LesionFilling error

[enrit]

Hi All,

I have segmented MS lesions and as a next step, I am using LesionFilling to 'paint' the lesions on a T1 image.

However, I am encountering the following errror -

LesionFilling 3 T1.nii.gz Lesion.nii.gz test.nii.gz
libc++abi: terminating with uncaught exception of type itk::ExceptionObject: /Applications/ANTs/build/staging/include/ITK-5.2/itkConnectedComponentImageFilter.hxx:145:
itk::ERROR: itk::ERROR: ConnectedComponentImageFilter(0x7fb53a605330): Number of objects (258) greater than maximum of output pixel type (255).
Abort trap: 6

Can you suggest how do I fix it ?

Many thanks

Rito

[ntustison]

I'm guessing you have more than 255 separate "lesions" in your mask which means that for the output image type (char?) specified in the LesionFilling program, that's causing the connected components filter to throw the exception. For the short term, you might want to run your mask through LabelClustersUniquely to return only 255 lesions and then run LesionFilling on that pruned mask. A longer term solution would probably be to change the output image type in LesionFilling if the shorter term solution works as expected.

[BennoGesierich]

Hi All,

I encountered the same bug: With more than 255 lesion clusters in the (binary) input lesion mask, I get following error:

terminate called after throwing an instance of 'itk::ExceptionObject'
what(): /opt/build/staging/include/ITK-5.3/itkConnectedComponentImageFilter.hxx:144:
ITK ERROR: ConnectedComponentImageFilter(0x2fe50b0): Number of objects (257) greater than maximum of output pixel type (255).

However, I would assume that the problem is not the data type of the "output image", but instead the data type of the image internally used to label the lesions in the input mask.
Instead, the only output image provided is the lesion filled grey scale image, which is of type float64.

I tried to provide the input lesion mask as uint16, but this didn't help.

Would you be able to change the data type used internally for labelling the lesions in the input lesion mask to something larger than char, e.g. uint16?

Thank you very much, in advance!
Benno

[cookpa]

Working on it now - you might also be interested in the inpainting utility in ANTsPyNet

https://github.com/ANTsX/ANTsPyNet/blob/master/antspynet/utilities/inpainting.py