Dear Marko and Stef,
Thank you both so much for your help and suggestions.
Marko - I will send you a separate link to the TPMs etc. It would be great if there was a new version floating around already that sorts this out. I did just check the latest available and it precedes December 5th so it doesn’t
look like it’s up publicly yet.
Steff - that’s v interesting. Though moving forward rather than backwards in version number will (hopefully!) be the most likely eventual solution, I wouldn’t mind fiddling around with the old one anyway
☺ I’ve had a look and can’t see an archived version for 12.3 anywhere. Any chance you still have a copy somewhere you could send/link me to?
Super thanks again to both of you and fingers crossed one of these versions sorts this out!
Arjun
------------------------------------------------------------
Arjun Sethi, PhD
Postdoctoral Research Associate
Forensic & Neurodevelopmental Sciences
Institute of Psychiatry, Psychology & Neuroscience
King’s College London
16 De Crespigny Park
London SE5 8AF
Tel: +44(0)207 848 5267
From: Stef Meliss <[log in to unmask]>
Date: Wednesday, 16 January 2019 at 10:01
To: Marko Wilke <[log in to unmask]>, "[log in to unmask]" <[log in to unmask]>, "Sethi, Arjun" <[log in to unmask]>
Subject: Re: [SPM] Problem with CAT12 segmentation using custom Cerebromatic TPMs
Dear Arjun, dear Marko,
To add to this: I came across the same issue when segmenting data of an older population. Interestingly, I have only encountered this error message when using CAT12.5, but not in version 12.3, although I have used the same TPM, participant
data, and parameters in both versions.
Do you have any thoughts on this?
Best regards,
Stef
___________
Stef Meliss
Doctoral Researcher
Motivation Science Lab
School of Psychology and Clinical Language Sciences
University of Reading
RG6 6AL (Sat Nav: RG6 7BE)
+44 (0) 118 378 7937
From: "SPM (Statistical Parametric Mapping)" <[log in to unmask]> on behalf of Marko Wilke <[log in to unmask]>
Reply-To: Marko Wilke <[log in to unmask]>
Date: Wednesday, 16 January 2019 at 09:31
To: "[log in to unmask]" <[log in to unmask]>
Subject: Re: [SPM] Problem with CAT12 segmentation using custom Cerebromatic TPMs
Hello Arjun,
sorry to hear that you are having issues with the COM toolbox.
I have had feedback from some people that the generated TPMs are not
working well, but that was usually due to people assessing their own
small population using the "Estimate regression parameters" without
changing the estimation parameters (which are designed for a huge
sample). This does not seem to be the case here.
Also, if you checked the TPM for overall appearance, and they work when
fed to spm12, they seem to generally work alright (which is a relief to
me ;) If you want, you can send me the file off-list and I can look at them.
My best guess would be that, as indicated in the error message, the
affine registration is the issue. Not surprisingly, children's brains
are different (the whole point of COM), so too stringently using the
affine parameters from an adult population may bring about trouble. I
have had a similar failure a while ago, and the solution as usual came
from Christian with a new, more robust version of the affine estimation
part of CAT12. I have not seen this error since updating (on December
5th; I am not sure these changes are already available in the public
version).
Also, w.r.t. your bonus question, COM will also generate SHOOT templates
for you if you so desire. They are actually quite similar to DARTEL ones
but contain an extra 1-sum(1:n) tissue class volume. If you are
paranoid, you could generate both DARTEL and SHOOT templates using COM
(using the same parameters, of course) and supply both to CAT12.
Hope this helps
Marko
Sethi, Arjun schrieb:
Dear SPMers,
I’ve been bashing my head against this one for a while now, and thought
I’d ask the hive mind.
I’m analysing data from a study of kids (mean age around 8.6 years
[range: 5-11]) with CAT12 and making TPMs and dartel templates with COM.
