LoadImage support for TIFF files

[bhashemian]

Is your feature request related to a problem? Please describe.
Currently due to the extra consideration needed for TIFF files of histopathology WSI images in Camelyon dataset (huge images), we were using PatchWSIData to load these images. However, with other datasets like PANDA dataset for prostate cancer, the size of images allow for loading the entire image into memory. Thus, it might be beneficial to load such images by just using LoadImage transform similar to radiology use cases.

[Nic-Ma]

Hi @drbeh ,

I think LoadImage should already support TIFF based on ITK, could you please help try it?
If facing any issue, please let me know.

Thanks.

[bhashemian]

Thanks @Nic-Ma! That'd be great! Will try it.

[myron]

@Nic-Ma
I tried LoadImageD for .tiff (WSI) file

Initially I got error
RuntimeError: can not find a suitable reader for file: ('/PandaChallenge2020/train_images/17cedc9620be373816ca026a94e0d6c3.tiff',). Please install the reader libraries, see also the installation instructions: https://docs.monai.io/en/latest/installation.html#installing-the-recommended-dependencies. The current registered: [<monai.data.image_reader.ITKReader object at 0x7f3792d803a0>, <monai.data.image_reader.NumpyReader object at 0x7f3792d80760>, <monai.data.ima ge_reader.PILReader object at 0x7f3792d80490>, <monai.data.image_reader.NibabelReader object at 0x7f3792d80c70>, <monai.data.image_reader.ITKReader object at 0x7f3792d80 e20>]

which is confusing, it says I have ITKReader registered, but I didn't have itk installed. (we need a better error message)

After intalling (pip intall itk)

I get and image (16384, 17664, 4) float32
but it should be (16384, 17664, 3) uint8

3 channels (not 4) and uint8 (not float32). This works correctly with other readers such as tiffile

is there a way to fix it? thanks

[Nic-Ma]

Hi @myron and @drbeh ,

Thanks for your feedback.
I am working on this ticket now to improve the image reader.
And BTW, I think PILImage is a better reader for your case? I will change the reader for TIFF.
For example:

from PIL import Image
import numpy as np

img = np.as_array(Image.open("/XXX/image1.tiff"))

Thanks.

[myron]

PIL can't read all tiff images (mine are multi page WSI images), but tiffile can
Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/opt/conda/lib/python3.8/site-packages/PIL/Image.py", line 3030, in open raise UnidentifiedImageError( PIL.UnidentifiedImageError: cannot identify image file '/train_images/340236fc3587d38939e3dd669bbfe3e8.tiff'

[bhashemian]

Hi @Nic-Ma, thanks for taking care of this. I think it's better to have either tifffile or WSIReader (which uses cuCIM or OpenSlide) to handle TIFF files. What do you think?