[caret-users] [Fwd: Re: Problems with CARET]
Donna Dierker
donna at brainvis.wustl.edu
Tue Apr 7 15:37:47 CDT 2009
Markus,
I want to retract two statements:
* "There is definitely something wrong with the NIfTI header in
SWI_Images.hdr": Rather than "is definitely", I think "may be" is more
apropos.
* "mayo_analyze read the header and interpret the pixdim as cubic 1.0":
I misread the output. AFNI's mayo_analyze says the pixdims are cubic 0.5.
Both Caret and AFNI still show the segmentation volumes misaligned with
the anatomical (in the original_files directory -- not in the
corrected_files). This remains true.
These are very large volumes (364x448x144x4), but my machine loaded all
five segmentations plus the anatomical underlay concurrently, with no
out of memory problems. I also was able to reconstruct a VTK model from
the segmentations without out of memory errors; however, only the
biggest object is reconstructed. I don't know of a way to get Caret to
reconstruct a VTK model from multiple objects.
I dug out these hints, which David passed on last year when doing
something similar with subcortical structures:
> Generate surface from segmentation, save coord, topo
> Compute folding
> Surface ROI: threshold nodes with folding > 1
> Dilate selected nodes twice
> Smooth selected nodes 100 iters
> This gets rid of spikes while preserving main shape features
> Save as vtk model
Finally, here is a related hint from John Harwell:
> To map functional data to the VTK models, "caret_command
> -volume-map-to-vtk-model" has been added. It takes as input a VTK
> model file, a functional volume file, and the name of one of the
> default color palettes. The output is a VTK model file with color
> assignments made using the functional volume.
Donna
On 04/07/2009 11:52 AM, Donna Dierker wrote:
> Hi Markus,
>
> (Besides the list, I'm cc'ing you directly, to see if this reduces the
> time lag. The list server delivers within minutes on my end.)
>
> There is definitely something wrong with the NIfTI header in
> SWI_Images.hdr. At first, I thought perhaps the problem was that
> Caret doesn't read NIfTI .hdr/.img pairs (but I thought John added
> that recently), or that it is an oblique dataset, as AFNI warns:
>
> *+ WARNING: If you are performing spatial transformations on an
> oblique dset,
> such as
> /tmp/university_duisburg_essen/original_files/439-alle_kerne.nii,
> or viewing/combining it with volumes of differing obliquity,
> you should consider running:
> 3dWarp -deoblique
> on this and other oblique datasets in the same session.
>
> But the fact that both Caret and mayo_analyze read the header and
> interpret the pixdim as cubic 1.0 is a clue (see attached mayo_analyze
> output, and compare with the attached nifti_tool output). In AFNI,
> when I view the corrected 439-alle_kerneneu.nii overlaid on the
> c_SWI_Imagestest.nii, they align fine, but when I view
> 439-alle_kerne.nii overlaid on SWI_Images.hdr, they do not (captures
> attached).
>
> It looks like SUIT_Isolate is doing more than just converting
> SWI_Images.hdr to NIfTI, so you might want to do something less
> drastic, but you somehow need to fix the bad .hdr.
>
> Donna
>
> On 04/07/2009 05:26 AM, Markus Thuerling wrote:
>> Hello Donna,
>> thanks for your great help. I used the link to upload the compressed
>> file university_duisburg_essen.zip. This contains two folders. The
>> first, named as original files, contains the old SWI_Images (img/hdr
>> format in which the conversion was not possible - either the voxel
>> size growth from 0.5x0.5x0.5mm^3 to 1x1x1mm^3 or in CARET the offset
>> occurs) and the old nucleii from the cerebellum in the nii-format.
>> The second (named as my_corrected_files) was formed with the software
>> SUIT_Isolate (this was developed from Joern Diedrichsen to isolate
>> the cerebellum) to get smaller files (reduced the size from 45
>> megabyte to 25 megabyte). Here it was possible to convert the
>> img/hdr- to nii-format without problems (i don't no why). I must open
>> the nucleii-files in MRICRON under draw\open voi and saved it again
>> to avoid offsets in CARET. So problem 1 could be solved, but i cannot
>> explain it. With the steps in CARET that i told you in the first
>> email to get the surface from the nucleii i got the same problem,
>> that the surface is only gray at one part of the surface of the nucleii.
>> The OUT OF MEMORY-problem occurs sometimes in SPM. It is usual to use
>> such great files in CARET?
