deroneriksson / python-wsi-preprocessing Goto Github PK

I am using this code for processing Whole Slide Images (.svs format). This code running well when NUM_TOP_TILES = 50 to NUM_TOP_TILES = 12000. As Whole Slide Images are very large in size, in that case if I increase NUM_TOP_TILES>12000, I'm getting OverflowError: size does not fit in an int error. This is very common issue with Pillow when large images were used. There are many solutions are available in the github and StackOverFlow but I'm not able to solve my issue. As per my analysis,Image.fromarray() of Pillow in the below function causing this issue. It will be a great help if anyone suggest any solution to solve this issue. I have given related function and error traceback for your reference.

Function:

def np_to_pil(np_img):
  """
  Convert a NumPy array to a PIL Image.
  Args:
    np_img: The image represented as a NumPy array.
  Returns:
     The NumPy array converted to a PIL Image.
  """
  if np_img.dtype == "bool":
    np_img = np_img.astype("uint8") * 255
  elif np_img.dtype == "float64":
    np_img = (np_img * 255).astype("uint8")
return Image.fromarray(np_img)

Error Traceback

Traceback (most recent call last):

File "WsiToImagePatch.py", line 35, in np_to_pil return Image.fromarray(np_img) File "/usr/local/lib/python3.5/dist-packages/PIL/Image.py", line 2536, in fromarray return frombuffer(mode, size, obj, "raw", rawmode, 0, 1) File "/usr/local/lib/python3.5/dist-packages/PIL/Image.py", line 2479, in frombuffer return frombytes(mode, size, data, decoder_name, args) File "/usr/local/lib/python3.5/dist-packages/PIL/Image.py", line 2412, in frombytes im.frombytes(data, decoder_name, args) File "/usr/local/lib/python3.5/dist-packages/PIL/Image.py", line 812, in frombytes s = d.decode(data) OverflowError: size does not fit in an int

Hello!
Is there any option where we can rejoin the tiled images?
I am feeding the tiled images to a CNN (input size is small) and recombine each processed tiles to original image.

Hi @deroneriksson

I have gone through your blog and found it very useful. However I am facing issues regarding the WSI Level for Leica .scn images. Whenever i use the function slide_info() from the openslide package to get information about he dimensions, it gives incorrect information. The information doesn't seem to match the information from the image.

I am hereby attaching the details of 2 WSI and their information from the slide_info() function and from the actual WSI. The width and height at each and every level dont seem to match at all. As a result I get tiles with huge y-coordinates and few tiles are completely black. For two different WSI , openslide slide_info() function gives same level values and same dimensions which is not possible at all.

Request you to kindly inspect the issue and let me know about it, I am unable to proceed with my research. What is the workaround for such issues?

Regards,
Hitesha

nims-002.txt

nims-004.txt

In tile_extraction/filter.py, in line 286, sk_morphology.remove_small_holes is called with the parameter name min_size.
This probably worked for earlier versions of the sk_morphology function, but it now raises an error, as the function now expects the parameter area_threshold, which is probably equivalent with what min_size once meant.
-> filter.py needs to be adapted