The problem is I keep on getting an ‘Empty segmentation error’ (see
underneath the main body of this email, or attached error file; Versions
are: Matlab r2017a, SPM 7487, com 1.1, CAT12.5 r1364). This only occurs
when I’m using the TPMs created using COM. The segmentation actually
works fine with the standard SPM TPMs in CAT12, and the custom COM TPMs
seem to produce a reasonable segmentation when using the standard SPM
segment module. In short, it seems like there is nothing fatally wrong
with either component alone (CAT12, COM TPMs), but they do not seem to
want to play nice together.
Here are a few troubleshooting steps that I have run so far:
Within CAT12:
1. I have ensured the origin is appropriately close to the AC on the
T1w (moving) image and checked that this lines up well with the TPMs
2. I have fiddled with pretty much every available GUI option to no
avail (Affine Regularisation method, Strength of inhomogeneity
correction, Affine preprocessing, skull stripping method)
3. In addition to the above points re: the specificity of this problem
to CAT12 and the COM TPMs, I thought I’d also mention that the
custom DARTEL templates from COM also work fine in CAT12 segmentation.
4. The study is longitudinal in nature, but this error occurs
regardless of whether I use the regular or longitudinal segmentation
options. I’m doing most of my troubleshooting with the
cross-sectional segment to keep things simple.
Within COM (using unified segmentation regression parameters):
1. I’ve tried with a variety of age sets as input to make sure there
was no unexpected behaviour associated with a specific combination.
Beyond that all our data are from male subjects collected at 3T, so
these are the only changes to regression parameters I have tried)
2. I’ve tried using the affine only regression parameters, but still no
dice.
3. I noticed that the final COM TPM is visually quite different (more
‘detail’ within the brain, which appears as greyscale rather than
just zeros) from the default SPM TPMs, and surmised this is because
of how the COM TPMs are calculated (which ensures that TPM values
sum to one, and adds the remainder to the last tissue class). When I
change this so that the remainder isn’t added (and the final TPM is
more visually similar to the SPM default), there was still no change.
If anyone has any input it would be hugely appreciated as I’m a little
stumped at this point. It’s looking more and more to me like there is
nothing wrong with the segmentation or affine registration (as SPM does
this just fine, it would seem), but some kind of write error or
orientation issue that isn’t immediately obvious or observable (wrt the
latter, I’ve also checked the header info of the SPM and COM TPMs and
there were no differences that stuck out to me). I’ve noticed there have
been a few empty segmentation errors reported about this on the board
but I haven’t seen any resolution applicable here.
Best wishes,
Arjun
PS I also have a bonus (☺), hopefully much simpler question: - I just
wanted to confirm that when the DARTEL registration option is selected,
that the shooting template within the GUI is not used. I notice that the
DARTEL does run some ‘fast shooting’ and just wanted to clarify this
does not rely on the shooting template input.
------------------------------------------------------------
Arjun Sethi, PhD
Postdoctoral Research Associate
Forensic & Neurodevelopmental Sciences
Institute of Psychiatry, Psychology & Neuroscience
King’s College London
16 De Crespigny Park
London SE5 8AF
Tel: +44(0)207 848 5267
------------------------------------------------------------------------
15-Jan-2019 12:12:27 - Running job #10
------------------------------------------------------------------------
15-Jan-2019 12:12:27 - Running 'CAT12: Segmentation'
Warning: Dartel template was changed: Please be aware that ROI analysis
and other template-specific options cannot be used and ROI
output has been deselected.