>>
>> PS: It could be that i see your emails one day later because the time
>> lag between Germany and USA. ;)
>>
>> Best regards Markus
>>
>>
>>
>>> -----Ursprüngliche Nachricht-----
>>> Von: "Donna Dierker" <donna at brainvis.wustl.edu>
>>> Gesendet: 06.04.09 21:07:10
>>> An: caret-users <caret-users at brainvis.wustl.edu>
>>> Betreff: [caret-users] [Fwd: Re: Problems with CARET]
>>>
>>
>>
>>
>>> This thread got started off-list, before I saw Markus' pending
>>> caret-users post and approved it.
>>>
>>> By forwarding this message to caret-users, I'm putting it on-list,
>>> because I think the problem will get solved quicker that way.
>>>
>>> There are two separate problems:
>>>
>>> 1. The segmentation and anatomy volumes are not in register; there
>>> appears to be an offset/translation, even after both volumes are
>>> converted to NIfTI and reopened/loaded in Caret.
>>>
>>> 2. Caret shuts down (out of memory) when reconstructing the
>>> segmentation into a VTK model (possibly because there is more than
>>> one object in the segmentation?).
>>>
>>> For problem #1, all I can do is have you upload both volumes and
>>> reproduce the problem here:
>>>
>>> http://pulvinar.wustl.edu/cgi-bin/upload.cgi
>>>
>>> I can also use nifti_tool to display the NIfTI header, which might
>>> yield some clues (e.g., there is a transform specified in one, but
>>> not the other).
>>>
>>> For problem #2, I'm curious what happens if you try Volume:
>>> Segmentation: Disconnect Islands before reconstructing the
>>> segmentation into a VTK model.
>>>
>>>
>>> -------- Original Message --------
>>> Subject: Re: Problems with CARET
>>> Date: Mon, 06 Apr 2009 18:23:15 +0200
>>> From: Markus Thuerling <mthuerling at web.de>
>>> Organization: http://freemail.web.de/
>>> To: Donna Dierker <donna at brainvis.wustl.edu>
>>>
>>>
>>>
>>> Hello Donna,
>>> thank you very much for the fast answer. I saved the SWI-datas as
>>> nii-file. The datas are displayed isotropic and i don't must change
>>> the voxel size. But when i do this (1x1x1mm^3 is not coorect, but
>>> negligible), the shift also occurs. My Volume Segmentations Files
>>> have a size of 23 megabyte. But the Out of Memory-Problem not always
>>> occur. It was a great problem as i used the Reconstruct into surface
>>> - step with more nucleii resp. with more files which contain one
>>> nucleus. It is possible to save the nucleii in one file, but then i
>>> don't have an idea to color different nucleii (provided that if it
>>> is possible).
>>> You can see i am not an expert in CARET ;)
>>> Best regards Markus
>>>
>>>
>>>> -----Ursprüngliche Nachricht-----
>>>> Von: "Donna Dierker" <donna at brainvis.wustl.edu>
>>>> Gesendet: 06.04.09 17:45:30
>>>> An: Markus Thuerling <mthuerling at web.de>
>>>> CC: John Harwell <john at brainvis.wustl.edu>
>>>> Betreff: Re: Problems with CARET
>>>>
>>>
>>>> Hi Markus,
>>>>
>>>> Is it possible to save SWI.hdr as NIfTI, rather than Analyze? The
>>>> translation offset is likely due to the fact that Analyze does not
>>>> provide adequate/reliable origin information. It also doesn't
>>>> prescribe orientation information reliably, so you could also have
>>>> a flip issue and not realize it. Usually Analyze *does* provide
>>>> accurate voxel dimensions, however, so I'm surprised you had to
>>>> correct the voxdims.
>>>>
>>>> If the .nii and .hdr files have the same dimensions, voxel
>>>> dimensions, and origins, then it should be possible to copy the
>>>> header info from the .nii file to the .hdr. But if they don't,
>>>> then I'm not sure how else you can get the correct origin info,
>>>> other than by going back to the original DICOM.
>>>>
>>>> I'm cc'ing John Harwell about Caret shutting down (out of memory)
>>>> after reconstructing the .nii segmentation into a solid structure
>>>> VTK model.
>>>>
>>>> If by color different nucleii, you mean can Caret parcellate the
>>>> cerebellum, the answer is no. Nor can it parcellate the cortex.
