[MINC-users] Best plane (axial or coronal) for CIVET processing?

Tue Mar 20 09:34:07 EDT 2012

Hello Kie Woo,

>From what you are describing, I think that the issue you came across
is not an issue with CIVET but with the format conversion you used,
and/or the way the Analyze data were resliced. I would guess that in
the reslicing the world coordinate information (and probably other
metadata) was not maintained; so when you then convert these two
resulting images to MINC independently the images would - from the
MINC/CIVET perspective - be different and may thus yield different
processing results.

If the images would actually be the same, you would be able to load
them both in "register" and they would align; if they don't do that
(as you say), you have at minimum lost the world coordinate
information in the conversion or reslicing process.

To properly test the impact of dimension ordering, you should convert
the image to MINC once, and then use the MINC tools to perform the
reordering, for example like so:

   mincreshape -transverse <coronal.mnc> <transverse.mnc>

which will properly maintain the image metadata. In this case the
results you get from processing <coronal.mnc> and <transverse.mnc>
using CIVET should be the same; these two images should also line up
in "register" and have the exact same voxel values at the same spatial
coordinates.

The important thing to remember is that MINC stores much more useful
information than most other file formats, so if those other formats or
tools do not support those features, doing multiple conversions will
almost certainly get you into trouble.

-- Alex

On Mon, Mar 19, 2012 at 6:52 PM, Nam, Kie Woo <kie_woo.nam at kcl.ac.uk> wrote:
> Dear Alex,
>
> I much appreciate your quick and detailed reply.
>
> I find your advice very useful, but my image has been acquired only in the
> coronal plane in UNC format. I've converted it into Analyze format in
> coronal (using "unc2analyze") and axial (using "coronal2axial") planes.
> After that, I've converted both Analyze images into MNC format using
> "ana2mnc".
>
> I've compared the axial and coronal Analyze images and checked that the
> signal intensities were the same in a few randomly picked voxels. I couldn't
> check the MNC images properly because of difficulty selecting exact same
> voxels in both images (using MNC viewing software like "Display" and
> "register"), but I couldn't tell any visual difference between the axial and
> coronal MNC images.
>
> Anyway, thank you very much again for your advice. For now, I'll just stick
> to the images in the original (i.e. coronal) plane.
>
> Best wishes,
>
> Kie Woo
>
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Sun, 18 Mar 2012 18:07:21 +0000
> From: "Nam, Kie Woo" <kie_woo.nam at kcl.ac.uk>
> Subject: [MINC-users] Best plane (axial or coronal) for CIVET
> processing?
> To: "minc-users at bic.mni.mcgill.ca" <minc-users at bic.mni.mcgill.ca>
> Cc: "Simmons, Andy" <andy.simmons at kcl.ac.uk>
> Message-ID:
> <194921E98A6CEC41B0C14C804A56CC90CFF0B4 at DBXPRD0310MB382.eurprd03.prod.outlook.com>
>
> Content-Type: text/plain; charset="Windows-1252"
>
> Dear experts,
>
> I know this mailing list is not particularly for CIVET pipeline, but could
> someone please advise me the best plane (axial or coronal) for images to be
> in for CIVET processing?
>
> I've recently run CIVET on an image in coronal and axial planes (i.e. 2
> identical images just in different planes), and noticed that the outputs
> differed. This was noticeable in the "clasp" images as well as the text
> files (example shown below).
>
> ...brainmask_qc.txt
>
> Axial image: native skull mask in stx space (10.83%)
>
> Coronal image: native skull mask in stx space (11.96%)
>
> My original images were in coronal plane, so I'm thinking of doing CIVET
> processing on coronal images. However, I'm not convinced whether this is a
> good choice. So, if anyone would please advise me on this, I would much
> appreciate it.
>
> Best wishes,
>
> Kie Woo
>
> Kie Woo Nam
> Postgraduate Research Student
> Division of Psychological Medicine
> Department of Psychosis Studies
> Institute of Psychiatry, King?s College London
> De Crespigny Park, Denmark Hill
> London SE5 8AF, UK
> Tel: +44 (0)20 7848 0061
>
>
> ------------------------------
>
> Message: 2
> Date: Sun, 18 Mar 2012 17:43:20 -0400
> From: Alex Zijdenbos <zijdenbos at gmail.com>
> Subject: Re: [MINC-users] Best plane (axial or coronal) for CIVET
> processing?
> To: MINC users mailing list <minc-users at bic.mni.mcgill.ca>
> Message-ID:
> <CABEEk15OETNOP61yqqVbDn0zTRxLOub16FsrJ3nB9ETEtc8utw at mail.gmail.com>
> Content-Type: text/plain; charset=windows-1252
>
> Hello Kie Woo,
>
> Are you referring to two images _acquired_ in different orientations
> (i.e., two different acquisitions), or two images simply (re)sampled
> (or reshaped) to have different dimension ordering? In the latter case
> you really should obtain identical results (unless you changed voxel
> sizes or manipulated the data in some other way); so I assume you are
> talking about the former, i.e., you have two different acquisitions.
> In this case you would expect to get slightly different results -
> there is an inherent scan-rescan variability associated with any type
> of image processing tool, simply because the image you provide is a
> different one. In other words, if you would take two axial or two
> coronal scans acquired in sequence, you would also see a difference in
> image processing results.
>
> The only reason I can think of why coronally scanned data would yield
> more or less accurate results as compared to axial images is if there
> is a difference in the image quality, in terms of intensity uniformity
> for example. Otherwise I would say the results should be equivalent.
> Being consistent however is always good; for example, I would
> definitely not change your acquisition protocol in the middle of a
> longitudinal study, if that is what you are contemplating.
>
> -- Alex
>
> On Sun, Mar 18, 2012 at 2:07 PM, Nam, Kie Woo <kie_woo.nam at kcl.ac.uk> wrote:
>> Dear experts,
>>
>> I know this mailing list is not particularly for CIVET pipeline, but could
>> someone please advise me the best plane (axial or coronal) for images to be
>> in for CIVET processing?
>>
>> I've recently run CIVET on an image in coronal and axial planes (i.e. 2
>> identical images just in different planes), and noticed that the outputs
>> differed. This was noticeable in the "clasp" images as well as the text
>> files (example shown below).
>>
>> ...brainmask_qc.txt
>>
>> Axial image: native skull mask in stx space (10.83%)
>>
>> Coronal image: native skull mask in stx space (11.96%)
>>
>> My original images were in coronal plane, so I'm thinking of doing CIVET
>> processing on coronal images. However, I'm not convinced whether this is a
>> good choice. So, if anyone would please advise me on this, I would much
>> appreciate it.
>>
>> Best wishes,
>>
>> Kie Woo
>>
>> Kie Woo Nam
>> Postgraduate Research Student
>> Division of Psychological Medicine
>> Department of Psychosis Studies
>> Institute of Psychiatry, King?s College London
>> De Crespigny Park, Denmark Hill
>> London SE5 8AF, UK
>> Tel: +44 (0)20 7848 0061
>> _______________________________________________
>> MINC-users at bic.mni.mcgill.ca
>> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>>
>
>
> ------------------------------
>
> _______________________________________________
> MINC-users mailing list
> MINC-users at bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>
>
> End of MINC-users Digest, Vol 80, Issue 9
> *****************************************
>