Hi all,

I'm trying to run StatNonParamTestPDM and I get the following error:
.....
Error while reading number, encountering non-number on line 103 feature: 3 numChar : 0
Error while reading number, encountering non-number on line 103 feature: 3 numChar : 0
Error while reading number, encountering non-number on line 103 feature: 3 numChar : 0
Error while reading number, encountering non-number on line 103 feature: 3 numChar : 0
data has been relabeled: -1 --> 0 ; -1 --> 1
#A: 104, #B: 0
tupel: 1 , numFeatures: 0
group sizes: 104/0 = 104
Features : 0 , real: 0
.Segmentation fault (core dumped)

I use the following command:

./StatNonParamTestPDM /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/meshList.txt -out /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/first_StatNonParamTestPDM_result -numPerms 20000 -signLevel 0.05 -signSteps 1000 -v

I've used the latest Shape Analysis extension in Slicer to produce .vtk files and I've converted them to .meta files as suggested here. I've followed the same format as the example given the Docs section here. This is part of my file:

#groupID scalingFactor meshfile
0 1 /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/all_meshes/C76_LH_pp_surfSPHARM_procalign.meta
0 1 /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/all_meshes/C77_LH_pp_surfSPHARM_procalign.meta
0 1 /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/all_meshes/C78_LH_pp_surfSPHARM_procalign.meta
0 1 /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/all_meshes/C79_LH_pp_surfSPHARM_procalign.meta
1 1 /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/all_meshes/NPI001_LH_pp_surfSPHARM_procalign.meta
1 1 /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/all_meshes/NPI002_LH_pp_surfSPHARM_procalign.meta
1 1 /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/all_meshes/NPI003_LH_pp_surfSPHARM_procalign.meta
1 1 /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/all_meshes/NPI004_LH_pp_surfSPHARM_procalign.meta
1 1 /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/all_meshes/NPI005_LH_pp_surfSPHARM_procalign.meta

Does anyone know what might be wrong?! Thanks a lot,
-Mojdeh

Hi again,

I think I figured my mistake, I was missing the "-surfList" flag after my input file. Now, the when I run the following command:

./StatNonParamTestPDM /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/meshList.txt -surfList -out /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/first_StatNonParamTestPDM_result -numPerms 20000 -signLevel 0.05 -signSteps 1000 -v


I see the list of subjects, but again I get an error, this time:
...
...
/data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/all_meshes/TLE_P5_LH_pp_surfSPHARM_procalign.meta 1 1
/data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/all_meshes/TLE_P6_LH_pp_surfSPHARM_procalign.meta 1 1
/data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/all_meshes/TLE_P7_LH_pp_surfSPHARM_procalign.meta 1 1
/data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/all_meshes/TLE_P8_LH_pp_surfSPHARM_procalign.meta 1 1
0 0
CellType required and not defined.
MetaObject: Read: MET_Read Failed
MetaScene: Read: Cannot open file
Error reading meshfile:
ITK error: itk::ERROR: SpatialObjectReader(0x17acc30): No groups were found in file

Could someone please tell me what's wrong? Thank you very much,
-Mojdeh

Hi Mojdeh,



It seems an I/O problem with your mesh files. Could you please provide more information on how did you create the *procalign.meta meshes?
Thanks!

Beatriz

Hi Mojdeh

The data needs to be in meta format. Did you convert the vtk data to meta (with VTK2Meta) or simply rename them? The error indicates that there were no surfaces in the file or that the file (e.g. /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/all_meshes/TLE_P5_LH_pp_surfSPHARM_procalign.meta) does not exist.


Martin

Thanks Martin and Beatriz for the reply!

Yup, that's how I converted the vtk files, using VTK2Meta:

mydir=$(ls -d *_surfSPHARM_procalign.vtk)

for f in $mydir; do echo $f; ./VTK2Meta /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Mesh/SPHARM/$f /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/${f%.vtk}.meta; done

Hi Mojdeh,


That is surprising. Could you please share one of your meta files?
Thanks,


Beatriz

And is the path in the file correct, i.e. points to a file that you have read access to.

Martin

Hi again,

Martin. I checked the files again, it's the correct path and it's read/write/executable by everyone! ... I even converted back one of the meta files to vtk and can view it with the ShapePopulationViewer !!!

I have attached one of my .meta files.

Thanks,
-Mojdeh

Hi again,

Just wondering if anyone has an explanation for what's wrong with my stat analysis? I really appreciate your advice!

Thanks,
-Mojdeh

I looked at the meta file, looks all okay. I can visualize it.

So, I don't think you have an issue with the file. Do you have by any chance empty lines in the file? What does the first line in the file look like? 

Try to remove everything from the file except for the lines that actually contain information.

Martin

Hi Martin,

Thanks so much for the reply! I don't see any empty lines! I used the example .txt file to create to my meshlist file. Could you take a look at the text list to see if there is anything wrong with it?

-Mojdeh

your meshList file somehow got corrupted. It starts with some wierd characters that you would not see in a standard editor:

if I do an "od -c " on your file, I get:
0000000 357 273 277 0 1 . 0 / d a t a / h
0000020 i p p o c a m p u s _ s h a p e
0000040 _ a n a l y s i s / S h a p e A
0000060 n a l y s i s _ r e s u l t s _
0000100 L H _ m i n u s 2 / S t a t s /
0000120 a l l _ m e s h e s / C 2 3 _ L
0000140 H _ p p _ s u r f S P H A R M _
0000160 p r o c a l i g n . m e t a \r \n

What you can see is that your file starts with 3 strange characters which messes up the reader. It's only the first line that seems to have this issue. What I would do it basically copy-paste the whole thing into a new text file and you hopefully should be fine. 

