Simpleitk use - 4 Strange question

1. data dtype Type problem

1.1 Determine the cause

Phenomenon description

• windows Next use numpy Handle Nifti Image and save ,
  • The resulting image is 49KB
  • ITK-Snap Can open
• linux Use the same code under numpy Handle Nifti Image and save
  • When getting the image 85KB
  • ITK-Snap Open error report ：itk::ERROR: NiftiImageIO(000001CB27E1CD80): Unknown component type: 0
    
  • however 3D Slicer Can open

---

Find out why

• Because the code is the same , Tested and saved numpy The data are the same , But the final file size is different .
• So consider data types ,dtype The problem of

---

experiment

Because one sentence in the code is ：

skull_striper_array = np.where(pred_label_array - skull_mask_array == 1, 1, 0)

So check the data types one by one , Find out

pred_label_array.dtype
> dtype('uint8')

skull_mask_array.dtype
> dtype('uint8')

skull_striper_array.dtype
> dtype('int64')

Therefore, the consideration is due to numpy Constants are introduced in , Cause the data type to change .

skull_mask_path = "./skull_striper_mask.nii.gz"
skull_mask_image = sitk.ReadImage(skull_mask_path)
print(skull_mask_image.GetPixelIDTypeAsString())
> 64-bit signed integer

see linux The data type of the next image , Is, indeed, int64, To int32 after ,itk-snap You can open the processed image .

---

Consider different environments numpy Default data type for constants

• linux In the environment ：numpy-1.19.2、simpleitk-2.1.1.2

  test =np.array([1])
  test.dtype
  > dtype('int64')
  

• windows Environmental Science ：numpy-1.19.2、simpleitk-2.1.1.2

  	test =np.array([1])
  	test.dtype
  	> dtype('int32')
  

1.2 numpy The default dtype type

stay numpy file -Data types in , There are the following descriptions ：

NumPy Provide numpy.iinfo and numpy.finfo To confirm Numpy The maximum value of integers and floating-point numbers in .

np.iinfo(int) # Bounds of the default integer on this system.
iinfo(min=-9223372036854775808, max=9223372036854775807, dtype=int64)
np.iinfo(np.int32) # Bounds of a 32-bit integer
iinfo(min=-2147483648, max=2147483647, dtype=int32)
np.iinfo(np.int64) # Bounds of a 64-bit integer
iinfo(min=-9223372036854775808, max=9223372036854775807, dtype=int64)

So test on your own machine ：

• windows On ：
  
• linux On ：
  
• So you know , When I use the following code , After importing an integer

  skull_striper_array = np.where(pred_label_array - skull_mask_array == 1, 1, 0)
  

• windows The default integer （int） yes int32 position , and linux The default integer (int) yes int64 position

StackOverflow There is a similar problem on , See Specifying default dtype for np.array(1.)

1.3 itk-snap open int64 The problem of format

The phenomenon is described above ：

• 64 Bit image use ITK-Snap Open error report ：itk::ERROR: NiftiImageIO(000001CB27E1CD80): Unknown component type: 0
  
• however 3D Slicer Can open

So consider whether ITK-Snap I won't support it int64 Bitwise image . No valid relevant content has been found for the time being .
TBD