📢Update on the Testing Phase progress: Fixing OpenSlide issues

Dear participants,

We were able to reproduce the error and find the solution. We are able to process slides with Ubuntu 20.04, openslide 1.3.1, openslide-python 3.4.1.

We recommend using these versions. We were not yet able to find solutions with other versions.

If your container was getting an error "Not a JPEG file starts from 0x00..." You can now resubmit it to both test and sanity check phase - it should work now.

I will follow up on possibility of extending deadline for those who already faced issues with test files shortly. But if you can submit today, please do so.

Looking forward to your submission,

Kind regards,

Khrystyna

Hi, maybe it would help to re-save those slides with ASAP (or pyvips if it works).

The following Dockerfile, along with the requirements.in file, should be error free. We looped over the 2 slides some people are facing issue with and were able to successfully open each tile at the 3 highest levels (0, 1, 2).

Link to files

Dear participants,

Also, find the https://github.com/DIAGNijmegen/LEOPARD-challenge-baseline/tree/main/docker with ASAP installed.

Kind regards, Khrystyna

In case you want to use the Docker file with ASAP installed and switch to wholeslidedata to read WSIs, I've put hereunder the equivalence between openslide & wholeslidedata when reading a (512, 512) tile in level 0. Please note the use of center=False when using wholeslidedata!

openslide

import openslide

wsi = openslide.OpenSlide(filepath)
width, height = 512, 512
patch = wsi.read_region((x, y), 0, (width, height))

wholeslidedata

import wholeslidedata as wsd
from PIL import Image

wsi = wsd.WholeSlideImage(filepath, backend="asap")
target_level = 0
target_spacing = wsi.spacings[target_level]
width, height = 512, 512
patch = wsi.get_patch(x, y, width, height, spacing=target_spacing, center=False)
patch = Image.fromarray(patch).convert("RGB")

One team who was facing issues was able to submit to the test set using this base docker .