Back to the main page.
Bug 1816 - extend ft_volumesegment with FSL BET and FAST
| Status |
ASSIGNED |
| Reported |
2012-11-07 15:47:00 +0100 |
| Modified |
2013-01-28 16:00:25 +0100 |
| Product: |
FieldTrip |
| Component: |
forward |
| Version: |
unspecified |
| Hardware: |
PC |
| Operating System: |
Mac OS |
| Importance: |
P3 enhancement |
| Assigned to: |
Ümit Aydin |
| URL: |
|
| Tags: |
|
| Depends on: |
|
| Blocks: |
1822 |
| See also: |
|
Robert Oostenveld - 2012-11-07 15:47:41 +0100
these would be alternatives to the existing brain/skull/skin and csf/grey/white segmentations.
Lilla Magyari - 2012-11-20 15:06:17 +0100
Dear Umit,
thanks for sending the dti and T1, T2 images yesterday. Now, I also assign this bug for you. This is about the FSL segmentation. I made a test script again, you can find it in test/test_bug1816. It gives an overview how the new segmentation method could fit into the FieldTrip pipeline.
Lilla
Ümit Aydin - 2012-11-29 08:40:02 +0100
Dear Lilla,
thanks for the test_bug1816, I will try to come up with the code as soon as possible but it might take until January because currently I am involved in some paper projects which are very urgent.
Best,
Ümit Aydin - 2013-01-28 16:00:25 +0100
Hi
I created a function that should work with test1816. I also made some changes in the test file and send those to your e-mail.
Currently you can obtain the labelled MRI as the output and it also saves separate binary masks for skin, innerskull, outerskull, white matter, gray matter and CSF in nifti format.
You can use only T1 or both T1 and T2 for segmentation.
You can select to have just skin, inner skull and outer skull or in addition to those wm,gm,CSF.
There are also a bunch of fsl options that can be used and you can find them in the help of the function I wrote. To activate them you can set them as 1 like cfg.fsl.bet.B = 1.
For the wm, gm and CSF segmentation it stops after segmentation, shows the segmentation and asks for the user to define which layer is wm and which one is gm. There you can enter 3 values.
1 --> darkest compartment
3 --> lightest compartment
2 --> in between 1 and 3.
I included this user interface because the values it assigns to compartments changes every time. But once you answer this then the labels are handled consistent within the code.
It also saves the partial volume images for wm and gm.
According to your tests we can think about which of those outputs should not be saved and calculated. As an output of the function I gave out the labelled MRI that includes all compartment that is required by the user and the corresponding labels. One can also access separate binary images of the compartments by loading their respective nifti images.
Please note that for our dataset using both T1 and T2 results in worse results than just T1 for outerskull compartment. The reason for that is our T2 sequence is programed to emphasize skull-spongiosa and the bet algorithm mistakens that white skull spongiosa layer with white scalp layer. In our procedure I use the outerskull of T1. And the innerskull from just T2. But for normal T1 and T2 sequences using both should result in better outcome.
One further point is the coordinate systems. It uses the nifti coordinate system.
Sorry for the delay and best regards,
Ümit