Error in "Pipeline T1 2007"

[aurely1310]

Hello,

(you'll be used to see me on this forum! )

I have another question. i would like to do sulci recognition but have some problems. I made the same steps as in the tutorial

-import T1 MRI (OK)
- prepare subject to anatomical pipeline (OK)

- "Pipeline T1 2007" ==> BUT here...there is an exception message and something's strange...
When the window appears, the field left_hemisphere_point is empty! I have to define it again...Is it normal ?

Then, all the steps are correct until...
"Segmentation of brain mask" ==> The step is validated but with warning messages :

( Traitement Pipeline T1 2007 2 démarré sur 2010/05/10 16:48
OK, the image seems to be in the correct orientation. )

------------------------------
! VIP Warning:
Risky regularization, gray matter to close to 0, skip!
------------------------------
------------------------------
! VIP Warning:
Empty volume in VipGet3DConnex
------------------------------
------------------------------
! VIP Warning:
Empty volume in VipGet3DConnex
------------------------------
------------------------------
! VIP Warning:
Empty bucket list in VipGetNthBiggest3DBucket_S16BIT
------------------------------
------------------------------
! VIP Warning:
Nothing is left in VipConnexVolumeFilter
------------------------------
------------------------------
! VIP Warning:
Empty volume in VipGet3DConnex
------------------------------
------------------------------
! VIP Warning:
Empty volume in VipGet3DConnex
------------------------------
------------------------------
! VIP Warning:
Empty volume in VipGet3DConnex
------------------------------
------------------------------
! VIP Warning:
Empty volume in VipGet3DConnex
------------------------------
------------------------------
! VIP Warning:
Empty volume in VipGet3DConnex
------------------------------

Then, the failure is in the next step :

error1.jpg (22.78 KiB) Viewed 10456 times

error2.jpg (41.24 KiB) Viewed 10451 times

error3.jpg (56.61 KiB) Viewed 10450 times

What does it mean ?
Could you help me, please?

Thank you very much,
Aurélie

[Dominique Geffroy]

Hello,

The fact that the left_hemisphere_point field is empty in the T1 pipeline is not a problem. This information is only used the first time you run the Prepare subject step to check the orientation of the image, then it is not useful anymore, the needed coordinates are stored in a .APC file.

Sometimes, the pipeline can fail on some sets of data. The algorithms have a lot of parameters, the default values are generally ok but not always. The steps T1 Bias Correction and Histogram Analysis are the more likely to fail if the contrast in the image is low for example. First you can try to choose the alternative method of Bias correction step : unroll the item and choose "Vip bias correction" instead of "1 - T1 bias correction".
If it doesn't work better, try to do the same with "Histogram analysis" step : unroll the item and choose "2 - Histogram analysis" instead of "Vip Histogram analysis" .

Dominique

[aurely1310]

Hello,

i did what you said (and each possible configuration) but unfortunately the problem was here again...

During the operation called "segmentation du masque de cerveau", this message appears :

------------------------------
! VIP Warning:
Risky regularization, gray matter to close to 0, skip!
------------------------------
------------------------------
! VIP Warning:
Empty volume in VipGet3DConnex
------------------------------
------------------------------
! VIP Warning:
Empty volume in VipGet3DConnex
------------------------------

However, the step is accepted, but the execution of the program always stop at the next step : "separation du masque de cerveau". i tried several options but nothing changes...

Do you have another idea ?
Thank you very much,

Aurélie.

[Dominique Geffroy]

Hi,

You should look at the results of the different steps to try and see where the problem is. Does the bias correction look correct ? And the brain mask ?
In the bias correction step, you can try to change some parameters like the field_rigidity. You can find information about this step in the documentation of the process : http://brainvisa.info/doc/brainvisa/en/ ... ction.html

Dominique

[aurely1310]

I don't know how to see if the bias correction is correct.

But I think that the problem comes from the histogram : the next step seems to be curious...
I can not show the histogram (only the values, when I click on the pen). Is it normal?

And for the bias correction, what is the "coulour code" ?

Thank you,
Aurélie.

[Jean-Francois Mangin]

But I think that the problem comes from the histogram : the next step seems to be curious...
I can not show the histogram (only the values, when I click on the pen). Is it normal?

yes the pen is used for manual correction (you can guess the stats using the profile window of anatomist for instance)

And for the bias correction, what is the "coulour code" ?

I do think there is some documentation in front of the bias correction procedure?
Basically the trick is to tune the color map in order to have a mix of two colors in white matter.
If the correction is OK, the colors should be randomly located everywhere in white matter.

[Dominique Geffroy]

Hi Aurely,

Maybe this post by Denis could help you with the pipeline : http://brainvisa.info/forum/viewtopic.p ... 1446#p5268

Dominique

[aurely1310]

Hello, thank you for your answer.

It seems that it was also a problem with Windows...I'm now working on linux, and all seems to be good...
i just changed the parameter "histogram".

Now, you know what you can ameliorate

Thank you,

Aurelie.

[gzr]

I know you said you resolved the issue by using a Linux system, but I think I may have just encountered the same problem and figured out a solution so I thought I'd share in case it's helpful (FYI, I'm working on a Mac):

I've had a few of my files hit a fatal error in the "Split Brain Mask" step of the Pipeline. Here's an example of the resulting error message:
"in 4 - Split Brain Mask 8 System command exited with non null value : 111
RuntimeError
neuroProcesses.py (1834) in _processExecution:
result = process.execution( self )
SplitBrain.py (112) in execution:
apply( context.system, call_list+option_list )
neuroProcessesGUI.py (1323) in system:
ret = apply( ExecutionContextGUI.system, (self,) + args, kwargs )
neuroProcesses.py (1928) in system:
raise RuntimeError( _t_( 'System command exited with non null value : %s' ) % str( ret ) )"

Although I am not fluent in French, it seems like one of your error messages may be the same thing ("Separation du masque due cerveau"). Based on the documentation that accompanies that step in the "Components" drop-down, I ran the "Correction Split Brain from Brain Mask" and verified (visually) that the split was relatively accurate. I then reran the Pipeline, de-selecting everything up to and including the "Split Brain Mask" step. (Incidentally, anyone know if there's a trick to de-selecting/unchecking things in the Pipeline? It seems finicky and sometimes difficult to do!)

So far I've tried this for 2 files -- for one of the files, all seems fine/resolved; for the other, the "Correction" is still failing and giving error messages. That particular image seems to be of poorer quality than the rest, so I'm guessing it's an issue of artifact/quality.

[Jean-Francois Mangin]

Basically, there is two key tricks to be known for overcoming usual failures of the current pipeline:
1) When the histogram analysis fails, whatever how, try one by on the values of the menu
dedicing for the histogram sampling (2, 4, 8...). Most of the time, one will work.
2) When the split procedure fails, try one by one other values for the barycentre parameters
deciding for the threshold defining white matter. You should find a value leading you
to something you can use further (manual correction for instance:-)

Fortunately, we have finally decided to work on the robustness of the pipeline. I hope that
next year, this stuff will work most of the time without tuning...

Additional efficient trick used by some of you:
skip the first part of the pipeline using another software:
For instance, putting a skull stripped image as input seems to be
a good approach. You can also provide the brain mask at the right place
and use a grey/white segmentation to construct the histogram analysis file.
The command VipGreyStatFromClassif can be used for this purpose.