> In cat_run>update_job (line 331)
In cat_run (line 159)
In cfg_run_cm (line 29)
In cfg_util>local_runcj (line 1717)
In cfg_util (line 972)
In cfg_ui>MenuFileRun_Callback (line 710)
In gui_mainfcn (line 95)
In cfg_ui (line 53)
------------------------------------------------------------------------
CAT12.5 r1364: 1/1: ./04_T1/ABC002A/anat9/ABC002A_4903_run1_T1w.nii
------------------------------------------------------------------------
SANLM denoising (medium): 16s
Internal resampling (1.20x1.05x1.05mm > 1.00x1.05x1.05mm): 2s
APP: Rough bias correction:
Initialize 5s
Estimate background 32s
Initial correction 15s
Refine background 32s
Final correction 12s
Background correction 0s
Final scaling 1s
99s
Coarse affine registration 11s
Affine registration 20s
SPM preprocessing 1 (estimate 1): 32s
SPM preprocessing 1 (estimate 2): 119s
SPM preprocessing 1 (estimate skull-stripped): 104s
SPM preprocessing 1 (estimate skull-stripped): 105s
SPM preprocessing 2 (write):
Write Segmentation 18s
Update Segmentation
------------------------------------------------------------------------
CAT Preprocessing error for ABC002A_4903_run1_T1w:
------------------------------------------------------------------------
Empty Segmentation:
Possibly the affine registration failed. Pleace check image orientation.
Tissue class: BG CSF GM WM
HDH HDL
Rel. to image volume: 33.70 0.00 0.00 0.00
37.86 28.43
Rel. to brain volume: Inf NaN NaN NaN
Inf Inf
Tissue intensity: 0.47 748.77 1522.66 2296.56
1986.58 1138.68
------------------------------------------------------------------------
216 - cat_main
721 - cat_run_job1070
16 - cat_run_newcatch
398 - run_job
161 - cat_run
29 - cfg_run_cm
1717 - local_runcj
972 - cfg_util
710 - MenuFileRun_Callback
95 - gui_mainfcn
53 - cfg_ui
------------------------------------------------------------------------
Warning:cat_io_report:CATgui: Incomplete report creation in
cat_io_report because of incomplete CAT parameters.
Print 'Graphics' figure to:
/data/project/abc/03_Sorted/04_T1/ABC002A/anat9/report/catreport_ABC002A_4903_run1_T1w.pdf
15-Jan-2019 12:21:51 - Failed 'CAT12: Segmentation'
Error using cat_main (line 216)
Empty Segmentation:
Possibly the affine registration failed. Pleace check image orientation.
Tissue class: BG CSF GM WM
HDH HDL
Rel. to image volume: 33.70 0.00 0.00 0.00
37.86 28.43
Rel. to brain volume: Inf NaN NaN NaN
Inf Inf
Tissue intensity: 0.47 748.77 1522.66 2296.56
1986.58 1138.68
In file "/data/project/abc/MatlabPath/spm12/toolbox/cat12/cat_main.m"
(v1361), function "cat_main" at line 216.
In file
"/data/project/abc/MatlabPath/spm12/toolbox/cat12/cat_run_job1070.m"
(v1348), function "cat_run_job1070" at line 721.
In file
"/data/project/abc/MatlabPath/spm12/toolbox/cat12/cat_run_newcatch.m"
(???), function "cat_run_newcatch" at line 16.
In file "/data/project/abc/MatlabPath/spm12/toolbox/cat12/cat_run.m"
(v1332), function "run_job" at line 398.
In file "/data/project/abc/MatlabPath/spm12/toolbox/cat12/cat_run.m"
(v1332), function "cat_run" at line 161.
The following modules did not run:
Failed: CAT12: Segmentation
--
____________________________________________________
Prof. Dr. med. Marko Wilke
Facharzt für Kinder- und Jugendmedizin
Leiter, Experimentelle Pädiatrische Neurobildgebung
Oberarzt der Abteilung Neuropädiatrie
Universitäts-Kinderklinik
Marko Wilke, MD, PhD
Pediatrician
Head, Experimental Pediatric Neuroimaging
Consultant in Pediatric Neurology
University Children's Hospital
Hoppe-Seyler-Str. 1
D - 72076 Tübingen, Germany
Tel. +49 7071 29-83416
Fax +49 7071 29-5473
____________________________________________________