iMath TruncateImageIntensity segfaults on certain input

[Philip Cook]

As reported in ANTsX/ANTsPy#597

To Reproduce

See ANTsX/ANTsPy#597 for data and code

System information (please complete the following information)

• OS: Mac
• OS version: 12.7
• Type of system: Desktop
• CPU architecture: x86

ANTs version information

• ANTs code version: 2.5.1.post24-g6c29d9d
• ANTs installation type: Compiled from source

[Philip Cook]

@stnava this appears to be an error with images that have negative intensities. The range of your T1w is [-32768, -31932]. The seg fault goes away if I multiply it by -1.

[Philip Cook]

The mask step will not contain any voxels in an image that is all negative. Should I change it to default to use the full range of pixel type?

ANTs/Examples/iMathFunctions1.hxx

Lines 622 to 630 in 2236a75

 typedef itk::BinaryThresholdImageFilter<ImageType, MaskType> ThresholdFilterType; 
 typename ThresholdFilterType::Pointer                        thresh = ThresholdFilterType::New(); 
 thresh->SetInput(image); 
 thresh->SetLowerThreshold(1e-6); 
 thresh->SetUpperThreshold(itk::NumericTraits<PixelType>::max()); 
 thresh->SetInsideValue(1); 
 thresh->SetOutsideValue(0); 
 thresh->Update(); 
 mask = thresh->GetOutput(); 

[stnava]

thanks @cookpa for searching this out