>>>> (But in the latter case there are "atlas goodies" that help one
>>>> probabilistically determine where any given node is located.) But
>>>> if you mean is it possible to render a node-ROI or node-color
>>>> mapping that already exists, then the answer is yes. (I mean find
>>>> the label: no; render the label: yes.)
>>>>
>>>> If your NIfTI dentate.nii file has both left and right dentate
>>>> structures included in the segmentation, then I wonder if this
>>>> might be causing the out of memory error above (i.e., Caret is used
>>>> to dealing with single object segmentations). If Caret does handle
>>>> multi-object segmentations, then I'm not sure whether it wants to
>>>> save each object in its own file, or can support storing multiple
>>>> objects in a single VTK file. This is another good question for John.
>>>>
>>>> Donna
>>>>
>>>> On 04/06/2009 10:26 AM, Markus Thuerling wrote:
>>>>
>>>>> Dear Sir or Madam,
>>>>> i am PhD student at the University Duisburg-Essen, Department of
>>>>> Neurology.
>>>>> I want to show the surface of the nucleus dentatus in the
>>>>> cerebellum with CARET v5.61. My problems are shifts between the
>>>>> anatomical volume (the cerebellum was measured with a SWI
>>>>> sequence, saved as hdr / img-file) and the segmentation volume
>>>>> (the dentatus, saved in a nii-file), the right part of the
>>>>> dentatus (and the same problems with other nucleii) is only saved
>>>>> and i want to color different nucleii.
>>>>> I used CARET with the following steps:
>>>>> File\Set Current Directory
>>>>> File\Open Data File\SWI.hdr as Volume Anatomy File
>>>>> Create a spec file with structure cerebellum or all and unknown
>>>>> space (are these inormations important?)
>>>>> Without radiological orientation
>>>>> File\Open Data File\dentate.nii file as Volume Segmentation File
>>>>> Change under D/C the Primary Overlay to Segmentation
>>>>> Volume\Segmentation\Reconstruct into Surface\Solid Structure VTK
>>>>> Model (or Brain Model Surface)
>>>>> to get the Surface of the nucleus. This step is problematic,
>>>>> because CARET often shut down (Out of Memory) or don't realise
>>>>> this step. I must often repeat open the Volume Anatomy File to get
>>>>> the Model SURFACE&VOLUME.
>>>>> The SWI-image file is not correct displayed so i changed the voxel
>>>>> size to 0,5x0,5x0,5mm^3 (the correct voxel size). After this step
>>>>> i have a shift between Anatomy and Segmentation File.
>>>>> It is possible to color different nucleii?
>>>>> In the appendix you can find two screenshots. You can see a gray
>>>>> and a red nucleus. When i want to save the files, or use the
>>>>> procedure to the next nucleus, CARET saves ONLY the right nucleus
>>>>> and i can not explain why.
>>>>> Thank you in advance Markus Thürling
>>>>>
>>>>>
>>>>>
>>>>> ____________________________________________________________________
>>>>> Psssst! Schon vom neuen WEB.DE MultiMessenger gehört? Der kann`s
>>>>> mit allen: http://www.produkte.web.de/messenger/?did=3123
>>>>>
>>>>>
>>>>> ------------------------------------------------------------------------
>>>>>
>>>>>
>>>>>
>>>>> ------------------------------------------------------------------------
>>>>>
>>>>>
>>>>>
>>>>
>>> ________________________________________________________________________
>>>
>>> Neu bei WEB.DE: Kostenlose maxdome Movie-FLAT!
>>> https://register.maxdome.de/xml/order/LpWebDe?ac=OM.MD.MD008K15726T7073a
>>>
>>>
>>>
>>> _______________________________________________
>>> caret-users mailing list
>>> caret-users at brainvis.wustl.edu
>>> http://brainvis.wustl.edu/mailman/listinfo/caret-users
>>>
>>>
>>
>>
>> ____________________________________________________________________
>> Psssst! Schon vom neuen WEB.DE MultiMessenger gehört? Der kann`s mit
>> allen: http://www.produkte.web.de/messenger/?did=3123
>>
>>
>> _______________________________________________
>> caret-users mailing list
>> caret-users at brainvis.wustl.edu
>> http://brainvis.wustl.edu/mailman/listinfo/caret-users
>>
> ------------------------------------------------------------------------
>
> ------------------------------------------------------------------------
>
More information about the caret-users
mailing list