Unable to upload the image volume, file too big. Use TCIABrowser extension to download LIDC-IDRI-00314 volume used in the illustration.

Hi Andrey,

We have encountered similar issue in the past regarding using different labelmap for statistics. the solution is to resample the labelmap according to input image.

We have created a utility function in SlicerRT for this purpose. see [1]. the intention is to integrate this into Slicer's Volumes module logic. if you think it is useful after you review it, then we can start the process to integrate it into trunk.

Thanks,

Kevin

[1] https://www.assembla.com/code/slicerrt/subversion/nodes/1872/trunk/SlicerRt/src/SlicerRtCommon/vtkVolumesOrientedResampleUtility.cxx
[2] http://www.na-mic.org/Wiki/index.php/Common_resampling_and_conversion_utility_functions_in_Slicer
[3] https://www.assembla.com/spaces/slicerrt/tickets/303#/activity/ticket:

Kevin, thank you for adding this information, I was not aware of that project. I think the right thing is to get Steve on board, since he is the author of LabelStatistics and has a vision what is the right thing to do for the trunk. I know Steve added error checks to LabelStatistics to ensure orientation is consistent, but I am not sure if he would prefer to add the resampling function. I will assign to Steve.

Yes, it seems logical to add this resampling directly to the LabelStatistics module and invoke if if the geometries are different.

I had used the brains resample cli with an identity transform to fix this when it came up for Andrey, but I agree it would be better for users if it happened automatically.

I don't have plans to work on LabelStatistics right now, but probably will later when the qiicr pet/ct metrics are ready to integrate. If someone else wants to make a patch and pull request I'd be happy to review it.

Hi Steve,

Thanks for your info. I was not aware of using brains resample cli for this purpose. seems you have already found a solution.

so now there are 2 options:
option 1: transfer the resampling code (based on VTK) in SlicerRT to the Volumes module logic class.
option 2: using brains resample cli (based on ITK) in labelstatistics module logic class.

option 2 requires minimal changes to slicer trunk which seems to be less work and requires less maintenance.

option 1 is good if this utility function is required in a lot of c++ loadable modules then direct access is more convenient.

so which one do you guys prefer? once decided, I can work on the solution and create a pull request.

Thanks,
-Kevin

Hi Kevin - I agree that your option 1 is a more general solution - I think the issue comes up often enough that people would want to use it for various purposes (I've written such a utility a couple times myself and never put it in volumes logic -- shame on me!).

But if you would consider doing the integration using your existing SlicerRT code that would be great.

Hi Steve,

Thanks for the comments. I will start to work on a topic branch and ask you to review it once it is ready.

Kevin

Topic pushed to https://github.com/kevinwangcanada/Slicer/tree/3753-LabelStatistics-reports-incorrect-stats

Can you guys review it please?

Thanks,

Kevin

I added some comments on the commit -- generally looks good though.

Thanks Steve for the quick review. I have changed code accordingly. in summary, I added a new method to clone the volume without copy the imagedata as I did not want to change the API. if it works ok, then we can discuss how to refactor the code to have a single clonevolume logic.

I have pushed the code to github: https://github.com/kevinwangcanada/Slicer/tree/3753-LabelStatistics-reports-incorrect-stats

committed in r23491 and r23492.

Thanks Kevin!

Thanks Steve for quick integration.

Hi Steve,

sorry there is a mistake in the patch. I forgot to declare the method static. could you please fix that?

Thanks,

Kevin

Hi Steve,

sorry there is a mistake in the patch. I forgot to declare the method static. could you please fix that?

Thanks,

Kevin

Done: http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&revision=23534

thanks for the quick fix!

Thanks a lot for working on this! Sorry I am unable to test to confirm resolution and close the issue, since Slicer is unusable on Mac at the moment. Does not start up.