is regarding the group analysis, because I have a big data set including around 700 subjects, I need to automatically do the analysis and this threshold seems a little subjective. The best threshold varies subjects to subjects. I wonder how to manage this problem in the process of automation. I would be grateful if you could help me.

Best regards

Maedeh 

[Frank Yeh]

The data has only 30 directions in each of the two shells, and the signal quality will drop a lot if there is a movement.

The data you sent me seems to have artifacts due to either eddy current or subject movements. Increasing the anisotropy threshold may not help much in this case.

Have you run the SRC quality control? http://dsi-studio.labsolver.org/Manual/Parse-DICOM#TOC-Batch-Quality-Control-for-SRC-files 

Some of the data may have quality issues that need to be removed.

Frank

--
You received this message because you are subscribed to the Google Groups "DSI Studio" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dsi-studio+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/aec7265c-1507-4af7-8323-2bca044e7003n%40googlegroups.com.

[maede...@gmail.com]

Dear Frank

I see what you mean.Once I had done eddy correction using fsl but I didnt much help and reviewing the manual of dataset I got some slight preprocessing steps have been done so ....but you are completely right. I did a src quality control on my dataset. It seems the neighboring DWI correlations are low although I don exactly know its normal threshold. And some bad slices are discovered in some data. The report.txt has been drop boxed to you. I would be grateful If you could give me a hint what I can do to solve the problem.

Thanks a million 

[maede...@gmail.com]

Dear Frank

I got how to solve the problem of bad slices, I just drop boxed you 2 files (one of them is the fib file of a subject in which DSI discovers 2 bad slices in its raw data, the second one is the fib file of that subject after eddy correction in which dsi doesnt show any bad slices )

To me there are no differences...!!!

But I would be grateful if you check and let me know if applying (top up,eddy) has been useful ,so I do the process for the other problematic subjects.

by the way the tracking threshold I set for the eddy corrected fib is 0.02756 . I would be more thankful if you let me know if the threshold is ok.

Thanks more

Maedeh

On Monday, April 12, 2021 at 5:33:21 PM UTC+2 Frank Yeh wrote:

[Frank Yeh]

Here's my suggestion on dropping outliers: http://dsi-studio.labsolver.org/Manual/Parse-DICOM#TOC-Batch-Quality-Control-for-SRC-files

On Tue, Apr 13, 2021 at 3:50 AM maede...@gmail.com <maede...@gmail.com> wrote:

Dear Frank

I see what you mean.Once I had done eddy correction using fsl but I didnt much help and reviewing the manual of dataset I got some slight preprocessing steps have been done so ....but you are completely right. I did a src quality control on my dataset. It seems the neighboring DWI correlations are low although I don exactly know its normal threshold. And some bad slices are discovered in some data. The report.txt has been drop boxed to you. I would be grateful If you could give me a hint what I can do to solve the problem.

Thanks a million 

On Monday, April 12, 2021 at 5:33:21 PM UTC+2 Frank Yeh wrote:

The data has only 30 directions in each of the two shells, and the signal quality will drop a lot if there is a movement.

The data you sent me seems to have artifacts due to either eddy current or subject movements. Increasing the anisotropy threshold may not help much in this case.

Have you run the SRC quality control? http://dsi-studio.labsolver.org/Manual/Parse-DICOM#TOC-Batch-Quality-Control-for-SRC-files 

Some of the data may have quality issues that need to be removed.

Frank

On Mon, Apr 12, 2021 at 7:03 AM maede...@gmail.com <maede...@gmail.com> wrote:

Dear Frank

There are 2 issues regarding tracking threshold I have problem with.

First, for some data, I have problem selecting the proper tracking threshold. Specially when there are lots of false positive in CSF, removing all false positives may cause removing the fibers in some parts like cerebellum. I will send one of my fib files to which I have set the threshold 0.08150. I would be grateful if you could check and let me know if it is a good choice or not. It will help me a lot for the other subjects.

The second one is regarding the group analysis, because I have a big data set including around 700 subjects, I need to automatically do the analysis and this threshold seems a little subjective. The best threshold varies subjects to subjects. I wonder how to manage this problem in the process of automation. I would be grateful if you could help me.

Best regards

Maedeh 

--
You received this message because you are subscribed to the Google Groups "DSI Studio" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dsi-studio+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/aec7265c-1507-4af7-8323-2bca044e7003n%40googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "DSI Studio" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dsi-studio+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/25ea159c-7d9e-4b15-aece-ebfc20fb7ae5n%40googlegroups.com.

[maedeh khalilian]

Dear Frank

Thank you very much. I got how to solve the problem of bad slices, I  drop boxed you 2 files (one of them is the fib file of a subject in which DSI discovers 2 bad slices in its raw data, the second one is the fib file of that subject after eddy correction in which dsi doesnt show any bad slices )

To me there are no differences...!!!

But I would be grateful if you check and let me know if applying (top up,eddy) has been useful ,so I do the process for the other problematic subjects.

by the way the tracking threshold I set for the eddy corrected fib is 0.02756 . I would be more thankful if you let me know if the threshold is ok.

Thanks more

Maedeh

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/CAG_QrtJVyQKdLU5%3DanP_iu_OOrcJBq-nneE0OyEZBVRHK8mV3g%40mail.gmail.com.

[Frank Yeh]

Thank you very much. I got how to solve the problem of bad slices, I  drop boxed you 2 files (one of them is the fib file of a subject in which DSI discovers 2 bad slices in its raw data, the second one is the fib file of that subject after eddy correction in which dsi doesnt show any bad slices )

To me there are no differences...!!!

GQI reconstruction is robust against bad slice, and thus I usually don't do this correction.

 

But I would be grateful if you check and let me know if applying (top up,eddy) has been useful ,so I do the process for the other problematic subjects.

I would suggest still doing both of them just for publication purposes. You don't want a reviewer give you a hard time just because you skip these.

 

by the way the tracking threshold I set for the eddy corrected fib is 0.02756 . I would be more thankful if you let me know if the threshold is ok.

I would set it to 0, and DSI Studio will use the default otsu threshold.

Frank

 

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/CAHd3-cxGFhE%2B%2BGmD9_NcraiHdP7DxNjv0OOd042Gkv%2B2xETJMw%40mail.gmail.com.