Martin

Thanks a lot Martin, I did as you suggested and now it seems the command is running, but I still get an error before it starts:

CellType required and not defined.
MetaObject: Read: MET_Read Failed
data has been relabeled: 0 --> group A = 0 ; 1 --> group B = 1
#A: 36, #B: 67
this version does not support writing out covariance ellipsoid stats
tupel: 3 , numFeatures: 3006
group sizes: 36/67 = 103
Features : 3006 , real: 1002
........

should I ignore the first two lines?! what does this mean?

Not sure why it has the first error, maybe one of your surfaces does have a problem.

Are the numbers correct for the two groups? Or should there be one more surface?

Martin

weird! :( yes the numbers are fine, 103 in total (26/67 groups A/B)!

What would be the best way to check if the results are ok? is it possible to get this error and still have correct results at all?!

Hi Mojdeh,

What is the exact commandline you are using? From what software package?
Thanks,


Beatriz

Hi Beatriz,

This is the command I run:

./StatNonParamTestPDM /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/meshList.txt -surfList -out /data/hippocampus_shape_analysis/ShapeAnalysis_results_LH_minus2/Stats/first_StatNonParamTestPDM_result -numPerms 20000 -signLevel 0.05 -signSteps 1000 -v

and I've attached my new meshList file again.

I downloaded the software (StatNonParamTestPDM for linux64 (scientific linux)) from here: https://www.nitrc.org/frs/?group_id=308

Thanks,
-Mojdeh

Hi again,

Just to update: I figured out the problem, one of the meta files had a problem (didn't convert back to .vtk) which is weird because no error message was given when I initially converted the .meta file to .vtk! so I excluded that file and StatNonParamTestPDM seems to run without any errors. This is the output I got on the command line:

data has been relabeled: 0 --> group A = 0 ; 1 --> group B = 1
#A: 36, #B: 66
this version does not support writing out covariance ellipsoid stats
tupel: 3 , numFeatures: 3006
group sizes: 36/66 = 102
Features : 3006 , real: 1002
........................................
writing avg shape diff pvalue 0.0077
writing raw pvalues
writing statistic (T^2)
writing T statistic
writing Mean Diff statistic
computing FDR
computing Min distribution


However, now there seems to be another problem! (sorry for so many questions) ... When I want to convert the "meanAll.meta" file to .vtk to visualize I get the error:

CellType required and not defined.
MetaObject: Read: MET_Read Failed
Aborting: No Points in GRID

conversion works fine with meanA and meanB meta file thought! any thought on why?

Thanks,

-Mojdeh

Not sure, StatNonParamTestPDM is an older software that is not really maintained anymore and it could be that the overall average surface created is corrupt. 

You can though easily create it yourself by averaging the two group average surfaces using MeshMath

MeshMath meanA.vtk meanAll.vtk -avgMesh meanB.vtk

Martin

Hi Martin,

Thanks for the reply! The reason I was looking at the average images was to make sure everything went fine with the command. Maybe I try using shape MANCOVA instead, would you recommend that? is there a tutorial for that one that explains the outputs and how interpret them?

Thanks,
-Mojdeh

If you have no covariates (such as age or gender) to correct for, then StatNonParamTestPDM should be perfectly fine. Also, when the 2 group averages are okay, then you should be fine as well. Do the computed results make sense?

Martin

Hi again,

I have no covariates (for now) ... could you tell me how to view the text files (Tstat, corrPval, etc) on the averages in ShapePopulationViewer? or if there is any other way of viewing them? Sorry for too many questions, but could you also tell me which of these stat values are most commonly used for interpretation?

Thanks!

The text files that are generated go along with the "old style" visualization in KWMeshVisu (a tool that is no longer available/supported). But you can add the information in those text files to the surfaces via MeshMath and visualize them in ShapePopulationViewer.

MeshMath meanA.vtk meanA_withPval.vtk -KWMtoPolyData myPval.txt myPval 

The information in the generated files  and what kind of visualizations can be generated can be found in the insight journal article.

Usually you can use the rawP files for hypothesis generation, or the FDRP files for corrected hypothesis testing.

Martin

Thanks so much for the reply Martin!

I did as you suggested but I'm not sure if I'm getting the correct output! When I run the command (./MeshMath first_StatNonParamTestPDM_result_meanA.vtk first_StatNonParamTestPDM_result_meanA_withCorrPval.vtk -KWMtoPolyData first_StatNonParamTestPDM_result_corrPval.txt corrPval) I get a 0 as the output on the command line! and when view the generated vtk file it seems to be the exact same as the meanA file! Also, looking at the corrPval and rawPval files there seems to be 1002 numbers (all in one line), but the FDRPval is empty (only three lines:
NUMBER_OF_POINTS = 0
DIMENSION = 1
TYPE = Scalar) ... I've also attached these files. Could you take a look at them to see if they make sense or there was an error in the previous step?

really appreciate your help,
-Mojdeh

and the FDRPval file!