Issue in patch extraction from images with more than 3 channels

[Arwa-AlRubaian]

• TIA Toolbox version:1.3.3
• Python version: 3.10
• Operating System: Ubuntu

Description

I was using (patchextractor=patchextraction.SlidingWindowPatchExtractor) to extract patches on images with 6 channels. Then when iterating over the patches using :
for patch in patchextractor:
#(some code ...)

the following erroroccur:
Traceback (most recent call last):
for patch in patchextractor:
File "//site-packages/tiatoolbox/tools/patchextraction.py", line 167, in next
return self[n]
File "//site-packages/tiatoolbox/tools/patchextraction.py", line 179, in getitem
return self.wsi.read_rect(
File "//site-packages/tiatoolbox/wsicore/wsireader.py", line 2846, in read_rect
return self._read_rect_at_resolution(
File "//site-packages/tiatoolbox/wsicore/wsireader.py", line 970, in _read_rect_at_resolution
return self.read_bounds(
File "///site-packages/tiatoolbox/wsicore/wsireader.py", line 3057, in read_bounds
return utils.transforms.background_composite(image=im_region)
File "///site-packages/tiatoolbox/utils/transforms.py", line 41, in background_composite
image = Image.fromarray(image)
File "/****/site-packages/PIL/Image.py", line 3083, in fromarray
raise TypeError(msg) from e
TypeError: Cannot handle this data type: (1, 1, 6), |u1

[John-P]

We do not currently have support in place for multi-channel images as we have been focused on RGB bright-field imaging. However, we would love to add support for this and getting patch extraction working for multi-channel images would be a good start. I think the trickiest part would be to define how patches get written out to disk. Do you have a preferred format for this? Would a multi-channel TIFF or npy file be suitable for your needs?

[rogertrullo]

Maybe this should be left up to the user? as long as the patch_extractor returns a np array e.g. H,W,num_chans it should be good?

[shaneahmed]

• Allow the user to select the background colour for background_colour = [1 1 1].
• Instead of mode, alpha_channel = None if not background composite required, otherwise alpha_channel = -1 by default.
• background_composite is defined here

  tiatoolbox/tiatoolbox/utils/transforms.py

  Line 11 in 1598970

  def background_composite(